There was so much to digest during IonQ's Analyst Day but two things have become very clear. Under Niccolo deMasi and senior leadership the company has shifted from a collection of highly gifted, specialized quantum operators into an effective, efficient, and excellent oprating unit. In this sense, IonQ is no longer a quantum computing company. It has become the industrial base for the quantum industry.
The company demonstrated this on September 8 to a room that recorded the guidance number and left. The stack is closed — chips, processors, the interconnect that links them, sensing, security, satellites, and the foundry that fabricates the parts. Thirteen layers advanced in a single afternoon. None retreated. Four of the results announced are of a kind that independent layers cannot produce on their own; they exist only where the layers meet.
Nine witnesses spoke whose credibility originates entirely outside IonQ, and none of whom has an obvious reason to flatter it: a 2025 Nobel laureate whose own work is in a competing qubit modality, a sitting foreign minister, a former director of the National Geospatial-Intelligence Agency, a utility now running its own algorithms on the hardware, and a customer who reproduced an IonQ result with an independent team. Each addressed a different layer. Seven questioners then took the floor. Not one asked whether the physics works. That question has closed, and the room conducted itself as though it knew.
The afternoon's most consequential disclosure appeared on no slide. It was organizational.
IonQ was assembled at speed: twelve acquisitions closed in under three years, ten of them since Niccolò de Masi became chief executive in February 2025. The standing expectation for companies built this way is well founded and rarely wrong. Acquired teams retain their own cultures, their own roadmaps, their own engineering conventions. Work is duplicated. Integration costs arrive later and larger than underwritten. Federations of subsidiaries are the ordinary outcome because coherence is genuinely difficult to manufacture.
That expectation no longer describes this company. The teams IonQ bought contain some of the deepest talent alive in trapped ions, photonic interconnect, quantum networking, quantum sensing, space systems and semiconductor fabrication, and they now operate as a single unit against a single roadmap on a single cadence. The evidence is not a management assertion; it is the sequencing itself. Thirteen layers advanced together because one organization moved them together, and four of the results exist only because assets that belonged to different companies eighteen months ago are now engineered against one another. Integration risk was the largest unpriced liability on this balance sheet. It has been retired. It has not been marked.
The risk profile has shifted accordingly. What IonQ carried until recently was scientific risk — the possibility that the physics would never scale, which markets treat as unquantifiable and therefore price as fatal. That risk is discharged. What remains is ordinary industrial risk: yield, cadence, and commercial conversion. These are difficult problems. They are not mysterious ones, and they are borne by the only company that owns a quantum-capable foundry operating at scale, one already running two-thirds full on third-party wafers.
Which brings the discussion to the wrong opportunity. Qubit counts and a guidance raise are what the room took away. The real opportunity lies in the six intersections set out in Section 7 — products no single-layer competitor can construct at any price:
Cryptanalytic credibility that converts directly into remediation revenue. A foundry that earns regardless of which qubit modality ultimately prevails. Radar analysis that requires ownership of both the constellation and the computer reading it. Chip-scale sensing. A networked path into pharmaceutical chemistry. And an AI relationship already operating in both directions, with classical machine learning improving quantum system performance while quantum resources are directed at machine learning workloads.
Not one of these is available to a holding company. Each is a product of the unified engineering organization described above, which is the fact the market has not yet connected to the opportunity set.
Opening Observations, Strongly Supported by the Evidence
A note from the author, before the technical argument begins.
Investor days are usually about the future. A company tells you what it intends to build, you decide whether to believe it, and everyone goes home. That is not what happened at the New York Stock Exchange on September 8. IonQ spent nearly three hours describing things that already exist, already run, and in several cases already have customers — and it brought other people onto its stage to say so.
HOW TO READ THIS REPORT Observations 1–9, below — the case of what happened in plain language, no specialist knowledge assumed. Sections 1–10 — the argument, with every analytical claim marked FACT, INFER, ARG or UNDISC. Sections 11–27 — every speaker examined, 112 questions, with what each answer does for the stack stated underneath it. Section 9 — the specific things that would break this view. Appendix B — the full proceedings indexed under twenty-three categories, so any citation resolves. A reader who wants the case can stop after Part One. A reader who wants to break it should start at Section 11. |
What follows in this report is detailed and, in places, technical. This opening section is not. These are the observations a careful reader should take away even if they read nothing else, written plainly, with the evidence for each one sitting in the sections that follow.
1. IonQ stopped describing a plan and started describing a machine
A quantum computer is not one invention. It is a stack of them — the chip, the way the chip is manufactured, the way errors are caught and corrected, the software that turns a problem into instructions the chip understands, the wiring that connects one machine to another, and the applications customers actually pay for. Every company in this field is strong at some of those and weak at others.
On September 8 IonQ put working evidence on the record at every one of them, on the same afternoon, with dates attached. That has not happened before, at this company or any other. If you read one exhibit in this report, read the thirteen-layer table in Section 1 and then the layer-by-layer evidence in Section 4.
2. The strongest proof was a piece of software, and it is easy to explain
Thirty years ago a mathematician showed that a large enough quantum computer could break the encryption that protects most of the internet. Ever since, researchers have calculated how big such a machine would need to be. Those calculations are a serious and crowded field — Google’s Craig Gidney has published widely cited numbers, and a Sydney company called Iceberg Quantum published a lower one in February 2026 — and every one of them is written against an idealized machine. A square grid of qubits. A stated error rate. A stated cycle time. Assumptions, carefully declared.
IonQ wrote its version against a machine with a name. Not an idealized grid, but the Superion architecture it is building, running the error-correction scheme it published in April, on chips coming off the fab it now owns. Every operation, in order, down to the physical level: 19,397 qubits, running for about twenty-six days. And it did something the field mostly does not — in Ballance’s words, it proved rather than assumed a lower bound on the probability that the whole computation actually succeeds.
Every prior calculation of this kind describes an idealized machine. IonQ’s describes the one on its own wafers — and a machine you can compile against, operation by operation, has stopped being a research project. |
The qubit count is not the headline it looks like either, and it cuts the other way from how most coverage read it. Previous estimates for this same problem on a trapped-ion machine ran between 1.2 million and 9.4 million qubits. IonQ got it to 19,397 — sixty to five hundred times less — by optimizing the algorithm, the compiler, the architecture and the error correction together instead of one at a time. What matters is that the specification of IonQ’s own machine — Superion, Walking Cat, SkyWater wafers — is now complete enough to compile against, operation by operation. This is the observation the report rests on, and most people watching missed it because it looks like a cryptography story rather than an engineering one. It is an engineering one.
3. They published what they cannot break, alongside what they can
IonQ sells security products. A company in that position has every commercial reason to leave a threat vague and frightening. IonQ did the opposite. It named the exact target, stated plainly that the exposure is to authentication and integrity rather than to the confidentiality of your data, and named the two standards — ML-DSA and SLH-DSA — that this class of result does not touch at all. It also drew the distinction that matters commercially: a signature compromise is exploitable going forward, not retroactively against traffic recorded today. It also briefed the United States government before publishing, rather than after.
IonQ released the resource estimate while withholding the specific circuits. That is the established convention for offensive cryptanalysis rather than a lapse — the release says so, and comparable published work does the same — so outside experts can check the accounting without being handed an implementation. It also briefed the United States government before publishing rather than after. Companies selling fear do not behave this way. This is what a research organization looks like when it happens to have a sales team.
4. IonQ owns the factory, and two thirds of it is running for other people
Quantum computing has a bottleneck that almost nobody outside the industry talks about: somebody has to actually manufacture the chips, and very few places on earth can. IonQ bought the one that can. Then it did something unexpected — it kept the doors open.
There are nine quantum companies manufacturing at that fab today, and roughly two thirds of the quantum production running through it is for somebody other than IonQ. Some of those customers build competing technology.
Every other part of IonQ does better when its competitors do worse. The factory does better when they do better. Whichever approach to quantum computing eventually wins, IonQ gets paid for building it. |
5. A competitor with a Nobel Prize stood on IonQ’s stage and endorsed the arrangement
John Martinis shared the 2025 Nobel Prize in Physics for experiments demonstrating quantum tunnelling and energy quantization in an electric circuit — the work that superconducting qubits, a rival approach to IonQ’s, are built on. He co-founded a company, Qolab, that competes with IonQ on technology. He appeared at IonQ’s investor day to say that IonQ owning this factory is good for the industry, and his company signed a multiyear manufacturing agreement to build there.
People do not do that as a favour. They do it when they have looked closely and concluded that the arrangement genuinely serves them. It is the single most credible endorsement available anywhere in this event, precisely because it came from someone with reason to say the opposite.
6. IonQ pointed its quantum computer at its own satellites, and it worked
IonQ owns satellites that photograph the earth using radar, which sees through cloud and darkness. It also owns quantum computers. On September 8 it showed a hybrid workflow in which its own radar imagery was processed with an IonQ quantum computer in the loop: a structural change at an air base, identified closer to ground truth than either classical method managed — 0.41 on the accuracy score IonQ used, against 0.24 and 0.16 for the two classical baselines. The satellite was not steered by the quantum computer; the imagery it had already captured was run through one.
A satellite company without a quantum computer cannot build this. A quantum company without satellites cannot build it either. One company owns both, and it is not a coincidence — it is what owning the whole stack is for.
7. Eleven businesses went in. One operating team came out.
This is the part that deserves more credit than it will get, and the deal count is only the setup. Since 2025 IonQ has closed ten transactions — Qubitekk, ID Quantique, Lightsynq, Capella Space, Oxford Ionics, Vector Atomic, Skyloom, Nexus Photonics, Seed, and SkyWater — with eight of them inside a single fifteen-month window. Absorbing one acquisition of that size usually takes a company two to three years, and usually only half of it works.
The test is not the press releases. It is whether people from six different origins can hold one roadmap for three hours, in public, under questioning. Three founders did not leave; they were promoted into the acquirer. The co-founder of Oxford Ionics, the British ion-trap and electronic-qubit-control company bought in September 2025, now owns the entire computing roadmap, including hardware he did not build. The chief executive of the semiconductor company acquired on July 31, 2026 stood up thirty-nine days later and presented the manufacturing plan for the whole company — then took the hardest question of the day and answered it without hedging. The co-founder of Lightsynq owns the interconnect programme and the new merchant foundry line.
The evidence is in the handoffs rather than the biographies. Compute passed to manufacturing in the middle of an argument. Three executives from three acquired companies built one answer on cycle time without contradicting each other. The chief executive deferred a question about Nexus Photonics to the person who now runs it. Five people made the same convergence argument in five different professional vocabularies and nobody corrected anybody. Federated companies defend their perimeters; this one did not appear to have any.
IonQ says this itself, in the Superion release rather than from the stage. De Masi credits the product to the two acquisitions jointly — Oxford Ionics enabling trapped-ion control with standard electronics, SkyWater unlocking manufacture at semiconductor cost and scale — and calls Superion 256 the first quantum computer platform designed to be built by the hundreds rather than one at a time. The flagship product is presented as the output of an integration, not of a laboratory.
The exception proves it. SkyWater kept its own name deliberately, because it has to sell to IonQ’s competitors. Separation exactly where separation has commercial value, and one company everywhere else.
Thirty-nine days after SkyWater closed, nobody in that room could have told which parts of the company arrived when. |
The roll-up is no longer the story. The operating system is. Acquisitions assemble parts; September 8 showed the parts taking one another’s questions. Section 3 sets out the handoffs in full.
8. The people who built America’s space and intelligence institutions were sitting together
The final panel had three men on it. One founded the United States Space Force. One ran the National Geospatial-Intelligence Agency, which turns satellite imagery into national decisions. One ran the quantum program at Sandia National Laboratories for two decades and then directed IARPA, the agency that decides which intelligence research becomes real capability.
Domain, mission, and the path between research and use. A company might recruit one such person. Having all three, and having them describe the same company in the same terms on the same afternoon, says something about access that no competitor can currently match.
9. The nature of the risk changed, and that matters more than any single announcement
For years the honest question about quantum computing has been whether the physics will cooperate. That question is unbounded — nobody can tell you when a physics problem will yield, or whether it will.
Listen to what IonQ’s own executives said worries them now. Not the physics. Yields. Schedules. Making the manufacturing boring. Seven analysts had the floor and not one of them asked whether the physics works; they asked about tolerances, cycle times and 2027 revenue.
A company whose remaining problems are physics problems is a science project. A company whose remaining problems are manufacturing problems is an industrial business. IonQ crossed that line, and said so out loud. |
Read that as a statement about what management and the analysts treated as residual risk, not as a claim that the physics is finished. It is not. The component number is strong and it is published: a 99.99% two-qubit gate fidelity world record set in October 2025 on arXiv, using the same Electronic Qubit Control technology that sits at the heart of the Superion chip, and reproduced on the Superion unit cell. The deck shows that figure as a completed milestone on the path to deployment rather than a target. What has not been published is the median across the assembled 256-qubit device, and no date has been given for it. That is a narrower gap than a reader might infer from how much attention it gets. That is the real technical hole, and Section 9 sets out what would change the view.
Why This Puts the World on Notice
Put those observations together and the conclusion is not really about IonQ’s share price. It is about where this industry now runs through.
A company that owns the manufacturing base, sells it to its own competitors, has the deepest security position, holds the satellites and the computer, and can convene the founders of America’s space and intelligence institutions on one stage, is not a participant in a race. It is the ground the race is being run on. Governments deciding how to build sovereign quantum capability, enterprises deciding when to replace their encryption, and rival companies deciding where to manufacture are all now making decisions that route through the same place.
That is what it means to put the world on notice. Not that IonQ won something on September 8, but that everyone else’s plans now have to account for what IonQ has already built.
Contents
Front Matter
Opening Observations, Strongly Supported by the Evidence 2
Why This Puts the World on Notice 5
What This Does Not Mean 5
The Cast — What Was Acquired, and Who Stood On Stage 8
The Thesis 9
The Second Argument 9
Why This Report Matters 9
Standing on the Record 10
How This Report Is Built 10
Method and Evidence Standards 11
Part One — The Argument
1. The Market’s Model of IonQ Is Out of Date 13
2. Four Results That Prove It 18
3. One Company on One Stage 23
4. The Stack, Layer by Layer (Layers 1–10) 29
5. Nine Witnesses Whose Standing Does Not Come From IonQ 37
6. What a Closed Stack Unlocks, Ecosystem by Ecosystem 40
7. The Compounding Layer: Where the Value Actually Sits 49
8. What This Changes 51
9. What Would Change the View 56
10. The Verdict 57
Part Two — The Examination
The Method of the Examination 58
11. Niccolò de Masi — Chairman and Chief Executive Officer 59
12. Dr. Chris Ballance — President, Quantum Computing 61
13. Tom Sonderman — SkyWater and IonQ Foundry Operations 64
14. Inder Singh — Chief Financial and Operating Officer 67
15. Jordan Shapiro — Quantum Platform 70
16. Mihir Bhaskar — SkyWater Quantum Solutions 73
17. Chad Sakac — Go-to-Market 76
18. Qolab — Dr. John Martinis and Alan Ho 79
19. John Lokada — ServiceNow 81
20. Robert Long — EPB 83
21. Dr. Matthew Rabinowitz — Natera and MyOme 85
22. Alessio Butti — Italy 87
23. Gen. John W. Raymond 89
24. Robert Cardillo 90
25. Rick Muller 92
26. The Analysts — What Seven Questioners Chose to Ask 94
27. What the Examination Establishes 96
Appendices
Appendix A — Speaker Citation Index (25 voices) 98
Appendix B — The Full Proceedings, Categorized (23 categories) 100
The Cast — What Was Acquired, and Who Stood On Stage
Ten transactions since late 2024. The column that matters is the last one: not what IonQ bought, but who from each business was on the stage on September 8 and what they now run. This is the compressed version of the evidence set out in full in Section 3.
Acquired | Closed | Who appeared on September 8 | What they run now |
Qubitekk | Late 2024 | — (EPB spoke to the relationship) | Quantum key distribution and networking hardware; the original partner in EPB’s 2016 fiber QKD work |
ID Quantique (~86%) | Apr 30, 2025 | — (Shapiro presented the portfolio) | Clavis QKD devices, the hardware layer of the Congruity360 rollout |
Lightsynq | May 30, 2025 | Mihir Bhaskar, co-founder | Photonic interconnect and quantum memory; now also leads SkyWater Quantum Solutions, the merchant foundry line |
Market-intelligence business | Jun 9, 2025 | — | Commercial intelligence function |
Capella Space | Jul 11, 2025 | — (Shapiro presented the SAR result) | The synthetic-aperture radar constellation behind the Miramar demonstration |
Oxford Ionics ($1.075B) | Sep 16, 2025 | Dr. Chris Ballance, co-founder | President, Quantum Computing — the entire compute roadmap, including hardware he did not build |
Vector Atomic | Oct 2, 2025 | — (Shapiro presented the clocks) | Quantum optical clocks, gravimeters, gyroscopes; the sensing pillar |
Skyloom | 2026 | — | Optical terminals; the free-space link Shapiro named as the next networking step |
Nexus Photonics | 2026 | — (Bhaskar answered for it) | Integrated photonics design and IP; moving sensing devices from hand assembly onto chip |
SkyWater ($1.8B, announced Jan 26, 2026) | Jul 31, 2026 | Tom Sonderman, chief executive | Foundry operations for the whole company — and the merchant business selling to IonQ’s competitors |
[FACT] Three founders of acquired companies were given businesses inside the acquirer rather than laboratories: Ballance, Sonderman and Bhaskar. Between them they presented compute, manufacturing and interconnect — three of the four pillars — and took six of the seven analyst questions.
[INFER] SkyWater is the only acquisition that kept its own name, and the reason is commercial rather than sentimental: it sells to nine quantum companies, most of which compete with IonQ. Separation exactly where separation has value; one company everywhere else.
Close dates are from IonQ’s Form 10-Q. Where a column reads with a dash, no principal from that business appeared under their own name; the capability was presented by whoever owns the pillar it now sits inside, which is itself part of the argument in Section 3.
The Thesis
THE ARGUMENT IonQ's full stack is not a roadmap that will one day close. It closed. September 8 was the disclosure, not the milestone. The consensus model of this company — a quantum computing business that bought a fab, a security company and a sensing company, and now has to make them work together — is obsolete. It was reasonable on September 7. It is wrong today, and everyone still using it is mispricing the same asset for the same reason. The evidence is not a single announcement. IonQ put working results on the record at every layer of the stack in one afternoon, and produced three of them that cannot exist unless the layers were already integrated. A first end-to-end compilation of Shor's algorithm is not an algorithm result. It is an integrity test of the entire stack, because you cannot compile an algorithm down to every physical operation without a settled architecture, a working error-correction scheme, a real compiler and an accurate device model all existing at the same time. The company's own framing makes the point sharper than the stage did: the paper proved, rather than assumed, a lower bound on the probability that the full computation succeeds. A photonic interconnect above one kilohertz built by connecting a compute-optimized quantum device to a memory-optimized quantum device is not a networking result; it is a systems-architecture result. And a hybrid quantum-classical workflow processing IonQ's own satellite radar data through IonQ's own quantum computer is not an application result; it is the stack consuming itself and producing something neither half could produce alone. None of these is a component result. Each one requires three or more layers to be simultaneously real, and each one would have been impossible to produce eighteen months ago. Every ecosystem opportunity in this report follows from that. |
The Second Argument, Which Matters Almost as Much
Any company can assert that it has a full stack. No company can fake nine outside parties, none of whom it employs, each standing behind a different layer of it on the same afternoon. That is not a testimonial exercise. It is a distributed audit, and IonQ passed it in public.
A Nobel laureate in superconducting qubits validated the silicon layer. The chief executive of a competing quantum company validated the foundry's unit economics. An Italian government undersecretary validated the sovereign distribution layer. A public utility validated the networking and applications layers with a system it operates and algorithms it wrote itself. The founder of a multi-billion-dollar genomics company validated the applications layer with a result his own team reproduced. A ServiceNow futures director validated the enterprise-workflow layer. A former director of the National Geospatial-Intelligence Agency validated the intelligence-analysis layer. And six sell-side analysts, whose job is to find the weak joint, spent their questions on manufacturing tolerances and cycle times rather than on whether the physics works.
[ARG] The distribution is the evidence, and it is the part most readers will miss. Weakness in a stack shows up as clustering: corroborators gather around the strong layers and go silent around the soft one. There was no silence. Nine parties, nine layers, no gaps — and one of them holds the 2025 Nobel Prize in Physics, awarded for the work that underpins the competing qubit modality. Section 5 names every witness, states each one’s relationship to IonQ, and identifies the layer each one carries.
Why This Report Matters
This series spent eighteen months building the case that IonQ was becoming a vertically integrated quantum platform, out of filings, patents, papers and personnel records — inference work, carefully tiered, because the company would not say it plainly. That phase is over. IonQ said it plainly, showed the parts running, and put outside parties on stage to speak to them. The inference was right, and it was too cautious.
So the question changes. It is no longer whether the platform is real — that argument is finished. It is what a working platform makes possible that a pile of components did not. This report answers that ecosystem by ecosystem, and then goes after the second-order opportunities that exist only where two or more layers meet. That is where the durable value in a vertically integrated business sits, and it is the part nobody else is modeling.
How This Report Is Built
The report is in two parts, and they do different work. Part One makes the argument across ten sections: what the market’s model gets wrong, the four results that prove it, the stack layer by layer, what a closed stack unlocks, the concessions, what would change the view, and the verdict.
Part Two is the evidence, examined rather than summarized. Every speaker is taken one at a time and put under a structured set of questions — the ones actually asked from the floor, and the ones this author puts to their prepared remarks — with what each answer does for the stack and for the business model stated underneath it. Sixteen examinations, one hundred and twelve questions, closing with what the examination establishes that no single speaker’s remarks would support alone.
A reader who wants the case can read Part One and stop. A reader who wants to test it should read Part Two, because that is where the case can be attacked.
Standing on the Record
This report does not begin from zero, and a reader is entitled to weigh it against what this series has already published and staked its name on. The Quantum Technology Integration Series has produced a continuous, primary-sourced body of work on IonQ over roughly eighteen months. Much of it argued, from filings and patents and papers rather than from company statements, for exactly the conclusion this event confirmed. The table below sets out what was claimed, when, and what September 8 did to it.
Prior work | What it argued | What Investor Day did |
The Quantum Full Stack (Part III of the series, 122 pp.) — blended platform scores placing IonQ at 8.7 against IBM 7.1, Google 6.1 and Quantinuum 5.9 | That IonQ led the field not on any single dimension but on the completeness and coherence of its stack | Confirmed at every layer, and by parties outside the company. The scoring gap now looks conservative rather than aggressive |
The Integrated Leader: Why IonQ Is Best Positioned for the Fault-Tolerant Era — the flywheel argument, built on explicit dependency mechanisms rather than a checklist of assets | That IonQ’s assets compound through describable mechanisms, each layer improving the next | Ballance made the identical argument on stage, calling it a virtuous flywheel of innovation obtainable only by working across the entire stack [Ballance 18:36]. The Shor’s compilation is the worked proof the flywheel argument predicted would eventually appear |
The Full Platform (38 pp., August 11, 2026) — eleven-ecosystem map and the thesis that IonQ had become the world’s only vertically integrated quantum-infrastructure platform | That the platform was real but management had not yet said so plainly | Management said it plainly and adopted the same structure, elevating the foundry to a named pillar. The eleven-ecosystem map required one addition — merchant supply — covered in Section 6.1 |
The IonQ Delivery Ledger, through Edition 1.2 — sixteen commitments met, eight open, zero slipped, with sales and deliveries scored separately | That IonQ’s distinguishing commercial characteristic is that it delivers what it says on the date it says | de Masi named the say-do ratio from the stage and claimed twenty to twenty-five consecutive quarters of hitting technical and commercial milestones [de Masi 2:45:40]. Four new dated commitments were created; they are tracked in Section 10 |
The Quantum Networking Tsunami (v3, 109 pp.) | That quantum networking would commercialize faster than the market expected, with IonQ positioned across the entire layer | The interconnect crossed the kilohertz threshold that makes distributed quantum computing operational rather than theoretical (Section 2.2), and EPB expanded from network to computer to quantum memories |
The IonQ Master Energy Efficiency Report (60+ pp.), built on Ballance’s framing that compute is fundamentally a markup on electricity | That energy efficiency would become a primary commercial axis, not a technical footnote | Energy efficiency appeared in the product pitch, the cost-per-qubit chart, the wall-socket deployment argument and the LLM results — four independent places in one event, plus a stated projection of a greater-than-300× cost-per-qubit reduction |
The sixteen-nation flagship report (124 pp., 1,315 personnel records) | That IonQ was building sovereign relationships country by country, ahead of any competitor | Italy moved from presence to a national go-to-market partner announced alongside a sitting minister. That country entry now needs revising upward (Section 6.10) |
The IonQ–SkyWater regulatory memorandum — HSR Second Request, vertical merger analysis, and the national-security dimension | That the transaction would clear and that the national-security framing was the reason it would | Cleared July 31. Sonderman framed quantum as sovereign infrastructure and confirmed classified work inside the foundry segment [Sonderman 31:11; Singh 47:01] |
The CHIPS Act equity-participation analysis — that IonQ became a materially stronger candidate for federal equity participation once it owned an accredited domestic foundry | That the post-SkyWater asset base satisfies the program’s stated rationale better than it did before the close | Not addressed by management and not raised by any analyst. The question remains fully open on unchanged terms, and is the largest untested government thread in the portfolio |
The pharmaceutical account dossier program — AstraZeneca (248 pp.), Vertex (40+ pp.), and a 27-section master dossier framework | That IonQ’s pharmaceutical opportunity turns on specific unsolved chemistry, above all cytochrome P450, rather than on generic drug-discovery claims | IonQ put the machine class, the molecule class, the timeline and the value on the record in one passage [Chad 1:42:36]. This is the single most useful public statement the dossier program has received (Section 6.7) |
The patent portfolio work — 610 issued and 514 pending owned or controlled, plus 131 exclusively licensed, supporting the company-reported 1,200-plus figure | That IonQ’s IP position is far deeper than independent trackers report, because they count only directly assigned patents | Ballance identified quantum multiplexed I/O as covered by a unique patent portfolio and as the core enabling technology of the entire Superion line [Ballance 18:36]. The portfolio is now load-bearing for the roadmap, not merely defensive |
The space-based quantum infrastructure thesis — assembled capability, explicitly not announced intent | That IonQ has assembled the component parts for orbital quantum infrastructure without stating that ambition | Shapiro stated intent for free-space, airborne and eventually space-based quantum links [Shapiro 36:47]. The boundary this series maintains still holds and is restated in Section 6.4 |
[ARG] The point of that table is not that this series was right. It is that the record was built in public, dated, and specific enough to be wrong — and that the direction of the error was consistently the same. Where the prior work missed, it missed by being too cautious. That is the calibration a reader should carry into the argument that follows, and it is why this report is stated as forcefully as it is.
Where the prior work missed, it missed by being too cautious. |
Two documents in particular sit directly underneath this one. The Q2 2026 earnings call was corrected, indexed across seventeen categories and over two hundred line-referenced entries, and synthesized into an institutional report — which is why the financial continuity in Section 8 can be stated without re-deriving it. And the full proceedings of this event have been indexed across twenty-three categories, reproduced in Appendix B, which is the evidentiary substrate for every citation in this report.
Method and Evidence Standards
One note on sources. IonQ did not publish its Investor Day slides at the time of the event, and the 8-K filed that day carried only the guidance release as an exhibit; the 107-slide presentation is now publicly available. Where the deck states something the spoken remarks did not — the sourcing behind the addressable-market figure, the accuracy scores behind the radar demonstration, the cost-per-qubit arithmetic — this report uses the deck and says so. Where the deck and the transcript differ on a proper noun, the deck and the press releases govern. Nine slides are reproduced in this report where the exhibit carries the argument better than a description of it would; each is captioned with what it shows and, where relevant, what it does not.
Every claim in this report is anchored to a named speaker at a timed turn in the Investor Day proceedings, in the form [Ballance 18:36]. Appendix B reproduces the full categorized index of those proceedings — twenty-three category indices covering every ecosystem, platform, technology, market, customer, government relationship and forward-looking statement in the event — so that any citation here can be traced to its source turn and read in context. Where a claim rests on this series' prior published work rather than on the event, that is stated. The evidence tiers below apply throughout, and the calibration section is titled What Would Change the View. No numeric probabilities are assigned. No market sizing appears beyond figures stated on stage.
Tier | Meaning | Worked example from this report |
FACT | Stated on the record by a named speaker at the event, or independently documented. | The first end-to-end compilation of Shor's algorithm, and a 30× improvement in CCZ magic-state factories, were both stated by the President of Quantum Computing on stage. |
INFER | A reasonable reading of stated facts that no speaker stated directly. | That the Shor's compilation functions as an integrity test of the whole stack is an inference from what compilation requires, not a claim any speaker made. |
ARG | This author's argument or judgment, offered as such. | That the distribution of external corroboration across layers is itself evidence of stack maturity is an argument, and a reader may weigh it differently. |
UNDISC | Genuinely not disclosed. | Median gate fidelity across all 256 qubits was asked for directly and no date was given for its availability. |
PART ONE
The Argument
Why the Market’s Model of IonQ Is Out of Date
Ten sections building a single case: that on September 8, 2026 IonQ closed the last open layer of the quantum stack, and that the nature of the risk in this company changed from a physics question to a manufacturing one. The evidence for every claim made here is examined speaker by speaker in Part Two.
1. The Market's Model of IonQ Is Out of Date
[FACT] The deck also states the installed base, which the remarks did not. More than $350 million of lifetime revenue; customers in 25 countries who have run workloads; more than 1,800 unique IonQ Cloud users across six continents, excluding those arriving through AWS, Google or Azure; more than 33,000 lifetime hours of cloud uptime; five generations of commercial systems from Harmony in 2019 through Tempo in 2025; and more than 1,200 patents granted and pending worldwide — the same figure IonQ uses in its Form 10-K and its investor materials.
The competitive position stated as arithmetic. IonQ's $80.1 million of second-quarter revenue against $43.6 million for the rest of the publicly reporting pure-play industry combined — Infleqtion $12.6M, Quantinuum $8.0M, IQM $7.6M, Quantum Computing Inc. $5.6M, Rigetti $5.1M, D-Wave $3.1M, Xanadu $1.5M. These are the companies' own reported figures and revenue is not the only measure of a quantum business, but the gap is the reason a single-company comparison is now the wrong frame.
IonQ's own framing of the platform, Investor Day 2026 — four product pillars over a quantum foundry layer, with customer segments across the base. The foundry sitting underneath rather than beside the pillars is the structural claim this report is testing.
Thirteen layers moved in a single afternoon. Not one moved backward. |
Thirteen layers moved in a single afternoon. Not one moved backward. The table below sets the prior consensus read against what is now on the record, with the speaker who moved each one. Read the middle column and the right column together; the distance between them is the mispricing.
Layer | Prior defensible read | After September 8 | Evidence from |
Silicon and foundry | A newly closed acquisition carrying integration risk; a fab that would need to learn quantum. | A ten-year-old operating quantum foundry running thousands of quantum wafers, of which roughly two thirds are for parties other than IonQ, across nine customers, on a mature 90nm 200mm process — now branded and launched as SkyWater Quantum Solutions, a merchant business with a named leader. | Sonderman 31:11; Sonderman 1:11:46; SkyWater Quantum Solutions release |
Qubit control | Electronic qubit control understood from patents and papers; performance at scale unproven. | Better than four nines two-qubit fidelity on the unit cell, quantum multiplexed I/O as the array technology, and wafer-scale chip production already running. | Ballance 18:36 |
Device integration | The hardest unsolved problem in the roadmap. | Reframed as a classic semiconductor integration problem, with tiling tolerances studied and found comfortably inside industry standard, and cryogenic CMOS de-risk wafers already fabricated. | Ballance 1:23:45; Ballance 18:36 |
Systems | 256 as a roadmap milestone. | Ion qubits loaded in the first prototype systems across several US facilities, fully integrated 256-qubit QPUs fabricated, orders open, cloud availability planned, and a first system already sold in Q1 2026, to the University of Cambridge, before the platform was named. | Ballance 18:36; Superion release |
Architecture and QEC | Walking Cat as a published blueprint; unclear whether it would survive contact with hardware. | The architecture of the shipping 256 compute fabric and the stated path to 10K, with customized error-correction codes demonstrated at breakeven on real hardware. | de Masi 4:40; Ballance 18:36 |
Classical control software | Assumed to be a downstream problem. | A mega-quantum-operation real-time decoding stack published within the prior fortnight, tailored to a 10,000-qubit device and running in parallel with quantum hardware. | Ballance 18:36 |
Compilation and algorithms | Cost calculations existed in quantity, all written against idealized architectures. | An end-to-end compilation of Shor's algorithm against secp256k1, targeted at IonQ's own architecture rather than a generic one: 19,397 physical qubits, 1,457 logical qubits, about 39 million logical Toffoli gates, roughly 25.7 days per attempt — with a 30× improvement in CCZ factories falling out of the work. | Ballance 18:36; secp256k1 release |
Interconnect | Quantum interconnects benchmarked in hertz — orders of magnitude short of useful. | Above one kilohertz, achieved through a heterogeneous architecture pairing a trapped-ion compute device with a quantum memory. | Bhaskar 1:31:00 |
Photonic integration | Nexus Photonics understood as a component acquisition. | A named foundry platform, with on-chip laser technology pioneered by the field's originator, whose first commercial impact is chip-scale sensing. | Bhaskar 1:31:00; Bhaskar 1:27:36 |
Applications | Demonstrations rather than production workloads. | Protein folding to 16 amino acids on shipping hardware, a stated machine class for cytochrome P450 chemistry, and four best-paper awards from ten entries at IEEE Quantum Week. | Chad 1:42:36 |
Distribution | Cloud and direct sales, with sovereign interest asserted. | A national go-to-market partner in Italy announced alongside a sitting minister, plus cloud, on-premises and sovereign as three explicit deployment modes. | de Masi 2:27:33; Butti 2:21:37; Chad 1:42:36 |
Unit economics | Cost per qubit understood directionally as improving. | A stated projection of more than a 300× reduction in cost per qubit across the roadmap, attributed to moving from laser control to semiconductor electronics. | Superion release |
Trust and governance | Not previously treated as a stack layer at all. | A cryptanalysis result briefed to the senior-most levels of government before publication, described from the board seat as a deliberate act of partnership. | Cardillo 2:43:39; de Masi 4:40 |
[FACT] Thirteen layers moved in one afternoon, each on the record from a named speaker. No layer moved backward. The companion Master Index catalogues what was not addressed; nothing in that gap list concerns the technical stack.
1.1 — The Risk Moved Upward, and That Is Everything
The old summary: IonQ has the deepest quantum stack in the industry, and its deepest layers are the least proven. The new one: the deepest layers are now the most proven, and every open question has migrated upward into yield, cadence and commercial conversion. Those are ordinary industrial questions. They have ordinary industrial answers, and they are answered by semiconductor people, of whom IonQ now employs several hundred.
[ARG] That migration is the whole story, and it is worth being blunt about what it means. A company whose remaining risks are physics risks is a science project with a business attached, and it should be valued like one. A company whose remaining risks are yield and schedule risks is a manufacturer, and manufacturers are valued on entirely different terms. IonQ crossed that line and told the room it had crossed it, and almost nobody wrote it down. Everything in Sections 6 and 7 is downstream of that single fact.
His exact framing on the tiling question was that the residual concern is not any single technology risk but how to properly understand and execute against the risks, doing enough in parallel to make them boring — and that this is not a particularly exciting problem, which he offered as high praise [Ballance 1:23:45].
1.2 — What the Releases Added Beyond the Room
de Masi described the day as a six-announcement event, and the six releases carry material detail that no speaker stated aloud. This report is built on both. The items below are not in the transcript at all, and several of them strengthen the argument more than anything said on stage.
Disclosed only in the releases | Why it matters here |
IonQ sold its first 256-qubit system in Q1 2026 to the University of Cambridge, before the platform was publicly named, under a partnership spanning computing, networking, sensing and security, under a wider agreement covering quantum computing, networking and intellectual-property development. | The product had a paying customer two quarters before it had a name. That converts the launch from a roadmap announcement into a shipment schedule, and it means the early-2027 deployment date has a named counterparty behind it. |
The resource estimate targets secp256k1 — the 256-bit elliptic curve used by Bitcoin and other systems — and models 1,457 logical qubits, roughly 39 million logical Toffoli gates, and about 25.7 days per attempt. | The stage version gave one number. The release gives the full resource ladder, which is what makes the result checkable by outside cryptographers and therefore what makes it credible. |
The scope is elliptic-curve digital signatures, not confidentiality of stored or transmitted data, and IonQ stated that post-quantum signature standards including ML-DSA and SLH-DSA would not be affected by this type of attack. | This is a narrowing that IonQ imposed on itself. It is the strongest available evidence that the result was published as science rather than as marketing, and it materially qualifies the security framing used on stage. |
Six tapeouts completed in the first half of 2026; design cycle reduced from nine months to two; twelve times more wafer lots run over six months than at the previous foundry. | The cadence claim now has three independent measures rather than one analyst-supplied figure. Note the correction: IonQ's own number is nine to two, not the eight to two stated from the floor. |
Superion 10K is being developed in parallel with the 256 rather than sequentially, adding cryo-CMOS, with a laboratory fault-tolerance demonstration targeted for 2027 and a manufacturable commercial system in 2028. | Two new dated commitments, neither given from the stage. Parallel development of the next generation is also the clearest single piece of evidence for the cadence thesis in Section 2.4. |
A projected reduction of more than 300× in cost per qubit across the roadmap, from replacing laser control with semiconductor electronics. | The economic expression of the entire architectural bet, and the number against which the platform thesis will ultimately be judged. |
The Congruity360 agreement is valued at $8.18 million and covers Clavis quantum key distribution devices and Solteris network appliances. | The stage described the deal as the largest of its kind in the US without a figure. Eight million dollars is a useful calibration of what 'largest' currently means in this market. |
The Qolab agreement is a multiyear manufacturing agreement moving Quantum System-in-Package devices from custom development onto standardized SC250 wafer services, with SkyWater investing in dedicated Minnesota equipment upgrades. | This is what productization of the foundry actually looks like: a competitor migrating from bespoke engagement to a catalogued process, with the foundry putting capital behind it. |
[FACT] IonQ did not disclose gate fidelities, error rates or logical-qubit results for the Superion prototypes in the releases either. The gap identified from the stage is a genuine disclosure gap, not a transcript artifact.
2. Four Results That Prove It
Thirteen layers moving is a strong signal. It is not a proof, because layers can improve independently. What follows are four results that cannot be produced by independent layers. Each requires three or more to be simultaneously real and mutually consistent. These are the load-bearing evidence, and a reader who wants to reject the thesis has to break one of them.
2.1 — Shor's Compilation Is a Stack Integrity Test, and IonQ Passed It
For thirty years, Shor’s algorithm has been the benchmark against which scaled fault-tolerant quantum computing is measured, and people have spent careers optimizing and understanding the nuances of how to approach it [Ballance 18:36]. This report is careful about what is and is not novel here, because the surrounding literature is substantial and a reader who knows it will not accept a claim of primacy.
Detailed cost calculations for Shor’s algorithm are an established sub-field with a long lineage: Gidney and Ekerå put 2048-bit RSA at twenty million noisy qubits in 2019, Gidney brought that below one million in May 2025, Roetteler, Naehrig, Svore and Lauter did the elliptic-curve case in 2017, and in February 2026 Iceberg Quantum’s Pinnacle architecture claimed RSA-2048 under one hundred thousand physical qubits using quantum LDPC codes. So the exercise of costing an attack against a machine nobody has built is neither new nor rare.
The like-for-like comparison is the one that matters, and the paper states it plainly. Previous estimates for solving this same problem on a trapped-ion architecture required between 1.2 million and 9.4 million qubits. IonQ’s figure is 19,397. That is a reduction of roughly sixty to nearly five hundred times against the prior art for its own modality, achieved by optimizing the algorithm, the compiler, the architecture and the error-correction layer together rather than separately. Comparing IonQ’s trapped-ion elliptic-curve number against a superconducting or modality-agnostic RSA claim is not a comparison at all, and this report does not make one.
[FACT] The specific optimizations are named. Point-addition arithmetic circuits for pseudo-Mersenne primes were rebuilt by combining conditionally-inverted adders for in-place multiplication with a Karatsuba-split modular squarer, cutting the requirement from a prior baseline of 58 million Toffoli gates to 39 million, consuming 273 million T-states, at 1,457 logical qubits. The heuristic single-run success probability is 63.3 percent, and the paper supplies a provable lower bound on that probability rather than a heuristic argument — which is the methodological claim Ballance made from the stage.
[FACT] Two of the four authors of the canonical 2017 elliptic-curve resource estimate — Martin Roetteler and Michael Naehrig — appear on IonQ’s author list, alongside Dmitri Maslov, Nicolas Delfosse, Mark Webster and John Gamble. The prior art and the new work share authorship, which is the most concrete evidence available that IonQ bought the people who define this literature rather than merely citing it.
[FACT] What distinguishes the published work is the target, not the genre. Every estimate named above is written against an idealized architecture, with assumptions declared up front — a square grid of qubits with nearest-neighbour connections, a uniform physical error rate, a stated code cycle time, a stated reaction time. That is the standard and correct methodology for a machine that does not exist.
IonQ says the same thing about its own work, and the concession is worth quoting because it settles the question. The release states that the estimate follows recent published work rather than announcing a new attack, that IonQ’s logical-layer figures advance on similar research into optimized point-addition circuits and elliptic-curve resource estimates, and that what is new is the architecture-specific accounting — what the same algorithm costs on a high-rate error-corrected trapped-ion machine rather than on a surface code. A company claiming a first in its own headline conceded the lineage in its own body text.
[FACT] The point is sharpened by who now works there. Martin Roetteler, VP Quantum Applications R&D at IonQ and one of the three executives quoted in the release, is a co-author of the 2017 elliptic-curve resource estimate that is part of the prior literature this work builds on. The person who wrote the canonical earlier estimate is inside the company that produced the newer one.
IonQ compiled against a machine that has a name, a manufacturing process and a fab. The architecture is Superion, the error-correction scheme is the Walking Cat code published in April, and the chips are being produced by the wafer at a foundry IonQ owns. Ballance stated that this is not an estimate, that they know exactly how many qubits are needed, exactly what software is needed, and that the full model is implemented down to every single physical operation to be played on the device [Ballance 18:36]. The release adds the methodological claim that carries the most weight: IonQ proved, rather than assumed, a lower bound on the probability that the full computation succeeds.
WHY THIS IS A STACK RESULT, NOT AN ALGORITHM RESULT To compile an algorithm end to end, down to physical operations, four things must simultaneously exist and agree with each other: a settled hardware architecture with known native operations; a working error-correction scheme with characterized overheads; a compiler that can lower a logical circuit through that scheme onto that hardware; and a device model accurate enough that the resulting operation count means something. Any one of those being immature makes the exercise impossible rather than merely imprecise. The output would not be wrong — it would not compile. That IonQ produced a number rather than a range is therefore a statement about the stack, not about Shor's algorithm. The corroborating detail is the byproduct. Ballance reported over a 30× improvement in core error-correction modules such as CCZ factories, discovered in the course of doing the compilation. Improvements of that magnitude do not fall out of a paper exercise; they fall out of an engineering team pushing a real toolchain against a real target and finding slack in it. |
Against the prior art for its own modality, IonQ moved the requirement from millions of qubits to under twenty thousand — and compiled it against a machine it is building rather than one it assumed. |
The resource ladder, from the published paper
The release supplies what the stage did not. The target is secp256k1, the 256-bit elliptic curve used by Bitcoin and other digital systems, and the modeled attack is on the elliptic-curve discrete logarithm problem. A trapped-ion machine of 19,397 physical qubits is estimated to solve it in approximately 25.7 days per attempt, requiring 1,457 error-corrected logical qubits and about 39 million logical Toffoli gates.
Two features of how IonQ framed the work matter more than the numbers. The first is Ballance's stated claim that this is the first time anyone has taken a utility-scale quantum algorithm and estimated its cost without approximating away the parts that usually dominate a real machine's runtime, compiling every operation down to the actual error-correction primitives the architecture runs — and that the paper proved rather than assumed a lower bound on the probability that the full computation succeeds.
The second is the narrowing. IonQ stated that the modeled exposure is to digital signatures used to prove identity and authorize transactions, not to the encryption protecting stored or transmitted data, and that post-quantum signature standards including ML-DSA and SLH-DSA would not be affected by this type of attack. It also noted that no deployed digital asset or crypto platform was affected during the research, and that advance copies were shared with US government and industry representatives before publication.
IonQ published the limits of its own result alongside the result. Companies selling fear do not do that. |
[ARG] A company selling quantum security had every commercial incentive to leave the threat scope vague. IonQ instead published the curve, the runtime, the logical-qubit count, the gate count, and an explicit list of what the attack does not touch. Self-imposed narrowing of this kind is the strongest available signal that the result was released as science rather than as demand generation — and, second-order, it is what makes the number durable enough to become the industry reference rather than being picked apart within a week.
[INFER] That the compilation constitutes an integrity test of the full stack is this report's inference from what compilation requires. Ballance did not frame it that way — he framed it as the innovation flywheel playing out, which is the same observation in the company's own vocabulary [Ballance 18:36].
The commercial expression of the same work arrived from a different speaker on a different stage segment. Shapiro announced 19,397 qubits as IonQ's requirement to reach Q-Day, attributing the reduction to quantum hardware and quantum software working together [Shapiro 36:47]. de Masi described the same result as an IonQ system with about twenty thousand physical qubits recovering a 256-bit elliptic-curve key, done without approximation and to a level of detail the world had not seen [de Masi 4:40].
[ARG] Two executives from different parts of the company describing the same underlying artifact in two different vocabularies — a scientific claim and a market-timing claim — is the clearest single illustration of what a closed stack buys. The research output and the sales input are the same object.
2.2 — The Interconnect Crossed the Line That Kept Distributed Quantum Computing Theoretical
Bhaskar framed the problem precisely: quantum interconnects have historically been benchmarked in units of hertz, while the requirement for actually connecting quantum computers together is kilohertz [Bhaskar 1:31:00]. That is not a gap to be narrowed by tuning. It is a gap that has kept distributed quantum computing theoretical.
The announcement was that IonQ has built a photonic-based quantum interconnect above a kilohertz, characterized explicitly as a step change rather than an incremental advance [Bhaskar 1:31:00]. But the method is what makes it a stack proof. The result was not reached by turning knobs. IonQ connected its trapped-ion computing system to a quantum memory acquired through Lightsynq — a device optimized for computing working with a device optimized for networking, to create a system better than the sum of its parts [Bhaskar 1:31:00].
He then named what had actually been achieved: until this moment, two different types of quantum system had not been successfully combined with improved performance resulting [Bhaskar 1:31:00]. The analogy he used was the ordinary laptop — CPUs, GPUs and networking cards, each designed for a different job, all working together, taken entirely for granted [Bhaskar 1:31:00].
[FACT] This required three layers to work at once: the trapped-ion compute device, the acquired quantum memory, and the photonic layer joining them. It is the first public demonstration that IonQ's acquisitions compose at the physics level rather than merely at the corporate level.
Two different types of quantum system had never been combined to produce improved performance. Now they have — and the number is above one kilohertz. |
Bhaskar extended the claim beyond IonQ's own modality, stating that the milestone applies not only to trapped-ion systems and that this will be the leading interconnect solution for all modalities — trapped ions, trapped atoms, superconductors and more [Bhaskar 1:31:00]. He tied it directly to IonQ's status as a DARPA HARQ performer, noting it is exactly the subject of that program [Bhaskar 1:31:00].
The productive tension worth naming
Asked by Rosenblatt at what physical qubit count photonic interconnect becomes important, Ballance answered that on the core compute roadmap IonQ can scale well beyond millions of qubits without it on a five-year view, and that the customers who want photonic interconnects are typically not the ones who just want one big computer — they are building fleets and networks [Ballance 1:08:42; McPeak 1:08:17].
[ARG] This is not a contradiction and should not be read as one. IonQ built the networking layer as a product and an option, not as a dependency. Its own scaling story does not need the interconnect; the interconnect gets sold to everyone whose architecture does. That is a stronger position than either statement suggests alone — single-path risk removed from the roadmap, commercial optionality retained — and it is the posture a company adopts when it intends to sell into its competitors’ architectures as well as its own.
2.3 — The Stack Consumed Itself and Produced Something Nobody Else Can Build
Shapiro disclosed for the first time that IonQ is using its quantum computers with its own remote-sensing organization, noting that IonQ captures synthetic aperture radar data from across the globe on a daily basis [Shapiro 36:47]. He presented what he described as the first example, to IonQ's knowledge, of a hybrid quantum-classical workflow with SAR data, with the quantum computer placed in the loop processing data from IonQ's own platform [Shapiro 36:47].
The worked example was a structural change at Miramar Air Base. Three models were shown against ground truth. The one run on an IonQ quantum computer identified the structural change, sat closest to ground truth, and lacked the noise and fuzziness of the classical models [Shapiro 36:47]. His summary claim was that the model works better than classical and enables a level of insight no other SAR platform has today, and that IonQ is the only quantum computing company with this level of access to SAR data to run its quantum computers on [Shapiro 36:47].
THE MOST UNDERRATED RESULT OF THE DAY This is vertical integration producing something neither half of the company could produce alone. A satellite operator without a quantum computer cannot build it. A quantum computer company without proprietary radar data cannot build it. Better change detection from the same imagery belongs to whoever owns both, and exactly one company owns both. It is also the only result of the day that runs the stack backwards. Everywhere else the stack terminates in a customer. Here it terminates in IonQ's own capability, which is then sold to an entirely different customer set — one that buys on classification and mission fit rather than on price. Nobody covering this stock has modeled it, because until today nobody knew it existed. |
A satellite operator without a quantum computer cannot build it. A quantum company without proprietary radar data cannot build it. Exactly one company owns both. |
The most credible endorsement came from the board rather than from management. Cardillo, who ran the geospatial-intelligence business inside the US government, noted that synthetic aperture radar was a central theme of that work, applied against adversaries to build a picture of their capabilities and intentions [Cardillo 2:39:32]. His framing of the value was precise: taking an overwhelming amount of data and performing the synthesis between noise and signal — what he called a coherence-from-chaos service, finding the right information at the right time to make a better decision [Cardillo 2:39:32].
[ARG] When the former director of the National Geospatial-Intelligence Agency describes your quantum-enhanced radar analysis as the thing he spent a career trying to do, that is not a technical endorsement. It is a statement about what the intelligence community will pay for, from someone who used to sign the cheques.
2.4 — The Stack Has a Clock Speed, and It Is Faster Than the Industry's
This is the least glamorous proof and the most predictive one. A stack that is closed but slow is a laboratory. A stack that is closed and fast is a business. IonQ's is fast, and the margins are not close.
Ballance stated the ambition directly: taking innovation cycles that are typically years in this industry and striving to turn them into months, with big operational changes to how the company is run, structured, and how teams develop technology in order to achieve it [Ballance 18:36]. The result he reported is that over the last six months IonQ has built more ion-trap chips and ion-trap quantum processors than in the entire history of the company to date, learning and iterating faster across multiple generations in parallel [Ballance 18:36].
Sonderman supplied the manufacturing side of the same claim, describing the move out of the physics demonstration realm into manufacturing at scale in a highly developed operation delivering both the speed required for cycles of learning and the scale to run in parallel [Sonderman 31:11]. He quantified the parallelism against his own history: at AMD, competing against Intel, three nodes ran in parallel; at IonQ and SkyWater, five devices run in parallel [Sonderman 1:17:18].
The specific number came from the analyst bench. Choksi of Northland described the reduction in cycle time from eight months to two months as absolutely astounding and asked how it had been achieved [Choksi 1:09:29]. Neither Ballance nor Sonderman disputed the figure; both explained it. IonQ's own release puts the design cycle reduction at nine months to two, and adds two measures the room never heard: six tapeouts completed in the first half of 2026, and twelve times more wafer lots run over six months than at the company's previous foundry.
Ballance's explanation was cultural rather than technical, and it is worth recording as evidence. Working with two foundries late the prior year, the major industrial foundry pushed delivery timelines out by a couple of months on a six-month schedule over a chemical supply problem. SkyWater faced the identical challenge, told IonQ on a Tuesday that a supplier's one-month quote was unacceptable, found a supplier in Korea with stock by Thursday, and had it installed in a machine yielding results by Sunday — while the other partner was still holding a two-month slip [Ballance 1:10:04].
Sonderman's addition was the node disclosure that explains the cost structure underneath the cadence: this is done on 90-nanometer technology in a 200-millimeter fab, a very mature environment, against others who claim you need 28 nanometers and below or even two nanometers — a claim he called ridiculous [Sonderman 1:11:46].
[INFER] Mature-node manufacturing is the reason the cadence is affordable. Fast iteration at an advanced node would be prohibitively expensive; fast iteration at 90nm on 200mm tooling is ordinary semiconductor practice. The cycle-time claim and the node disclosure are the same fact viewed from two angles, and neither speaker connected them explicitly.
Singh closed the loop on why this compounds: having SkyWater allows IonQ to run multiple prototypes at one time, which others cannot do, so the company finds out which approach works and which does not much more quickly [Singh 47:01]. Sonderman stated the objective plainly — the goal is to continue to move the roadmap to the left, and he believes they are well on the way [Sonderman 31:11].
3. One Company on One Stage
There is a proof in this event that has nothing to do with physics, and for a company built through acquisition it is one of the most consequential. IonQ has acquired ten businesses since late 2024, eight of them inside a single fifteen-month window. What it put on stage on September 8 was not ten businesses. It was one company, with one roadmap, one commercial agenda and one vision, presented by people who came from six different places and spoke as though they had always worked together.
That is an achievement, and it is worth naming as one.It is also the kind of achievement that cannot be manufactured in a press release, because it shows up in how people behave on a stage rather than in what they say from it.
The deal count is the setup. Contradiction-free handoffs, in public, under questioning, are the proof. |
3.1 — The Hardest Thing to Fake
Assembling world-class parts is a capital allocation exercise. Making them behave as one company is not, and there is no shortcut to it. It requires shared roadmaps that people actually build against, retained founders who choose to stay, commercial motions that route customers across business lines rather than defending them, and a leadership team that can hold one story for three hours without drifting into six.
None of that can be asserted. It has to be watched. An investor day is one of the very few occasions on which an outside observer can watch an entire executive team operate in public, under questioning, for long enough to see whether the coherence is real. That is why the format of this one matters as much as its content, and why this section treats the presentation itself as evidence.
[ARG] Every technical achievement in this report was produced by an organization. If the organization is one company, those achievements compound and the next generation arrives faster. If it is a federation of excellent parts, they do not. Section 3 argues IonQ demonstrated the first, and it argues it from behavior rather than from claims — which is the only way the argument can honestly be made.
3.2 — Fifteen Months
The most remarkable fact about the company that appeared on that stage is how recently it did not exist. These are closing dates from IonQ’s own filings, not announcement dates.
Business | Closed | Who represented it on stage | Elapsed to Investor Day |
Qubitekk | Late 2024 | Referenced by EPB as now inside the IonQ family [Long 2:08:09] | About 21 months |
ID Quantique (majority stake) | April 30, 2025 | Clavis QKD devices, at the centre of the Congruity360 agreement | 16 months |
Lightsynq Technologies | May 30, 2025 | Mihir Bhaskar, presenting the kilohertz interconnect and leading the new merchant foundry business | 15 months |
Capella Space | July 11, 2025 | The SAR data behind the hybrid quantum-classical workflow [Shapiro 36:47] | 14 months |
Oxford Ionics | September 16, 2025 | Dr. Chris Ballance, presenting IonQ’s entire compute roadmap | Under 12 months |
Vector Atomic | October 2, 2025 | The clocks and sensing portfolio, and a named cross-portfolio solution [Singh 47:01] | 11 months |
Nexus Photonics | 2026 | The integrated photonics platform inside SkyWater Quantum Solutions | Months |
SkyWater Technology | July 31, 2026 | Tom Sonderman, presenting the manufacturing base of the entire company | 39 days |
Two entries deserve to be read twice. Oxford Ionics closed on September 16, 2025. Less than a year later its co-founder was on stage owning the compute roadmap for the entire company — including hardware he did not build — and handing off mid-argument to a manufacturing executive whose company IonQ had owned for thirty-nine days.
Thirty-nine days. SkyWater closed on July 31, 2026, and on September 8 its chief executive described SkyWater’s roadmap, IonQ’s roadmap and the wider industry’s roadmap as one continuous argument — then took the hardest question of the day, on conflict of interest with the eight other quantum companies in his fab, and answered it without hedging [Sonderman 1:13:58].
[FACT] IonQ completed six acquisitions during 2025 alone, for total consideration of $2,659.0 million, per its own quarterly filing. Qubitekk preceded them in late 2024; Skyloom, Nexus Photonics and SkyWater followed in 2026. That is ten businesses across roughly twenty months, of which the eight above were represented directly on the Investor Day stage.
[ARG] Integration on this timetable is not normal and should not be described as though it were. The standard corporate expectation for absorbing a single acquisition of consequence is eighteen to thirty-six months, and the standard outcome is partial. IonQ absorbed eight in fifteen, kept the founders, and had them presenting a unified technical roadmap in public. Whatever else this event demonstrated, it demonstrated an operating capability in integration rarer than any of the physics discussed on the same stage.
3.3 — What the Stage Actually Showed
Sixteen people delivered remarks across nearly three hours. Between them they came from at least six previously independent companies. Not one of them presented as a subsidiary.
Speaker | Came from | Presented as |
Dr. Chris Ballance | Co-founder and CEO, Oxford Ionics — closed September 2025 | President, Quantum Computing — owning IonQ's entire compute roadmap, including hardware he did not build |
Tom Sonderman | CEO, SkyWater Technology — closed July 31, 2026 | Running the foundry inside IonQ, framing SkyWater's roadmap as serving IonQ and the wider industry simultaneously |
Mihir Bhaskar | Co-founder, Lightsynq — closed May 2025 | Leading interconnects and photonics across IonQ, and stepping into leadership of the merchant foundry business at SkyWater |
Rick Muller | Sandia National Laboratories; Director, IARPA | IonQ, speaking to how government buys mission solutions rather than technology |
Robert Cardillo | Director, National Geospatial-Intelligence Agency | IonQ board and Executive Chairman of IonQ Federal, speaking to applications and governance |
Inder Singh | CFO, Arm | CFO and COO, presenting one consolidated P&L with one reporting taxonomy across both companies |
The retention record underneath that table is the part most worth pausing on. Founders of acquired companies routinely leave within twelve to eighteen months, and the industry treats their departure as unremarkable. Ballance did not merely stay — he was given the compute roadmap for the whole company. Sonderman did not merely stay — he runs the manufacturing base. Bhaskar did not merely stay — he was handed a new business line. That is three founders of three acquired companies holding three of the most important operating roles at the acquirer, presenting one roadmap between them.
3.4 — The Handoffs Were the Evidence
The most persuasive evidence was structural rather than stated. Across the afternoon, speakers repeatedly picked up each other’s material, credited each other by name, and answered questions that belonged to someone else’s domain. Loosely joined organizations do not do that. They defend their perimeters, because in a federation the perimeter is where the budget is.
Moment | What it demonstrated |
Ballance closed his roadmap section by handing directly to Sonderman to explain how the chips get built [Ballance 18:36] | The compute roadmap and the manufacturing roadmap were presented as one continuous argument, not two adjacent ones |
Shapiro opened by thanking Sonderman and telling the SkyWater team, on behalf of all of IonQ, that they were thrilled to have them joining the ranks [Shapiro 36:47] | A president of a different business line welcoming the acquired company publicly, unprompted |
Asked about foundry cycle time, Ballance answered first on SkyWater's culture, then handed to Sonderman for the manufacturing detail [Ballance 1:10:04; Sonderman 1:11:46] | The acquirer's executive advocating for the acquired company's operating culture, from direct experience as its customer |
Singh backed both of them with the procurement history of why SkyWater was chosen over other foundries during the Oxford Ionics acquisition [Singh 1:12:19] | Three executives from three different origins constructing one answer to one analyst question |
de Masi routed the Nexus Photonics question to Bhaskar rather than answering it [de Masi 1:27:29; Bhaskar 1:27:36] | A CEO deferring to the technical owner in front of investors, which only happens when the owner is trusted with the narrative |
Shapiro interjected to point the room at a photonic integrated circuit in Bhaskar's showcase [Shapiro 1:28:48] | Cross-selling another executive's material without being asked |
Chad referenced Alan Ho's remarks approvingly and noted IonQ is doing this not just for trapped ions but for superconducting modalities [Chad 1:42:36] | Go-to-market absorbing a foundry partnership announcement into the commercial narrative within the hour |
Cardillo built his board remarks on Shapiro's SAR slide, and credited Muller by name for leading the government conversations [Cardillo 2:39:32; Cardillo 2:43:39] | The board speaking to operating detail from two different business lines |
de Masi credited Monroe and Bhaskar together for the multimodal networking position [de Masi 2:45:40] | The founder's legacy work and an acquired founder's work presented as one continuous lineage |
Bhaskar supplied the line that captures it. Concluding his segment, he said that whether it is through the scale-out solutions or the scale-up semiconductor ecosystem, all of the roads to scaling quantum are running through the team in this room [Bhaskar 1:42:03]. Singh said the same thing in corporate language: SkyWater and IonQ have come together as one company called IonQ [Singh 47:01].
[ARG] Nine cross-references between executives from six previously separate companies, none of them scripted as a segue, is not a communications achievement. It is what an actually-integrated operating team sounds like when it talks in public. This is the single strongest piece of evidence available anywhere that the roll-up became a company, and it was produced by the format rather than by the message.
3.5 — One Vision, Stated Five Ways, Without Divergence
Drift is the tell. Where business units are only loosely joined, each describes the company slightly differently, optimizing for its own market, and three hours is more than long enough for that to surface. It did not. Five executives made the convergence argument in five distinct vocabularies — technical, manufacturing, commercial, architectural and financial — and the arguments were mutually reinforcing rather than merely compatible. The full comparison is set out in Section 7.1, and the point of it here is narrower: nobody contradicted anybody.
The one apparent divergence is instructive because of how it resolved. Bhaskar presented the photonic interconnect as the unlock for data-center-scale quantum computing; Ballance told an analyst the core compute roadmap does not need photonic interconnects on a five-year view [Bhaskar 1:31:00; Ballance 1:08:42]. On a federated stage that reads as two business units talking past each other. On an integrated one it reads as a deliberate architectural choice — build the networking layer as a product for other people's architectures and as an option for your own, so that neither roadmap is hostage to the other. Section 2.2 argues it is the second, and the reason to believe that is that both men were in the room, both answers were given within thirty minutes of each other, and neither corrected the other.
3.6 — The Deliberate Exception: SkyWater Kept Its Name
One brand was not absorbed, and the exception proves the rule. SkyWater retained its identity and launched a business under it — SkyWater Quantum Solutions — precisely because a merchant foundry serving nine quantum companies, several of them IonQ's direct competitors, cannot credibly operate under the acquirer's brand. Sonderman gave the commitment that goes with it: information is compartmentalized, there is no cross-contamination, and every customer should be a hundred percent confident their roadmap will not be impeded by IonQ's needs [Sonderman 1:13:58].
[INFER] Keeping SkyWater’s name is brand architecture doing work — a deliberately separated identity maintained exactly where separation carries commercial value, inside an otherwise unified company. That is a more sophisticated outcome than blanket absorption would have been, and a more useful one: it lets IonQ own the manufacturing base while letting SkyWater sell to companies that would never buy from IonQ.
3.7 — Raymond, Cardillo and Muller: The Row That Should Not Be Possible
The day closed with three men on one panel, and the composition of that panel is a strategic disclosure in its own right. Read their records together.
Panelist | Record as stated on stage |
Gen. John W. Raymond | Approximately thirty-five and a half years in the Air Force, transitioning to the Space Force in 2019. Led the establishment of the United States Space Force and served as its first Chief of Space Operations, holding four-star rank in both services. De Masi additionally described him from the stage as the only officer to hold that rank in two services since 1947 — a characterization this report reports rather than asserts [de Masi 2:27:33; Raymond 2:30:47]. |
Robert Cardillo | Nearly four decades in the US intelligence community, spanning seven presidents, including four years providing intelligence in the Oval Office, ending as Director of the National Geospatial-Intelligence Agency. An IonQ director since 2024 and Executive Chairman of IonQ Federal [de Masi 2:27:33; Cardillo 2:33:00]. |
Rick Muller | Roughly twenty-one years leading the quantum program at Sandia National Laboratories, then Director of IARPA, before joining IonQ. The only one of the four men on that stage holding a doctorate in ion trapping [de Masi 2:27:33; Muller 2:35:04; de Masi 2:37:13]. |
Domain, mission, transition. A company can hire one of these people. IonQ has all three — and all three came to say the same thing about the same company. |
WHAT THAT PANEL ACTUALLY SAID These three men cover the three things a company must have to sell quantum into national security, and they are rarely available together at any price. Raymond is the domain. He built the military service that owns space, which is where sensing, timing and eventually quantum links have to operate. Cardillo is the mission. He ran the agency that turns overhead imagery into decisions, which is precisely the customer for the SAR result IonQ disclosed that afternoon — and he stood up and said so, pointing at Shapiro’s slide and calling it a coherence-from-chaos service [Cardillo 2:39:32]. Muller is the transition. He spent two decades funding quantum research at a national laboratory and then ran the agency that decides which intelligence-community research programs graduate into use. He knows, from both sides of the table, what makes a technology transition and what makes it die in the lab. Domain, mission, transition. A company can hire one of these people. Hiring all three, and having all three appear together to describe the same company in the same terms, is a statement about access that no competitor in this sector can currently match. |
What they each chose to say is as informative as their presence. Raymond spoke about institution-building, comparing IonQ's growth to standing up the Space Force by integrating people from the Air Force, Army, Navy, Marines, colleges and off the street into one high-functioning team, and naming integrity and mission focus as what attracted him [Raymond 2:37:39]. Cardillo spoke about outcomes, arguing that what intelligence professionals do does not matter unless it affects a better decision, and that IonQ was the only company he saw bringing all the pieces together [Cardillo 2:33:00]. Muller spoke about how government actually buys — that the government does not want a qubit, it wants mission solutions to its hardest problems, and that technologies alone do not get you far without people who understand the mission well enough to value them [Muller 2:35:04].
Cardillo also delivered the single most consequential governance statement of the day, unprompted, from the board seat. On the decision to publish the cryptanalysis result, he credited the team for conversations Muller led at the senior-most levels of government, on the explicit grounds that IonQ did not want to surprise its partners and wanted to be responsible on the safe-cracking side — and said he believes it raised the company's reputational credibility with its most important customers [Cardillo 2:43:39].
[ARG] That is the panel working as designed. Muller had the relationships to make the calls, Cardillo had the standing to judge whether they were the right calls, and Raymond had the institutional experience to say whether the organization making them is built to keep making them. A company that publishes a working attack on elliptic-curve signatures and comes out of it with more government trust than it went in with is demonstrating a capability, not just good manners — and that capability lives in the three men who closed the show.
Raymond's closing line is the one to keep. Taking the brains of the people the room had heard from, translating them into real capabilities, and putting them in the hands of the warfighter, he said, is nationally critical — and that is what IonQ does really well [Raymond 2:44:40].
3.8 — The Honest Limit on This Section
Panels are assembled and stages are curated, and this section should not be read as though the format were accidental. IonQ chose who spoke and in what order, and a well-run investor day is designed to project exactly the coherence described above.
Two things survive that discount. The first is that the cross-references in Section 3.4 were mostly unscripted — they occurred in Q&A, in interjections, and in ad-libbed segues, which are the parts of an event a communications team does not control. The second is that career records cannot be staged. Raymond founded the Space Force, Cardillo led the NGA, Muller led IARPA, Ballance co-founded Oxford Ionics, Sonderman led SkyWater, Bhaskar co-founded Lightsynq. Those facts are true regardless of the seating chart, and all six of those people now work toward one roadmap at one company.
4. The Stack, Layer by Layer
Section 1 sized the reassessment. Section 2 proved it. This section is the underlying evidence, layer by layer, in the words of the people who own each one. The ordering runs bottom to top — silicon to solution, Ballance’s own phrase for how IonQ works and the reason for this report’s title [Ballance 18:36]. A reader who wants to challenge the thesis should start here, because this is where the load is carried.
From silicon to solution, every layer had evidence — and the evidence was dated. |
Layer 1 — Sovereign silicon
[FACT] The deck also quantifies the parallelism and names it. Five quantum devices are being prototyped simultaneously: 256-qubit trap devices, 10K trap devices, low-loss wideband silicon nitride photonic integrated circuits, ultra-high-efficiency fibre-to-chip optical couplers, and quantum-memory-control devices. Alongside the twelve-times wafer-lot figure the deck claims a better-than-threefold acceleration in 256-qubit fab-out cycle times and more than a hundred learning cycles per year expected.
[FACT] Speaker notes left in the published file supply the underlying arithmetic, and they show the headline is understated rather than inflated. The note records roughly 25 lots start-to-finish over the six months from June to December 2026, against roughly 2 lots in a six-month period at the previous trap fabs. That is 12.5 times, which the slide rounds down to twelve; the note goes further and argues that even a tenfold statement would be reasonable and justifiable. The working was left in the file, and it is more conservative than the claim made from it.
The three sites and what each does: Texas as the high-volume 200mm foundry with a 114,000 square foot cleanroom; Florida for advanced packaging at 36,000 square feet; Minnesota for technology-as-a-service manufacturing and CMOS integration at 91,000 square feet. Florida and Minnesota both hold DMEA Category 1A Trusted Supplier accreditation — which is the substance behind the trusted-fab claim, and the part a competitor cannot acquire quickly.
Sonderman's framing of the transition, in IonQ's own words: physics problems from lab, engineering problems to fab — units of production moving from one trap and one system to dies per wafer and lots per month. This is the single slide that best states the argument in Section 1.1.
Sonderman's core claim is that the industry's long-standing problem has been getting from the lab to the fab, and that SkyWater was literally created to solve it through a technology foundry model and a technology-as-a-service business model that bridges the valley of death [Sonderman 31:11]. The differentiator he names is customization: where the specialty foundries chose standardization, SkyWater embraced customization and built an environment where innovators could have ideas created in a volume environment and run at scale [Sonderman 31:11].
The scale disclosure is the one that changes the model. Thousands of quantum wafers run in the fab today, of which a third are tied to IonQ, meaning the company is designing for manufacturability and will be ready when it is time to run at scale [Sonderman 31:11]. Nine quantum customers now work in SkyWater, including Qolab as of the day's announcement — and Sonderman was explicit that this is not simply driving IonQ's roadmap but literally creating a quantum foundry industry that will be foundational through the rest of the decade [Sonderman 31:11].
Asked directly by StoneX how the other eight customers view the conflict of interest, and how much lead time exists against IBM and GlobalFoundries, Sonderman answered on both counts: information is compartmentalized, there is no cross-contamination, and every customer should be a hundred percent confident their roadmap will not be impeded by IonQ's needs; moving fast with IonQ populates those capabilities across the other customers, on a rising-tide logic [Mobley 1:13:05; Sonderman 1:13:58]. On competition, he noted that IBM sold its microelectronics business to GlobalFoundries in 2015, and characterized GlobalFoundries — where he previously worked — as having capabilities but not being an innovation engine, a scaled specialty foundry that loves standardization [Sonderman 1:13:58].
Singh added the procurement view from the buyer's seat. During the Oxford Ionics acquisition, other candidate quantum foundries were evaluated; they could not move fast enough, did not understand what was needed, had no learning curve on quantum, were blown away by the future volume numbers, needed permission from the mothership, and simply moved too slowly [Singh 1:12:19].
The release gives the business its name and its shape. SkyWater Quantum Solutions was launched as a merchant foundry service supporting quantum companies from process development through manufacturing, working with trapped-ion, neutral-atom, photonic and superconducting computing companies as well as quantum networks and sensors. Its technology portfolio is broader than the two platforms described on stage: the SP90 photonics platform, the SC250 superconducting platform, advanced packaging, and manufacturing processes involving diamond and lithium niobate. Mihir Bhaskar — a co-founder of Lightsynq who subsequently led research and development at IonQ — runs the business.
[INFER] Giving the merchant foundry a brand, a named leader, a catalogued process portfolio and its own launch release is not how a company treats a capacity-utilization decision. It is how a company treats a business line. This resolves the ambiguity left open on stage and is the basis for treating merchant supply as a distinct ecosystem in Section 6.1.
[FACT] The trusted-fab argument was stated end to end: security from design through engineering, manufacturing and delivery to the customer, because customers need to be able to trust it — with SkyWater described as the only trusted fab of scale in the United States and competitors targeting 2029 [Singh 47:01].
Layer 2 — Qubit control and the device
The most useful slide in the deck for anyone tracking delivery: the Superion 256 path with a marker on September 8. Chip design, first fabrication, prototype and component testing, first ion trapped, dynamic circuit execution with mid-circuit measurement, 99.99% gate fidelity and cloud early access sit behind the marker. Quantum multiplexed I/O, performance tuning, integrated system validation, first customer system build and production start sit ahead of it. Five named steps remain before a customer system exists.
The cost curve underneath the roadmap. Roughly a fourfold cost-per-qubit reduction across three laser-controlled generations, against roughly ninety-five fold across three generations of semiconductor-based electronic qubit control — more than 300x in the release and 330x in the deck, alongside a roughly 400x increase in physical qubit count from Aria to Superion 10K.
Ballance built the Superion case on three constraints learned from five generations of shipped machines: radical simplification, because only the simple scales; learning over product cycles of months rather than years; and platforms where each generation builds on the last to tackle problems that cannot be tackled otherwise [Ballance 18:36].
The specific move is replacing lasers with electronics. Natural qubits have historically been controlled with lasers, which is excellent for the zero-to-one phase of getting first hardware built in a lab and challenging to mass manufacture; Superion replaces laser control with electronics built into the chip, removing the entire middle layer and connecting classical control electronics directly to the quantum chip, allowing the chips to be built in standard semiconductor ways [Ballance 18:36].
Electronic qubit control alone is insufficient, and Ballance was careful about this. The analogy he used is that there is a tremendous amount of technology between a single LED and an LED pixel array in a television display, and the equivalent technology for going from single unit cells to unit cell arrays is what IonQ calls quantum multiplexed I/O, around which it has built a unique patent portfolio [Ballance 18:36]. The unit cells achieve better than four nines on two-qubit gate fidelity, and the multiplexing is what takes them into scaled arrays [Ballance 18:36].
Muller supplied the modality argument from the national-security seat, recalling Chris Monroe's case for ions: they are naturally identical qubits and naturally quantum particles that can be controlled classically, which is what enables low-cost, low-power solutions that scale rapidly and can be deployed locally or remotely as the customer's own solution rather than an anonymous hosted service [Muller 2:41:49].
Layer 3 — Packaging, tiling and integration
The scaling ladder with a technique named at every rung: cryogenic CMOS to 10,000, tiling to 20,000, die shrink to 200,000, and 2.5D IC packaging beyond a million. Every step is a semiconductor integration technique rather than a physics advance, which is the whole of the argument in Section 1.1 expressed as a roadmap.
This was the layer most likely to be soft, and it received the most searching question of the day. Ellis of B. Riley asked what technology competencies would be needed to execute tiling at end-of-decade scale [Ellis 1:22:45].
Ballance's answer began with a point of pride that it is not a quantum problem but an integration problem — die bonding with tight enough tolerances, and designing the transfer, which is an electrostatic phenomenon, physically pushing qubits over the junction [Ballance 1:23:45]. He then gave the status: thorough studies of industry-standard tolerances have been done and IonQ sits comfortably within them, not pushing the limit of any one thing [Ballance 1:23:45]. The challenge he named is a classic technology system-integration problem — pushing up yields, making the packaging boring, ordering value-add steps sensibly — and these are classic semiconductor industry problems that SkyWater has focused on understanding [Ballance 1:23:45].
Bolton of Needham pressed the adjacent question on fidelity data across all 256 qubits and on crosstalk mitigation [Bolton 1:19:09]. Ballance described the mechanism rather than the date: a tremendous amount of work over twelve months building fast-turn prototype de-riskers, which IonQ calls short loops, to learn each element in isolation and decouple it, with many of those short loops displayed at the side of the room; the result is confidence in the ultimate integrated performance, with current work focused on integration, iteration, and validating production scale ahead of first deployments in early 2027 [Ballance 1:20:02].
[UNDISC] No date was given for when median gate fidelity data across the full 256-qubit device would be available. This is the single most important undisclosed technical item from the event and is carried into the calibration section.
Sonderman placed the packaging capability geographically: front-end capability in Minnesota, packaging in Florida, and the ability to scale in a post-Moore's-law realm through heterogeneous integration [Sonderman 31:11]. On Fab 25, he described it as part of the foundry ecosystem giving significant purchasing power and supply-chain optimization, with quantum work sitting in Minnesota and Florida [Sonderman 1:21:05]. Singh characterized Fab 25 as optionality that would require retooling if repurposed, which has not been necessary [Singh 1:22:07].
Layer 4 — Architecture and error correction
de Masi positioned Walking Cat as the world's first fault-tolerant architecture that is shovel ready, published in April, and done to a level of detail the world had never seen before [de Masi 4:40]. Ballance gave the technical content: the most detailed full-stack quantum computing architecture available, carrying the theme of radical simplification from hardware through error correction to compiler and application, including going from multiple different types of error-correction code to one unified code — simplifying the architecture, simplifying the compute fabric, and allowing faster scaling [Ballance 18:36].
The consequential statement is that this architecture and the compute fabric shown in the 256 system take IonQ to its ten thousand qubit system, with the big addition being integrated cryogenic CMOS to implement the multiplexing — and the de-risking of that CMOS is already built at SkyWater, with first wafers displayed in the room [Ballance 18:36]. Beyond that, two dies are tiled in 2.5D integration, and hundreds of thousands of qubits follow from tiling more dies and increasing per-die density [Ballance 18:36].
The release adds the development sequencing, which was not stated from the stage: Superion 10K is being developed at the same time as the 256-qubit system rather than after it. It will be the first hardware generation designed to run the Walking Cat architecture, adding cryogenic CMOS to the quantum chip, with early cryo-CMOS test chips displayed at the event. IonQ is targeting a laboratory demonstration of fault tolerance using the technology in 2027, followed by a manufacturable commercial system in 2028.
[FACT] Those are two new dated commitments — a 2027 laboratory fault-tolerance demonstration and a 2028 manufacturable commercial system — neither of which was given verbally. Both belong in the Delivery Ledger, and both are carried into Section 10.
On the error-correction side specifically, Ballance reported customized quantum error-correction codes running on real hardware with breakeven performance in the preceding six months [Ballance 18:36] — the threshold at which error correction stops costing more than it saves.
Layer 5 — Classical control and compilation
The layer most often assumed to be a downstream problem was addressed with a dated, recent result: within the prior couple of weeks IonQ put out the first mega-quantum-operation real-time decoding stack tailored to its ten thousand qubit device, showing it can be run using efficient classical computation in parallel with the quantum hardware [Ballance 18:36].
That result and the Shor's compilation are the two halves of the same claim — that the classical software surrounding the quantum device is being built ahead of the device rather than behind it. Ballance's own framing was that working across the full stack gives a virtuous cycle: solutions can be tailored to the hardware, and hardware and software can be tailored to see around the corner because the team knows what the next generation of quantum solutions will demand [Ballance 18:36].
Layer 6 — Interconnect and photonic integration
Beyond the kilohertz result covered in Section 2.2, Bhaskar disclosed the structural move: the launch of the industry's first dedicated quantum foundry platforms, which he distinguished from SkyWater's existing decade of quantum wafer delivery as going far beyond it — scalable, reproducible platforms creating an open ecosystem for quantum computing companies building atom, ion, superconducting, photonic or spin-based processors [Bhaskar 1:31:00].
The platform names from the release are SP90 for photonics and SC250 for superconducting routing, alongside advanced packaging and diamond and lithium niobate processes. The stage rendering of the superconducting process name should be read against the release.
His argument for why two platforms suffice is worth preserving because it is the cleanest strategic reasoning offered all day. There are only two physics ways to route quantum signals — light and superconducting electronics — just as there is copper and fiber in the data center; there is a quantum analog, and IonQ has the world's only quantum platforms for semiconductor manufacturing [Bhaskar 1:31:00].
On the photonics side, Professor John Bowers was present in the audience and credited as having pioneered putting laser systems on a chip, without which the massive scale-out of today's AI data centers and the optical interconnects connecting GPUs would not exist; Nexus has pioneered that same technology for putting quantum lasers on chip [Bhaskar 1:31:00].
Asked by Mizuho how Nexus fits the IonQ and SkyWater roadmap, Bhaskar described it as another one-plus-one-equals-thirty story: SkyWater has the ability to manufacture and Nexus brings a proven pedigree and intellectual property for design and integration of integrated photonic solutions, letting IonQ take the complexity of hand-assembled systems and integrate those components on a chip [Vijay 1:27:01; Bhaskar 1:27:36]. He was specific that the first impact is in sensing, where clocks and inertial navigation sensors are still built from discrete elements with wires coming out and hand assembly [Bhaskar 1:27:36]. Shapiro noted that an example photonic integrated circuit for the sensing unit was in the showcase [Shapiro 1:28:48].
Layer 7 — Networking and security
Shapiro's opening framing was that the advances on the computing side have implications not just for quantum computing but for quantum security, and that two trends are converging [Shapiro 36:47]. The first is the falling qubit requirement for Q-Day. The second is that governments are pulling migration requirements forward — most recently, in June, the US executive order pulled the migration timeline from 2035 to 2030 [Shapiro 36:47]. His conclusion was that leaders who are not transitioning now are already too late, and that harvest-now-decrypt-later means exposure exists today, not at some future date [Shapiro 36:47].
The product stack he described has four elements: quantum security posture management giving visibility into where an organization is vulnerable; a CNSA 2.0 and NIST-compliant path to update infrastructure using post-quantum cryptography; quantum key distribution as a physical solution spanning a network where it matters most, using the same underlying quantum mechanics to protect against eavesdropping; and orchestration and key management to evolve the architecture as standards change [Shapiro 36:47].
The commercial proof was a major agreement with Congruity360 for enterprise rollout of quantum security in the US — to IonQ's knowledge the largest deal of its kind in the United States, spanning enterprises, healthcare systems, financial institutions, government organizations and telcos — alongside existing rollouts with Singtel and SK Broadband [Shapiro 36:47].
On networking, Shapiro restated the AFRL work: the first company in an enterprise setting to interconnect two quantum computers, followed by providing AFRL with a quantum network for entanglement distribution that allows experimentation with different modalities of quantum computers on the same network [Shapiro 36:47]. The next step is to take a network that is fiber-based today and use IonQ's optical communication technology to get off the fiber and operate in free space, which unlocks quantum connections that are airborne and in the future space-based [Shapiro 36:47].
de Masi framed the resulting posture as defense in depth across hardware and software, with the safe-cracker-should-build-your-safe logic, and noted IonQ delivered news on the safe-breaking front that morning while offering unique advantages on the safe-making side [de Masi 4:40]. Singh gave the origin story from the board seat: creating machines that go beyond what humanity has seen requires responsibility, so when de Masi set the company's mission he decided it was important to have quantum cybersecurity — a very different game from any cybersecurity seen so far, because it protects against a quantum computer [Singh 47:01].
Layer 8 — Sensing and PNT
Shapiro described a portfolio built on the same physics: the world's most accurate commercial clocks, which are quantum optical clocks, plus time-transfer devices, atomic gravimeters that measure specific gravity at any point on Earth, and gyroscopes measuring inertial movement — a precise sensing capability especially relevant to positioning, navigation and timing [Shapiro 36:47]. The performance claim is that IonQ's quantum clocks are a thousand times more accurate than the best-in-class classical equivalent and lead the entire field in NIST's ensemble of clocks, being the best-performing clocks NIST has available in this form factor [Shapiro 36:47].
Deployment is not hypothetical: sensors are deployed at sea, in space, on the X-37B space plane, on land and in the air [Shapiro 36:47]. Government validation spans the platform — DARPA QBI for computing, sensing awards including the It's About Time program, networks, and a recent contract with the NRO on the space side [Shapiro 36:47]. de Masi named the three DARPA programs as HARQ, About Time and QBI [de Masi 4:40].
Singh gave the demand rationale in plain terms: certain government customers value sensing that cannot be jammed, GPS jamming is real, planes are walked off course and land in the wrong place, and that technology has become broadly available and is being used in conflicts more and more [Singh 47:01].
Layer 9 — Applications and solutions
Chad reframed the measurement problem first: customers very rarely ask about two-qubit gate fidelity or coherence time; they ask how long it will take to get to a solution valuable for their workload and what the economics are — time to solution, economics to solution [Chad 1:42:36]. The positioning he offered is sniper versus shotgun, with results sometimes at parity and sometimes 10×, 100×, 1,000× or 10,000× better, and — importantly — with the repository published so anyone could duplicate the results, alongside third-party validation [Chad 1:42:36].
The named workloads connect the applications layer back to the compilation layer. One of the workloads where IonQ is roughly 1,000 to 10,000 times faster is the quantum Fourier transform, which shows up in Shor's algorithm; quantum phase estimation shows up in chemistry and life-sciences use cases [Chad 1:42:36].
The life-sciences detail was the most specific: work with AstraZeneca continuing at three iterations; protein-folding work where oxytocin has nine amino acids and GLP-1s are roughly 30 to 50, with Forte Enterprise and Tempo hardware used on 12 then 14 amino acids and an August Synopsys update taking Tempo from 14 to 16 [Chad 1:42:36]. Einride logistics work was brought across to healthcare with CCRM, and quantum pattern-finding was applied to fMRI data for early-onset brain cancer and dementia [Chad 1:42:36].
The academic validation was quantified: at IEEE Quantum Week, 857 papers were submitted and 20 selected as best papers; IonQ submitted 13, 10 entered, and four of the ten won best in show — which he characterized as a 3.7-sigma event [Chad 1:42:36]. One example was quantum algorithms improving LLM behavior with 24 percent better accuracy, lower time to solution and better energy efficiency [Chad 1:42:36]. A paper with NVIDIA and Oak Ridge covered using LLMs to improve quantum algorithms, with GPU and QPU clusters working hand in hand [Chad 1:42:36].
The commercial anchor was cytochrome P450, which is not solvable today on classical methods exactly right and represents about two to three billion dollars a year of research; a 2025 Google/Caltech paper sits at a scale beyond a Superion 20K, but the work could be done with a 100K, or with five 20K systems in parallel, pulling a workload forward two years and worth hundreds of millions in net present value — with networking not a prerequisite but the pull-forward carrying immense economic value [Chad 1:42:36].
The commercial state of the Superion line was disclosed only in the release: orders are open, production systems will also be available through IonQ's cloud service, and IonQ sold its first 256-qubit system during the first quarter of 2026, before the platform was publicly named. The Superion launch release does not identify the buyer; the Q1 2026 results release describes the sale and the partnership around it, and contemporaneous reporting identified the customer as the University of Cambridge.
The organizational build behind it: the customer-engagement team more than doubled in the first half to about 70 people worldwide across forward-deployed engineering, customer solutions and applications R&D, almost entirely PhDs [Chad 1:42:36].
Layer 10 — Distribution, trust and governance
Singh described the commercial architecture as land and expand, so it does not matter where the discussion starts — increasingly it starts with security — and the question becomes how many products a customer uses [Singh 47:01]. He offered the Cisco of quantum as his preferred analogy over the Amazon of quantum, on the logic of putting together things that work together [Singh 47:01]. On pricing dynamics, he noted that by the time IonQ reaches a customer it is usually not a competitive situation because the modality has already been chosen, so the conversation becomes value and use cases, with forward-deployed engineers and app developers moving customers from the cost side to the revenue side of their equation [Singh 1:28:56].
Chad added the deployment-choice dimension: customers like choice of language, with all seven major open-source stacks supported, and choice of where — cloud, on-premises or sovereign [Chad 1:42:36]. He also reported the upgrade pattern that a platform thesis requires: customers who bought Aria and Forte bought Tempo and want Superion cloud access [Chad 1:42:36].
The governance layer is not conventionally counted as part of a technology stack, and in this market it should be. Cardillo described the handling of the Shor's result as the essence of being a good partner: Muller led many conversations with the senior-most levels of government, because IonQ did not want to surprise its partners and wanted to be responsible going forward on the safe-cracking part, which Cardillo believes raised the company's reputational credibility with its most important customers [Cardillo 2:43:39].
[ARG] For a company whose computing business sells to governments and whose security business sells against governments' adversaries, the ability to publish a cryptanalysis result without damaging either relationship is a genuine capability. It is not obvious that any competitor currently has the institutional relationships required to do the same thing.
5. Nine Witnesses Whose Standing Does Not Come From IonQ
The strongest evidence at this event was not anything IonQ said about itself. It was that nine parties, each with standing that IonQ did not confer, stood behind a different layer of the stack, on the record, in the same room, on the same afternoon. A distinction matters here and this report states it plainly: seven of the nine have no IonQ relationship of any kind. Two — Raymond and Cardillo — hold IonQ positions, Cardillo as a director since 2024 and Executive Chairman of IonQ Federal, Raymond as a director. Their standing was earned at the Space Force and the National Geospatial-Intelligence Agency, and it is real; their testimony is corroboration from inside the company rather than from outside it, and it is weighted accordingly here and throughout Part Two. The relationship column below states each one.
Nine witnesses. Nine different layers. No gaps — six of them with no IonQ relationship at all, and one holding a Nobel Prize in the competing qubit modality. |
Witness | Layer attested | Relationship to IonQ | What they actually said |
Dr. John Martinis — Nobel laureate; co-founder and CTO, Qolab | Silicon and fabrication | None — competitor | That the field is entering a new phase where most of the science is understood and the task is building a manufacturable system that works better scaled up — a transition that will happen if scientists start collaborating with people in the manufacturing space doing semiconductor fabrication [Martinis 1:37:17]. |
Alan Ho — CEO, Qolab | Foundry unit economics | None — competitor | That Qolab is proud to be an initial customer for the SC250 process, and that SkyWater provides predictable unit economics for manufacturing, which is required for investors to underwrite manufacturing at scale [Ho 1:40:01]. The release adds that the agreement is multiyear and moves Qolab's Quantum System-in-Package devices from custom development onto standardized SC250 wafer services, with SkyWater investing in dedicated Minnesota equipment upgrades. |
Professor John Bowers — pioneer of lasers on chip | Photonic integration | Co-founder, Nexus Photonics — acquired by IonQ | Present in the audience and credited by Bhaskar as the originator of the technology underpinning today's AI data-center optical interconnects [Bhaskar 1:31:00]. |
Alessio Butti — Undersecretary of State to the Presidency of the Council of Ministers for Technological Innovation | Sovereign distribution | None | That Italy adopted its first national quantum strategy under his direction, covering computing, communication and security, sensing, research, skills and industrial ecosystem — and welcoming IonQ Italia by name as reflecting a collaborative approach [Butti 2:21:37]. |
Robert Long — President of Strategic Initiatives, EPB | Networking and applications | Customer | That EPB operates the first nationally available quantum network and the first commercially available quantum computer, is writing its own algorithms and running its own programs for energy-grid optimization and energy security, and is expanding into quantum memories [Long 2:02:32; Chad 1:42:36]. |
Dr. Matthew Rabinowitz — Founder and Executive Chairman, Natera; MyOme | Applications and AI | Customer / collaborator | That his team emulated an IonQ paper's architecture and saw model power go from 0.70 to 0.83 AUC using a six-qubit then fourteen-qubit quantum fine-tuning head on an approximately 800-length embedding — explicitly a simulation of quantum hardware [Rabinowitz 2:12:54]. |
John Lokada — Futures Director, ServiceNow | Enterprise workflow | Customer | That security and optimization can no longer be two separate lanes and that secure optimization is the interesting path, with ServiceNow looking at quantum plus AI and having acquired Armis and Veza [Lokada 2:01:46]. |
Robert Cardillo — sixth Director of the NGA; IonQ director | Intelligence analysis | Director since 2024; Exec Chairman, IonQ Federal | That what excites him is the applications, customers and missions, because what intelligence professionals do does not matter unless it affects a better decision — and that IonQ was the only company he saw bringing all the pieces together [Cardillo 2:33:00]. |
Gen. John W. Raymond — founder of the US Space Force; IonQ board | Institutional execution | Board member | That what attracted him was the team, operating with the same integrity and mission focus he saw in the Space Force, and that translating this science into capability in the hands of the warfighter is nationally critical [Raymond 2:37:39; Raymond 2:44:40]. |
5.1 — A Competitor With a Nobel Prize Endorsed the Acquisition on the Acquirer's Stage
Martinis and Ho are not IonQ customers in the ordinary sense. Qolab, based in Madison, Wisconsin, builds superconducting qubits — a directly competing modality to trapped ions. Martinis, who was a co-developer of superconducting qubit technology in the 1980s, said he is very excited that IonQ acquired SkyWater and that IonQ is helping with extra focus on quantum computing, that this is a very good partnership and acquisition, and that it helps SkyWater build new capabilities in fabrication to properly scale up quantum computers and build them properly [Martinis 1:37:17]. He closed by saying he looks forward to working with SkyWater not simply to build a better and bigger quantum computer but to lay the industrial foundation for a new era of computing [Martinis 1:37:17].
[ARG] Understand what this actually is. A competitor endorsed a rival’s acquisition, on the rival’s stage, at the rival’s investor day, in front of the rival’s investors — and then signed a multiyear manufacturing agreement putting his own company’s roadmap on that foundry. That is only rational if he believes the acquisition improves the shared manufacturing base more than it advantages the acquirer. No assurance IonQ could write is worth as much as that signature, and no competitor of IonQ’s has one like it.
Bhaskar drew the strategic conclusion explicitly, suggesting investors ask themselves where the leaders and luminaries in the field are choosing to build their quantum technologies — and answering that whether it is Bowers in the audience or Martinis, it is happening at SkyWater [Bhaskar 1:39:29]. Ho's geographic point extends it: Qolab does all its design, packaging and cryogenic testing in Madison, Wisconsin, and having a quantum manufacturing ecosystem in the Midwest makes for a resilient quantum manufacturing ecosystem for America [Ho 1:40:01].
5.2 — The Customer Who Writes Its Own Code
EPB is the most probative customer testimony available, because EPB is not merely buying — it is operating. Long described a full fiber network in the Chattanooga area with the IonQ Forte Enterprise system being commissioned around the current quarter, making both network and computer available to the community and nationally, focused on energy-grid optimization and energy security while using essentially the full stack of IonQ services [Long 2:02:32]. Eight fellows funded via grant are trained on the IonQ training platform, and EPB writes its own algorithms and runs its own programs [Long 2:02:32].
The history matters for the networking layer. Ten years ago EPB partnered with Oak Ridge National Lab, Los Alamos National Lab and Qubitekk — now in the IonQ family — and successfully tested QKD across a 21-kilometer span on its fiber network [Long 2:08:09]. The ecosystem being built on top now includes an IonQ office and R&D facility in Chattanooga, a Vanderbilt satellite office bringing 250 faculty or staff over three to five years, UT Chattanooga partnerships, and supportive city, county and state government [Long 2:08:09].
[FACT] A customer that trains its own fellows and writes its own algorithms is evidence that the applications layer is usable by non-specialists. That is a different and higher bar than a vendor-run demonstration.
5.3 — The Customer Who Reproduced the Result
Rabinowitz's contribution was the only independent reproduction described at the event. He referenced a paper out of IonQ in May on fine-tuning a large language model to classify review sentiment, said his team emulated a very similar architecture, and reported the outcome: an approximately 800-length vector embedding from a standardized LLM, tested with six qubits on the quantum output layer as the fine-tuning head and then with fourteen, taking model power from 0.70 to 0.83 AUC [Rabinowitz 2:12:54]. He was scrupulous that this was a simulation of the quantum hardware, and framed the opportunity as adding more qubits to the fine-tuning head of existing large models to build disease-predictive models [Rabinowitz 2:12:54].
His conceptual argument for why quantum matters to AI is worth recording because it reframes the entire hybrid discussion. Taking protein folding as a cartoon example, a hundred amino acids with four permutations between each gives four to the hundredth hypotheses to evaluate — computationally non-tractable [Rabinowitz 2:03:45]. Classical systems like AlphaFold solve this but not robustly; what is incredible is that in nature the protein always finds the global minimum, by a combination of quantum tunneling and jiggling, and does not get stuck in a local minimum the way today's AI algorithms do [Rabinowitz 2:03:45].
The applications he named are specific and commercially live: protein binding, neoantigen prediction for personal cancer vaccines, predicting how patients respond to therapies, and prognostic applications [Rabinowitz 2:03:45]. Natera's data position underneath them is described as catching cancer recurrence years or months before clinical symptoms, seeing therapy response better than imaging, and determining in the adjuvant setting whether all cancer has been cleared and who will benefit from chemotherapy — with less than one percent of tumor mutations making good neoantigens, which is precisely the selection problem quantum algorithms might improve [Rabinowitz 2:12:54].
Each of these nine is examined in full in Part Two, where their remarks are put under a structured set of questions rather than excerpted.
5.4 — What the Analysts Chose to Ask
Seven analysts had the floor. Not one asked whether the physics works. |
Seven questioners took the floor, six of them named. Not one asked whether the physics works, whether fidelity claims are real, or whether trapped ions can scale — the questions that dominated this sector's coverage two years ago.
Analyst | What the question reveals about where the doubt now sits |
John McPeak, Rosenblatt [1:08:17] | At what qubit count photonic interconnect becomes necessary — an architecture-sequencing question that presumes the architecture works. |
Nihal Choksi, Northland [1:09:29] | How an eight-month-to-two-month cycle-time reduction was achieved, and at what lithography node — a manufacturing-process question. |
Gary Mobley, StoneX [1:13:05] | How the other eight foundry customers view conflict of interest, and lead time against IBM and GlobalFoundries — a competitive-moat question. |
Unidentified [1:16:18] | Whether there is a scale-out roadmap for photonic interconnect phases three and four, and whether manufacturing capability exists — a capacity question. |
Quinn Bolton, Needham [1:19:09] | When median gate fidelity data across 256 qubits will exist, crosstalk mitigation, and plans for Fab 25 — the sharpest technical question of the day, and still a validation-timing question rather than a feasibility one. |
Craig Ellis, B. Riley [1:22:45] | What competencies tiling requires at end-of-decade scale, and how customers engage across the three technology areas — a scaling-execution and commercial-motion question. |
Vijay, Mizuho [1:27:01] | How Nexus fits the roadmap, 256 pricing, and how to model fiscal 2027 — pure commercial modeling. |
[ARG] The migration of analyst doubt from physics to manufacturing and modeling is itself a data point about where informed external observers now believe the risk sits. It aligns exactly with the recalibration argued in Section 1, and it was arrived at independently by seven questioners whose incentive is to find the weak joint.
6. What a Closed Stack Unlocks, Ecosystem by Ecosystem
IonQ entered no new markets on September 8. That is not the point. A working stack makes opportunities addressable that were theoretical while the stack was open, and twelve ecosystems change value as a result. Each section below states what is now addressable, which layers it draws on, and who evidenced it.
6.1 — Semiconductor Foundry and Merchant Supply
Vertical integration expressed as throughput: twelve times more wafer lots delivered over six months than at the prior foundry, a design cycle compressed from nine months to two, and the 330x cost-per-qubit reduction the roadmap depends on.
The opportunity: to be the manufacturing bottleneck for an entire industry, and to be paid by everyone in it regardless of which modality wins.
This is the ecosystem that changed most today, and it changed because of a single number. Two thirds of the thousands of quantum wafers running at SkyWater are for companies other than IonQ [Sonderman 31:11]. Nine quantum customers are in the fab [Sonderman 31:11]. Singh acknowledged what is rarely discussed — that IonQ sells components to other quantum computing companies that already depend on it, and manufactures some of what they need in the SkyWater quantum foundry, with the intent to keep doing so [Singh 47:01]. de Masi named merchant supply as a business grown decisively this year, supplying leading players in the friendly quantum ecosystem with products including the world's most accurate atomic clocks, across the US, the Five Eyes and the allied world [de Masi 2:45:40].
Two thirds of the quantum wafers running in the only quantum foundry at scale in the world are for somebody other than IonQ. |
The release turns disclosure into a business. SkyWater Quantum Solutions was launched as a merchant foundry service running from process development through manufacturing, serving trapped-ion, neutral-atom, photonic and superconducting computing companies plus quantum networks and sensors, with a technology portfolio spanning the SP90 photonics platform, the SC250 superconducting platform, advanced packaging, and diamond and lithium niobate processes — run by Mihir Bhaskar, a Lightsynq co-founder who subsequently led R&D at IonQ.
Qolab is the launch customer and the proof of the model. Under a multiyear agreement, the Madison-based superconducting company will move its Quantum System-in-Package devices from custom development onto standardized SC250 wafer services, with SkyWater investing in dedicated equipment upgrades at its Minnesota fabrication facility. SkyWater stated it will use information-security controls, IP protections and internal program separation to protect customer projects.
[ARG] The migration from custom development to standardized wafer services is the single most important detail in this ecosystem, and it was absent from the stage. Bespoke foundry work does not scale and does not produce predictable margins. A catalogued process that a customer can order against does both — and it is exactly the predictable unit economics Ho said investors require to underwrite manufacturing at scale [Ho 1:40:01]. SkyWater putting its own capital into Minnesota equipment for a competitor's process is the strongest available evidence that this is a business rather than an accommodation.
What makes this newly addressable rather than merely newly disclosed is the platform launch. Bhaskar's two dedicated quantum foundry platforms — photonics and superconducting routing — turn bespoke foundry engagements into productized processes explicitly open to atom, ion, superconducting, photonic and spin-based processor companies [Bhaskar 1:31:00]. Ho's testimony explains the commercial mechanism: predictable unit economics are what allow investors to underwrite manufacturing at scale [Ho 1:40:01]. A foundry that can offer that to nine customers can offer it to ninety.
[ARG] This is the most attractive structural position in the sectorand the least appreciated. Every other IonQ ecosystem improves when competitors do worse. This one improves when competitors do better and build more. It is the only line in the company whose revenue is uncorrelated with the trapped-ion-versus-everything-else question that dominates sector debate — a genuine hedge sitting inside what the market still prices as a pure-play. Investors who buy IonQ for trapped ions are getting something they have not paid for.
Investors who buy IonQ for trapped ions are getting something they have not paid for. |
The countervailing requirement is trust, and it was addressed directly rather than avoided. Sonderman's compartmentalization commitment was made to competitors sitting in the room [Sonderman 1:13:58], and de Masi restated it in the closing, describing the commitment to existing SkyWater customers on team retention and unparalleled IP protections as absolute [de Masi 2:45:40].
6.2 — Defense and National Security
The federal programme map as IonQ presented it, spanning all four pillars: DARPA QBI Phase B in compute; DARPA HARQ and AFRL work in networking; DARPA It’s About Time at 125 quantum clocks, GRAVMAP, DIU CRUISE and DIU QuIX in sensing; SDA HALO with 84 optical terminals on orbit supporting Tranche 1, and NRO RCA in space. Read as an exhibit, it is the breadth argument made by someone else’s procurement decisions.
The opportunity: to sell mission outcomes rather than qubits, into an ecosystem that now has a legislated deadline.
Muller framed the buying behavior with more precision than any vendor pitch could: the government does not want a qubit, it wants mission solutions to its hardest problems, and what IonQ brings is a broad platform of different capabilities that can be combined to address them — but technologies alone do not get you very far without people who understand the mission well enough to know what those solutions are worth [Muller 2:35:04].
What the closed stack unlocks is that IonQ can now assemble those combinations. Shapiro's validation map spans DARPA QBI for computing, the It's About Time sensing program, the networks, and a recent NRO contract on the space side — and his claim is that this validates not only leadership in each individual technology but that the technologies are converging to offer solutions only IonQ can deliver [Shapiro 36:47]. Singh gave two worked examples of exactly that: a Vector Atomic clock on a space platform, and using quantum computing to improve imaging from a SAR satellite [Singh 47:01].
Raymond's frame is the demand-side one: translating the brains of the people heard from into real capabilities in the hands of the warfighter is nationally critical, and is what IonQ does really well [Raymond 2:44:40]. Cardillo added that the executive order demands a scaled ecosystem, and that the teammates invited over the past twelve months have not just complemented but truly scaled the company in response [Cardillo 2:39:32].
[FACT] Sonderman placed the foundry inside this ecosystem too: quantum is sovereign infrastructure, and protecting IP matters not just commercially but for things tied to the defense of the country [Sonderman 31:11]. Singh confirmed the quantum foundry and advanced technologies segment includes work for cleared government customers on classified programs [Singh 47:01].
6.3 — Quantum Security
The opportunity: to own the threat model and the remediation, in a market with a federal deadline five years earlier than previously planned.
The mechanism is now explicit and it is unusual. IonQ published the number that defines the threat — 19,397 qubits [Shapiro 36:47] — and sells the defense against it. de Masi named the logic: you want a safe cracker to be the one who builds your safe [de Masi 4:40]. Shapiro made the same point structurally, noting that in security IonQ understands the threats better than competitors because it also has quantum computing, and in computing it offers the solution to the problem it is itself creating [Shapiro 36:47].
The release gives the threat model its precise scope, and the scope is narrower than the stage implied. The target is secp256k1, the 256-bit elliptic curve used by Bitcoin and other digital systems; the modeled exposure is to digital signatures used to prove identity and authorize transactions, not to the encryption keeping stored or transmitted data confidential; and IonQ stated that post-quantum signature standards including ML-DSA and SLH-DSA would not be affected by this type of attack. The full ladder is 19,397 physical qubits, 1,457 logical qubits, about 39 million logical Toffoli gates, roughly 25.7 days per attempt.
[ARG] That IonQ published the limits of its own result alongside the result is commercially counterintuitive and analytically valuable. It also directly qualifies de Masi's remark from the stage about large language models making progress against recommended post-quantum algorithms: the company's own release names two post-quantum signature standards as unaffected by the attack it modeled. Those are different claims about different things, but a reader encountering both should hold the release's specificity above the stage's aside.
The demand-side catalyst is regulatory and dated: the June US executive order pulled the migration timeline from 2035 to 2030 [Shapiro 36:47]. The commercial proof that enterprises are moving is Congruity360, described as the largest deal of its kind in the United States [Shapiro 36:47], on top of Singtel and SK Broadband [Shapiro 36:47]. The release values that agreement at $8.18 million and specifies the content: Clavis quantum key distribution devices and Solteris network appliances, combining post-quantum cryptography with QKD to protect information moving between locations. Congruity360 is a data-management and governance company serving finance, healthcare, insurance, legal services, manufacturing, defense and higher education. Duration, installation schedule and 2026 revenue contribution were not disclosed.
de Masi added an accelerant that would extend the addressable urgency if it holds: classical AI is playing into this trend, with large language models making progress against some recommended post-quantum algorithms, and the combination of quantum computers and classical models in the wrong hands making the problem trickier [de Masi 4:40].
[UNDISC] That claim was made in passing and without a citation, and it sits awkwardly beside IonQ's own written statement that ML-DSA and SLH-DSA are unaffected by the attack it modeled. It is recorded here because it was said, not because it is established, and this series should not repeat it without independent verification.
[INFER] An $8.18 million agreement being characterized as among the largest commercial quantum-security deals in the United States is the most useful single calibration of this market's current size available anywhere in the day's materials. The ecosystem is early in absolute terms, whatever the strength of the strategic logic — and any model built on this segment should be anchored to that figure rather than to the deal's adjectives.
The enterprise buyer's own framing came from Lokada, who argued the industry must look beyond the letters PQC, which is largely classical anyway, and that trust is subjective while operational resiliency — not compliance — is the durable frame [Lokada 2:10:25; Lokada 2:19:00].
6.4 — Quantum Networking
The opportunity: to sell the connective layer to every other quantum company's architecture, having crossed the threshold that makes it functional.
Ballance's statement that IonQ's core compute roadmap does not require photonic interconnects on a five-year view, and that interconnect customers are the ones building fleets and networks rather than one big computer, is a description of a merchant market [Ballance 1:08:42]. Bhaskar's claim that the kilohertz milestone applies across modalities and will be the leading interconnect solution for trapped ions, trapped atoms, superconductors and more, describes the product for that market [Bhaskar 1:31:00].
The reference deployment is government and already running: the AFRL entanglement-distribution network allowing experimentation across different modalities of quantum computers on the same network [Shapiro 36:47]. The commercial reference is EPB, which started in networking, expanded to computing, and is now expanding into quantum memories in Chattanooga [Chad 1:42:36; Long 2:02:32].
The extension is the free-space path — getting off fiber using IonQ's optical communication technology, which unlocks airborne and, in the future, space-based quantum connections [Shapiro 36:47].
[ARG] This series has maintained a specific distinction and it holds after this event: IonQ has now stated intent to pursue space-based quantum communication links, and has not stated intent to build space-based quantum data centers. The first is quotable as company guidance. The second remains this author's analytical framing and must be labeled as such wherever it appears.
6.5 — Sensing, PNT and Timing
[FACT] The deck adds scale figures the spoken remarks did not. DARPA’s It’s About Time program is described as 125 quantum clocks, against the $58 million agreement value Shapiro stated from the stage [Shapiro 36:47]. On the space side, SDA HALO is shown with 84 optical communications terminals already on orbit supporting Tranche 1. Those are deployed unit counts rather than roadmap targets, which is what separates this pillar from the others.
The opportunity: to move from a boutique instrument business to a chip-scale volume business, on the back of the photonics layer.
Today the products are excellent and hand-built. Bhaskar described the current state candidly: on the screen was a quantum sensing device made of discrete components, integrated about as tightly as possible, but still showing a jumble of wires connecting everything together — with clocks and inertial navigation sensors still packaged at the level of discrete elements and still hand assembled [Bhaskar 1:31:00; Bhaskar 1:27:36].
What Nexus changes is the manufacturing basis: taking the most complex parts and putting them on a chip, at a fraction of the size, weight, power and cost, with much higher manufacturability, yield and reliability — Nexus brings the know-how and expertise, and SkyWater brings the ability to manufacture it all on US soil [Bhaskar 1:31:00; Bhaskar 1:27:36].
The demand is already proven at the low-volume end: clocks a thousand times more accurate than the best classical equivalent, leading NIST's ensemble in this form factor, deployed at sea, in space, on the X-37B, on land and in the air [Shapiro 36:47]. Singh's jamming rationale describes a demand curve that steepens with geopolitical conditions rather than with technology adoption [Singh 47:01]. Shapiro noted sovereign demand for advanced PNT to operate in GPS-denied environments as increasingly important in today's geopolitical landscape [Shapiro 1:25:01].
[INFER] This is the cleanest cost-curve story in the portfolio and the one with the shortest path to volume, because the product is already qualified and the constraint is purely manufacturing. No speaker framed it as the near-term revenue opportunity it appears to be.
6.6 — Space and Earth Intelligence
The Miramar change-detection result, quantified in the deck rather than on stage. Filtered F1 score — which combines missed changes and false detections into a value between 0 and 1 — came in at 0.41 for the quantum run on the QPU, against 0.24 for the classical Copula method and 0.16 for classical NLCD. A meaningful margin on a hard dataset, and well short of a solved problem.
The opportunity: quantum-enhanced geospatial intelligence, as a proprietary product rather than a service.
The SAR result covered in Section 2.3 is the whole argument [Shapiro 36:47], and Cardillo's coherence-from-chaos framing is the market articulation [Cardillo 2:39:32]. Singh listed imaging from space using AI as one of three customer-facing solution categories alongside security and computing in his opening [Singh 0:00], and later cited quantum-improved SAR imaging as a proof point of the solutions strategy turning into announcements [Singh 47:01].
The adjacency that follows is the free-space optical work [Shapiro 36:47], which places quantum links on the same platforms that carry the sensing payloads.
6.7 — Life Sciences and Healthcare
The machine class the applications actually need, as IonQ scoped it: roughly 1,300 to 5,000 logical qubits and 400 million to more than a billion Toffoli gates, running days to weeks. Set that against the 1,457 logical qubits and 39 million Toffoli gates in the cryptography result and the scale of the gap between a demonstrated compilation and a commercial workload becomes legible.
The opportunity: to be the computational substrate for the problems pharmaceutical and genomics companies already know they cannot solve, with a stated machine class and a stated date.
de Masi set the priority: life sciences is near and dear at IonQ, spanning oncology, predictive medicine and personalized medicine, and modeling quantum-mechanical things like materials and pharmaceuticals is the modality's superpower [de Masi 4:40].
The change today is that the gap between demonstrated and useful is now numbered. Protein folding stands at 16 amino acids on shipping hardware against a GLP-1 range of roughly 30 to 50 [Chad 1:42:36]. Cytochrome P450 chemistry — two to three billion dollars a year of research, not exactly solvable classically — requires a 100K system or five networked 20Ks, and delivers roughly two years of pull-forward worth hundreds of millions in net present value [Chad 1:42:36].
IonQ now states the machine, the molecule, the timeline and the value. That is a capital-planning conversation, not a technology conversation. |
WHY THE P450 DISCLOSURE IS THE MOST COMMERCIALLY USEFUL STATEMENT OF THE DAY Enterprise pharmaceutical buyers do not purchase quantum computing on the strength of a speedup claim. They purchase on the strength of a defensible answer to a single question: what machine, by when, for which molecule, worth how much. Until today, no vendor had put all four on the record simultaneously. IonQ now has. The molecule class is cytochrome P450 chemistry. The machine is a Superion 100K, or five networked 20Ks. The economic frame is two years of pull-forward against a two-to-three-billion-dollar annual research spend. That converts a technology conversation into a capital-planning conversation, which is the conversation that actually closes. The discipline note that belongs alongside it: the single-machine path and the networked path are not equivalent in maturity. The networked path depends on an interconnect announced today at above one kilohertz that has not been demonstrated at the fidelity or scale a distributed chemistry workload would require [Bhaskar 1:31:00]. Presenting them as interchangeable would overstate the evidence. |
The named engagements underneath: AstraZeneca continuing at three iterations, Synopsys updated in August, CCRM in healthcare, and quantum pattern-finding on fMRI for early-onset brain cancer and dementia [Chad 1:42:36]. The adjacency Rabinowitz opened — neoantigen selection for personal cancer vaccines, therapy-response prediction, and quantum fine-tuning heads on classical genomic models [Rabinowitz 2:03:45; Rabinowitz 2:12:54] — is a second front in the same ecosystem, and one that runs through data assets rather than through chemistry.
Rabinowitz also supplied the scale of the prize outside drug discovery: MyOme's whole-genome and electronic-medical-record modeling suggests over two hundred billion dollars of savings to the US healthcare system by predicting susceptibility to common and rare diseases and catching them earlier [Rabinowitz 2:03:45]. Cardillo named drug discovery among the mission outcomes that brought him to the company [Cardillo 2:33:00].
6.8 — Energy and Utilities
The opportunity: grid optimization as a repeatable, regulated-utility product, evidenced by a customer already running it.
EPB is not a pilot. Long described focusing on energy-grid optimization and energy security, using essentially the full stack of IonQ services, with eight fellows and EPB writing its own algorithms and running its own programs to service the energy needs of the industry [Long 2:02:32]. de Masi characterized EPB as one of IonQ's largest, closest and most historic partners, with a computer, a network, and its own applications and algorithms [de Masi 2:07:02].
The replication logic is the regional ecosystem: Chattanooga is accumulating an IonQ R&D office, Vanderbilt's satellite office, UT Chattanooga partnerships, and city, county and state backing, with Long describing momentum getting momentum [Long 2:08:09]. Every regulated utility with fiber has the same starting position EPB had ten years ago.
Adjacent to it, Chad reported hiring an oil-and-gas specialist who in month one met two of the largest US energy companies, both of which already had quantum science teams [Chad 1:42:36]. Shapiro described the oil-and-gas engagement as tri-product — quantum chemistry solutions, PNT for finding new wells, and quantum security for critical infrastructure [Shapiro 1:25:01].
6.9 — Enterprise Software and AI Convergence
The opportunity: to sit inside the hybrid architecture that enterprise software vendors are already designing toward.
Lokada's framing is the buyer's own: quantum will not help every problem, and people sometimes look at quantum as a very fast solution when several days is fine for the right strategic solution [Lokada 2:10:25]. What ServiceNow is designing for is the orchestration problem — several billion agents coming online over the next several years, and how to leverage probabilistic intelligence to deliver deterministic workflows [Lokada 2:10:25]. His requirement is the plumbing that integrates between today and tomorrow, GPUs where GPUs make sense and QPUs where they are needed [Lokada 2:10:25].
IonQ's answer is already architectural rather than aspirational. Singh described hybrid deployments as an AI factory with a GPU next to the quantum computer, and framed the sequencing plainly: if you are asking what comes after AI, quantum comes after AI, and if you are asking whether quantum can work with AI, IonQ is making it happen today in hybrid solutions [Singh 47:01]. Ballance reported demonstrated speedups of hybrid stacks with quantum processors, GPUs and CPUs working together, better than each part individually [Ballance 18:36]. Chad added the NVIDIA and Oak Ridge paper on using LLMs to improve quantum algorithms, with GPU and QPU clusters working hand in hand, and the result showing quantum algorithms improving LLM behavior by 24 percent on accuracy with better time to solution and energy efficiency [Chad 1:42:36].
[ARG] The AI convergence story is normally told as quantum one day helping AI. That version is wrong and this stage proved it wrong. The relationship is already running in both directions — LLMs improving quantum algorithms, quantum layers improving model accuracy — and a customer reproduced the second direction independently. Bidirectional relationships are far harder to displace than one-way dependencies, and this one has a two-year head start on anyone who wants to copy it.
6.10 — Sovereign and International
[FACT] The release that followed the next morning is larger than the stage remarks implied. IonQ Italia and Lutech announced a strategic collaboration with an initial term of twenty-four months, spanning quantum computing, quantum security, quantum networking and quantum sensing. Lutech brings industrial process knowledge, system integration, and established capability in AI, cloud, cybersecurity, high-performance computing, data engineering and managed services. Target sectors named are financial services and insurance, energy and utilities, telecommunications, manufacturing, healthcare and pharmaceuticals, transport, aerospace and defence, public administration, smart cities and critical infrastructure.
[FACT] The geography is not Italian. Initial activity covers Italy, San Marino, Vatican City, Spain, Portugal, Switzerland, Germany, France and the United Kingdom, with possible extension to further European and international markets. Nine jurisdictions, announced as an Italian partnership.
Marco Pistoia, chief executive of IonQ Italia, framed it in terms that match Butti’s almost exactly, which is the more interesting fact: the ambition is not to bring quantum technologies developed in the United States to Italy but to build them in Italy, alongside businesses, research centres and public administration, creating skills, intellectual property, applications and industrial value in the country. Giuseppe Di Franco, chief executive of Lutech, described the sector as entering the phase where technological availability has to become industrial capability. The release calls the model open technological sovereignty — access to the best global technology combined with local capacity to select, integrate, secure and govern it.
[INFER] A sovereign customer stating a requirement from the stage and a commercial partner announcing the delivery vehicle for it the following morning is a tighter loop than this ecosystem has previously shown. It also changes the shape of the opportunity: post-quantum security assessment and migration across nine European jurisdictions is a services annuity rather than a hardware sale, and it is the second business line in this report whose economics do not depend on IonQ’s own modality winning.
The opportunity: to become the default supplier for nations building sovereign quantum capability, through national partners rather than direct sales.
Italy is the template and it was assembled in public. Butti described a national strategy issued last July at the same time as Europe's, going beyond computing to encompass quantum communication and security, sensing, research, skills and industrial ecosystem, with a National Quantum Forum convening government, researchers, universities and industry [Butti 2:21:37]. His stated ambition — that Italy should not simply use quantum technology developed elsewhere but help create it — is the demand statement [Butti 2:21:37]. His stated preference for an open, competitive and technology-neutral ecosystem in line with the European Union is the procurement constraint [Butti 2:21:37].
IonQ's answer is a local entity plus a local channel. de Masi announced a partnership between IonQ Italia, run by Marco Pistoia, and Lutech for the nation's go-to-market of all of IonQ's quantum platforms [de Masi 2:27:33]. Butti welcomed IonQ Italia by name as reflecting the collaborative approach, noted the Q alliance of which IonQ is a founding member connects companies, universities and research institutions and gives researchers access to advanced quantum compute systems, and set the next milestone at the Como Lake Digital Innovation Forum at Cernobbio [Butti 2:21:37].
Shapiro described the pattern generalizing: IonQ offers a compelling set of solutions to any sovereign nation wanting everything from advanced PNT for GPS-denied environments to the computing power that will change its economy, and sovereigns are interested in all forms of quantum technology [Shapiro 1:25:01]. Muller's trust argument is the closing one — ion-based systems give a sovereign customer its own solution deployed locally or remotely, rather than an anonymous quantum service logged into on an anonymous host [Muller 2:41:49].
[ARG] The Italy structure is a template, and it is the highest barrier to entry in the entire portfolio. National subsidiary, national systems-integrator channel, government strategy alignment, university consortium membership, recurring ministerial forum. A competitor wanting to displace IonQ in Italy does not need a better machine. It needs to replicate five relationships, one of which is with a sitting minister who has already appeared on IonQ’s stage. That is not a sales problem. It is close to impossible.
6.11 — Cloud and Hyperscaler
The opportunity: the highest-margin, fastest-growing line in the business, now with a new product generation to sell into it.
This ecosystem received one sentence from de Masi — on the cloud since 2020 across Amazon, Google and Microsoft, generally the biggest cloud partners in the quantum space, with the business expected to remain exciting, high margin and growing quickly [de Masi 4:40] — and one much more consequential sentence from Singh, who described quantum services including QCaaS as growing faster than the computing business and more profitable than many of the things IonQ does [Singh 47:01].
The release confirms the supply side of that demand: orders for Superion 256 are open, and production systems will also be made available through IonQ's cloud service. Chad supplied the forward indicator: customers who bought Aria and Forte bought Tempo and now want Superion cloud access [Chad 1:42:36]. Ballance supplied the utilization evidence: there is not an hour that goes by without a job running on the IonQ quantum computer fleet [Ballance 18:36].
[ARG] This is the most underweighted ecosystem in the company relative to its economics, and it is not close. It got roughly one percent of stage time. On the CFO’s own characterization it is simultaneously the fastest-growing and among the most profitable lines in the business. The new segment taxonomy will expose it for the first time [Singh 47:01]. When it becomes visible, it will change how this business is valued — recurring, high-margin services revenue is priced on entirely different multiples from hardware.
6.12 — Academic, Workforce and Regional Ecosystem
The opportunity: to shape the talent and standards pipeline in the places where quantum manufacturing is being sited.
Three regional clusters were visible at this event: Chattanooga, with the IonQ R&D office, Vanderbilt, UT Chattanooga and EPB's fellows [Long 2:02:32; Long 2:08:09]; the Midwest, with Qolab's Madison design, packaging and cryogenic testing feeding SkyWater's Minnesota fab in what Ho called a resilient quantum manufacturing ecosystem for America [Ho 1:40:01]; and Italy, where Butti tied strategy explicitly to research and skills and argued that technology alone is not enough without investment in people, skills and trust [Butti 2:21:37].
A commercial fact of the day appeared only in the release, and the wording matters. IonQ states that Superion 256 is available to order now with customer deliveries in 2027, and that it pre-sold its first Superion 256 system in the first quarter of 2026 — before the platform was publicly named. The Superion release does not identify that customer, and this report does not name one.
[ARG] A system sold two quarters before it had a name, against a 2027 delivery, is a purchase made on the roadmap rather than on the specification. A buyer willing to commit before a specification is public is buying the delivery record, not the datasheet — which is the same say-do argument de Masi made from the stage, expressed as a purchase order.
The research-standing evidence is the IEEE Quantum Week result: 857 papers submitted, 20 best papers selected, IonQ submitting 13 with 10 entered and four winning [Chad 1:42:36]. de Masi's stated aspiration is that far more exciting applications will be created by people building businesses and doing research with IonQ machines at arm's length [de Masi 4:40].
7. The Compounding Layer: Where the Value Actually Sits
Section 6 treats the ecosystems separately, which is how everyone models this companyand why everyone gets it wrong. In a vertically integrated business the value concentrates at the intersections — in the products no single-layer competitor can construct at any price, because constructing them requires owning two things at once. Six such intersections are already visible. Every one of them is a moat.
The value is not in the twelve ecosystems. It is in the six places where two of them meet — and where no single-layer competitor can follow at any price. |
Intersection | Layers required | What it produces, and who evidenced it |
Cryptanalysis × security products | Compilation, architecture, security stack | The credibility that sells remediation. IonQ publishes the threat number and sells the defense, and the board's government pre-briefing preserved both relationships [Shapiro 36:47; de Masi 4:40; Cardillo 2:43:39]. |
Foundry × merchant supply | Silicon, packaging, photonics platforms | Revenue that is uncorrelated with which qubit modality wins, from nine customers on two productized platforms, endorsed by a competing modality's Nobel laureate [Sonderman 31:11; Bhaskar 1:31:00; Martinis 1:37:17; Ho 1:40:01]. |
Compute × proprietary radar data | Device, applications, satellite assets | Quantum-enhanced change detection that neither a satellite operator nor a quantum company could build alone, validated by a former NGA director [Shapiro 36:47; Cardillo 2:39:32]. |
Interconnect × chemistry | Interconnect, error correction, applications | A second path to P450-class chemistry through five networked 20K systems rather than a single 100K, shortening the timeline to a named commercial workload [Chad 1:42:36; Bhaskar 1:31:00]. |
Photonics × sensing | Photonic integration, foundry, sensing | Chip-scale clocks and inertial sensors replacing hand-assembled discrete devices, at a fraction of size, weight, power and cost, manufactured domestically [Bhaskar 1:27:36; Bhaskar 1:31:00; Shapiro 36:47]. |
Quantum × AI, bidirectionally | Applications, classical software, hybrid stacks | LLMs improving quantum algorithms and quantum layers improving model accuracy, with an independent customer reproduction of the second direction [Chad 1:42:36; Ballance 18:36; Rabinowitz 2:12:54]. |
7.1 — The Convergence Argument, in Five Voices
Five executives made the same argument in five different vocabularies, which is itself informative about whether it is a talking point or a description of how the company actually works.
Speaker | The argument as made |
de Masi [4:40; 2:45:40] | Vertical integration accelerates IonQ's own roadmap while supplying the allied industry — the platform is simultaneously an internal advantage and an external product. |
Ballance [18:36] | Working across the full stack gives a virtuous flywheel of innovation that can only be obtained by working across the entire quantum compute stack, letting the team see around corners. |
Shapiro [36:47] | Technologies converge to provide solutions otherwise unavailable — security understands the threat because computing creates it; the platform is stronger together. |
Bhaskar [1:31:00] | Heterogeneous architecture as literal engineering: a compute-optimized device and a network-optimized device combined outperform either alone, exactly as CPUs, GPUs and network cards do. |
Singh [47:01] | Land and expand across TAMs — entry through any product opens conversations across the others, so convergence is a commercial mechanism as much as a technical one. |
[ARG] Bhaskar’s version is the only one that is measurable, and that makes it the only one that counts. The other four describe convergence as strategy. His describes it as a benchmark with a number attached and a threshold crossed. Hold the other four to that standard, and hold this report to it too.
Shapiro's own summary of the commercial consequence is the most concrete: customers are approaching IonQ not just for one part of the platform but for integrated solutions, virtually globally and in virtually every industry — financial services firms discussing quantum computing applications and then hopping to the security team to talk to the CISO, oil and gas discussing chemistry, PNT and critical-infrastructure security in the same conversation [Shapiro 1:25:01].
8. What This Changes
The table below is the practical output of this report. The left column is the model most of the market is still running. The right column is the one the evidence in Sections 1 through 7 supports. The distance between them is the opportunity.
Dimension | Before September 8 | After |
Nature of the risk | Physics risk in the deep layers | Yield, cadence and commercial conversion risk in the shallow ones |
Nature of the business | Quantum computing company with acquisitions attached | Vertically integrated manufacturer with a platform attached |
Foundry economics | Assumed captive supply | Two thirds external, nine customers, two productized platforms |
Interconnect | Orders of magnitude short of useful | Above the threshold for distributed computation, on a heterogeneous architecture |
Compilation maturity | Resource estimates | An end-to-end compilation with a 30× module improvement falling out |
Life sciences | Demonstrations | A named machine class, a named molecule class, and an NPV frame |
Security timing | Q-Day is coming | 19,397 qubits, against a 2030 federal deadline |
Modality exposure | A bet on trapped ions | Partly hedged — the foundry and interconnect lines are modality-agnostic |
Proof standard | Company assertion | Nine external parties attesting to nine different layers |
8.1 — The Financial Position Underneath It
The guidance bridge as presented: $608 million street consensus for SkyWater's full year, less revenue already delivered January through July plus close adjustments, leaving roughly $240 million of standalone contribution across the five months IonQ will own the business.
The arithmetic stated plainly — $280 to $290 million for IonQ reaffirmed, $240 million from SkyWater for August through December, less $70 million of eliminated intercompany revenue, giving $450 to $460 million for the combined company.
A maturity thesis is not an earnings thesis, and the figures below are the reason this report does not attempt the second. IonQ produced $80.1 million of revenue in the second quarter of 2026, up 287 percent year on year, against an adjusted EBITDA loss of $120.3 million — $95.6 million excluding the $24.7 million of research spending on the SkyWater commercial relationship before that transaction closed. The GAAP net loss for the quarter was $1,867.7 million, or $5.08 a share, of which $1,576.2 million was a non-cash loss on the revaluation of warrant liabilities driven by the rising share price. That figure describes the stock rather than the operations, and this report treats adjusted EBITDA as the operating measure while stating the GAAP number rather than omitting it. Net cash used in operating activities was $254.8 million across the first half. Cash, cash equivalents and investments stood at $3.0 billion at June 30, 2026, and roughly $2.0 billion pro forma for the SkyWater consideration.
[ARG] Roughly $2 billion of post-acquisition liquidityagainst a quarterly adjusted EBITDA loss of that size buys years, not quarters, and Singh's stated philosophy — the strongest balance sheet required for success, not too strong and not too weak [Singh 47:01] — is consistent with it. But the platform being more advanced than modeled does not make the losses smaller. Both things are true simultaneously, and a report arguing the first should say the second plainly.
The platform being more advanced than modeled does not make the losses smaller. Both are true, and this report says both. |
8.2 — The Forecasting Record, and What It Implies for Guidance
A guidance number means something different depending on who issued it, and IonQ’s guidance has a documented history of being conservative. That record deserves weight.
[FACT] Q1 2026 revenue of $64.7 million exceeded the midpoint of the guidance range by 30 percent, on IonQ’s own statement, and was the fourth consecutive quarter of record results. Q2 2026 revenue of $80.1 million came in 20 percent above the midpoint of the previously provided range — IonQ had guided $65 to $68 million for the quarter — and was the fifth consecutive record quarter. De Masi described it as again exceeding guidance.
[FACT] Full-year 2026 guidance has been raised twice, from an opening range of $225 to $245 million in February, to $260 to $270 million in May, to $280 to $290 million in August, before the SkyWater contribution took the combined figure to $450 to $460 million on September 8. That is three upward revisions in seven months.
[FACT] The backlog supports it rather than merely accompanying it. Remaining performance obligations stood at $470 million at March 31, up 554 percent year on year, and $485 million at June 30, up 297 percent. For every dollar of revenue recognised in the first quarter, roughly $2.50 of new contracted backlog was added.
Two consecutive beats of 30 percent and 20 percent against the midpoint, three guidance raises in seven months, and backlog growing faster than revenue. This is a company that guides below what it delivers. |
[ARG] The reasonable working assumption is therefore that the $450 to $460 million combined figure is a floor rather than a central case, and that a beat of at least 20 percent against the midpoint of quarterly guidance is the pattern to expect until it breaks. This report states that as an expectation drawn from a documented two-quarter run and a three-raise sequence, not as a forecast with a probability attached. It is also a pattern with an obvious termination condition: the first quarter that merely meets guidance ends it, and the SkyWater contribution is a new and unproven variable in the arithmetic that has no beat history inside IonQ at all.
[INFER] There is a structural reason the pattern may persist through 2026 specifically. Guidance issued in February could not incorporate SkyWater, whose close was uncertain; guidance issued in May and August was constrained by the same transaction. A company that has just spent seven months guiding around an unresolved acquisition has been guiding with a deliberately wide margin of safety.
8.3 — The Metrics That Should Now Be Tracked
If the reframing is right, the leading indicators change. The following should replace fidelity-record watching as the primary tracked series in this coverage:
- Quantum wafer starts at SkyWater, and the IonQ versus external split — the single best proxy for both the foundry business and the roadmap cadence [Sonderman 31:11].
- Foundry customer count, currently nine [Sonderman 31:11], and any additions or departures.
- Device generations in parallel, currently five [Sonderman 1:17:18]; design cycle time, which IonQ puts at two months against nine previously; tapeouts per half, currently six; and wafer lots run, currently twelve times the prior foundry's rate over six months.
- Superion deployment schedule adherence against the stated early-2027 first customer sites [Ballance 18:36], and against the two new release-only commitments: a laboratory fault-tolerance demonstration in 2027 and a manufacturable commercial system in 2028.
- Quantum services growth relative to hardware, now separately reportable under the new taxonomy [Singh 47:01].
- Protein-folding amino-acid count, currently 16 against a 30-to-50 target [Chad 1:42:36] — the cleanest single proxy for chemistry readiness.
- Cost per qubit against the stated projection of a greater-than-300× reduction across the roadmap — the number the entire architectural bet ultimately resolves to.
- Multi-product customer penetration, which Singh described as the land-and-expand mechanism [Singh 47:01] and Chad evidenced through the Aria-to-Forte-to-Tempo-to-Superion upgrade pattern [Chad 1:42:36].
8.4 — How the Day Was Received, and the Result That Was Not IonQ’s
A report arguing that the market’s model is out of date owes the reader the market’s actual response. It was mild, and the reason is more interesting than it first appears.
IonQ traded up roughly two and a half percent intraday on September 8 — an intraday move, not a close, and this report does not present it as one. Sell-side reception was positive. Craig Ellis of B. Riley, who had asked about tiling and customer engagement from the floor hours earlier, reiterated a Buy with a $100 price target, calling the guidance raise the most notable item and noting the midpoint had risen by $170 million to $455 million. He expects concerns about revenue growth to ease rapidly and raises the possibility of annualized revenue approaching a billion dollars within roughly four quarters. Street consensus at the time stood at Strong Buy on eight Buys and one Hold, with an average target of $69.44.
The same day, the United States government finalized CHIPS Act quantum awards — $100 million each to firms including Rigetti and D-Wave, with the government taking minority, non-controlling equity stakes. Those two stocks rose roughly four and seven percent. On the day IonQ launched a sixth-generation platform, raised guidance, published a full-stack cryptanalysis result, signed a competitor as a foundry customer and launched a merchant foundry business, it was outperformed by two competitors whose news was a government cheque.
Why the Comparison Is Not What It Looks Like
[FACT] The September finalizations execute letters of intent announced months earlier. On May 21, 2026 the Department of Commerce announced $2.013 billion in planned CHIPS Act quantum funding across nine recipients — IBM at $1 billion for a standalone quantum foundry, GlobalFoundries at $375 million, six companies at $100 million each including Quantinuum, D-Wave, Infleqtion, PsiQuantum, Rigetti and Atom Computing, and Diraq at up to $38 million. IonQ was not among the nine. The selection decision predates the SkyWater close by more than two months.
[FACT] The vehicle behind those awards — Broad Agency Announcement 2025-NIST-CHIPS-CRDO-01, administered by NIST’s CHIPS Research and Development Office — accepts proposals on a rolling basis through 2029. Commerce’s own May release states the office continues to solicit proposals from eligible applicants and describes the nine letters of intent as an initial portfolio rather than a closed allocation. This is a channel that remains open, not a round that closed.
[FACT] The disclosed award structures involve issuing equity to the government. D-Wave’s award involved newly issued common stock; Rigetti’s used a discount-priced equity formula. These are share-issuance transactions, not grants.
[ARG] That combination supplies a mechanical explanation for IonQ’s absence that has nothing to do with eligibility or merit. Through the entire May window the SkyWater acquisition sat under active FTC Second Request review, with the transaction’s final structure and consideration unfixed. A company cannot cleanly commit to issuing new equity to the federal government while a transaction of that size is unresolved and its own share count is a live variable in the deal. This series has argued that reading since before the close and continues to label it inference from the published award structure rather than established fact — IonQ has never stated a reason, and none should be attributed to it.
[INFER] If that reading holds, the date that matters is not September 8 but July 31, when the transaction closed and the constraint lifted. The natural moment to petition is after the close and after the combined entity has a reported segment structure and audited numbers to show — which is to say, not thirty-nine days later at an event whose purpose was to introduce the combined company. An investor day is a poor venue for announcing a request that has not yet been granted.
The comparison that flattered Rigetti and D-Wave on September 8 was decided in May, while IonQ’s share count was still an open term in a $1.8 billion acquisition. |
[UNDISC] None of this was addressed from the stage and no analyst asked. Whether IonQ has petitioned, intends to, or has been told anything is unknown, and the public record does not say. This series recorded government equity participation as a pre-event prediction and scored it a miss; the miss stands, but the explanation now available makes it a timing miss rather than a positioning one. The trusted-fab argument in Section 6 would be genuinely weakened only if IonQ applies once the constraint has lifted and is passed over.
9. What Would Change the View
A thesis stated this forcefully is only worth reading if it comes with the conditions that would break it. These are specific, dated where possible, and checkable by anyone. Any one of them would require revising the analysis above, and this series will say so in public if one triggers.
Trigger | Consequence for this thesis |
Median gate fidelity across the 256-qubit device, when published, falls materially short of unit-cell performance | The central claim of Section 2 — that the layers are integrated, not merely individually strong — would be directly undermined at the device layer, and the recalibration in Section 1 would need reversing for Layers 1 and 2. |
The 2027 laboratory fault-tolerance demonstration or the 2028 manufacturable commercial system slips | Both are release-only commitments and both sit on the critical path from 256 to 10K. A slip on either would move the fault-tolerance timeline that underwrites the security, chemistry and sovereign arguments simultaneously — the single highest-leverage schedule risk in the portfolio. |
The system sold in the first quarter of 2026 does not reach delivery and operation | The launch customer failing to reach operation would undercut the strongest evidence that Superion is a shipment schedule rather than a roadmap. |
First Superion customer deployments slip beyond the first half of 2027 | The clock-speed proof in Section 2.4 fails. The whole reframing from science project to manufacturer rests on cadence, and a schedule miss on the first vertically integrated product is the most direct available counter-evidence. |
The kilohertz interconnect does not replicate at operationally relevant fidelity or distance | Section 2.2 weakens from a stack proof to a laboratory result, and the networked-20K path to P450 chemistry in Section 6.7 would need removing. |
SkyWater's external quantum wafer share falls materially, or a foundry customer departs citing conflict of interest | The merchant-supply opportunity in Section 6.1 collapses toward the alternative reading — that external volume was capacity utilization rather than strategy — and Sonderman's compartmentalization assurance moves from unverified to disconfirmed [Sonderman 1:13:58]. |
The 19,397-qubit figure is materially disputed by credible cryptographic reviewers, or the compilation is shown to rest on the same idealizing assumptions as the generic literature | Section 6.3's commercial logic fails at its foundation, because threat credibility is what sells the remediation, and the reputational exposure Cardillo described would become concrete [Cardillo 2:43:39]. |
FY2027 guidance, when it arrives, implies material deceleration in the standalone business | The platform maturity thesis would survive but its investment expression would not. Maturity that does not convert to revenue growth is an engineering achievement, not a business one. |
No external party independently reproduces a material IonQ result over the next twelve months | Section 5's argument — that distributed external corroboration is itself evidence — weakens considerably. Invited testimony at an investor day is a weaker form of the same claim than unprompted third-party replication. |
IonQ petitions for CHIPS quantum participation after the SkyWater close and is passed over | The clean test of the timing explanation in Section 8.4. A rejection after the share-issuance constraint has lifted would mean the absence was about IonQ rather than about the calendar, and the sovereign-infrastructure argument in Section 6 would need re-weighting. |
Anduril or Sandia relationships are disclosed as lapsed or materially reduced | Section 6.2's defense opportunity would need re-weighting, and the silence noted in the concessions would resolve in the adverse direction. |
10. The Verdict
WHAT THIS SERIES CONCLUDES IonQ is no longer a quantum computing company. It is the industrial base of an industry, and it spent September 8 proving it to a room that mostly wrote down the guidance number and left. The stack is closed. Thirteen layers moved in one afternoon and none moved backward. Four results were produced that independent layers cannot produce. Nine witnesses whose standing comes from somewhere other than IonQ — a 2025 Nobel laureate in the competing qubit modality, a sitting foreign minister, a former director of the National Geospatial-Intelligence Agency, a utility running its own algorithms, a customer who reproduced an IonQ result with his own team — each spoke to a different layer of it. Seven questioners had the floor and not one asked whether the physics works. The risk has moved. It is no longer physics risk, which is unbounded and unmanageable and priced accordingly. It is yield, cadence and commercial conversion — ordinary industrial risk, carried by a company that owns the only quantum foundry at scale in the world and runs it two-thirds full with other people’s wafers. The opportunity is not the one being discussed. The discussion is about qubit counts and a guidance raise. The opportunity is in the six intersections in Section 7, where IonQ can build products no single-layer competitor can construct at any price: cryptanalysis credibility that sells the remediation, a foundry that earns whichever modality wins, radar analysis requiring ownership of both the satellites and the computer, chip-scale sensing, a networked path to pharmaceutical chemistry, and an AI relationship already running in both directions. This report says plainly what it believes and shows its work. Sections 9 and 10 say just as plainly what would break it. Both are meant seriously. |
IonQ is no longer a quantum computing company. It is the industrial base of an industry. |
One closing observation about what kind of company this now is. The most revealing moment of the day was not an announcement. It was the President of Quantum Computing being asked what worries him about tiling out to end-of-decade scale, and answering that the concern is not any single technology risk — it is operational excellence, understanding the risks properly and doing enough in parallel to make them boring, and that this is not a particularly exciting problem, which he offered as high praise [Ballance 1:23:45].
Quantum computing companies do not talk like that. Manufacturers do. That sentence is the thesis of this report, delivered by the person best positioned to know it, and delivered in answer to a question rather than from a slide — which is usually where the true things are.
PART TWO
The Examination
Twenty-Five Voices, Question by Question
Part One argued a thesis. Part Two puts every speaker under examination, one at a time, and asks what each of them actually established. Sixteen examinations. One hundred and twelve questions. Under every answer, what it means for the stack and for the business model.
The Method of the Examination
Only seven people asked questions at the New York Stock Exchange on September 8, and they were all sell-side analysts working a single ninety-minute window. Every other speaker delivered prepared remarks. Part Two therefore uses two question types, and marks each one so a reader always knows which is which.
[FACT] ASKED FROM THE FLOOR — a question actually put to the speaker in the room, reproduced as asked, with the speaker’s live answer.
[INFER] PUT TO THE RECORD — an analytical question this author puts to the speaker’s prepared remarks. The speaker did not hear it. The answer is assembled strictly from what they said on stage, and nothing is attributed to them that they did not say.
The distinction is not cosmetic. A speaker who volunteers a number under no pressure and a speaker who concedes one under questioning are two different evidentiary events, and this examination never blurs them.
Quotation is reserved for the sentences where the exact wording is the evidence; everything else is summarized. Where the event record renders a proper noun incorrectly — an artifact of transcription rather than of speech — the name is given as IonQ’s press releases render it, since the releases govern. This affects Inder Singh, Mihir Bhaskar, Qolab, the SC250 platform and Lightsynq. No substantive wording has been altered.
Under every answer sits a shaded block explaining what that answer does for the full stack, for the business model, or for both. These are the author’s judgments, not the speaker’s claims, and they are written to be arguable. Where an answer matters mainly because of what it did not contain, the block says so.
11. Niccolò de Masi
Chairman and Chief Executive Officer, IonQ · Opening remarks 4:40–18:27; panel management throughout; close 2:45:40
Standing: Chief executive since the beginning of 2025. Architect of the acquisition program and the person who set the day’s frame.
Owns in the stack: Capital allocation, acquisition strategy, the platform narrative, government and sovereign relationships, and the six-announcement structure of the day.
Layers evidenced: Cross-cutting — L1 sovereign silicon, L6 interconnect, L7 networking and security, L10 distribution, trust and governance
Q1. Why does vertical integration matter enough to justify what you spent on it? [PUT TO THE RECORD]
He put it in the first sentence he spoke, and framed it as a category rather than a transaction: IonQ is not only the largest quantum company in history but, in his words, “the first one that is vertically integrated.” He then gave the reason twice — first that the company is “investing in driving merchant supply to accelerate the US industry and our allies,” and second that IonQ is “speeding up our own roadmap because of this vertical integration.” Two returns from one asset.
Why it matters to the full stack: This is the business-model claim on which the whole day rests, and it is unusual because the two returns are not additive but structurally different. One is a merchant revenue line that grows when competitors succeed. The other is an internal cycle-time gain that grows when IonQ pushes its own roadmap harder. A company normally has to choose between selling capacity and hoarding it. De Masi is claiming IonQ does not, and Sonderman spends his entire segment supplying the operational evidence for it.
Q2. How do you want investors to model this company — as a quantum computer maker or as something else? [PUT TO THE RECORD]
He rejected the single-product frame explicitly and named four: “We are focused as a company on these four areas in our platform: computing, security, sensing, and of course also our foundry, which is a new part of the family.” He reinforced it with a revenue comparison — before SkyWater, IonQ did roughly twice the revenue of the entire rest of the public sector in the second quarter.
Why it matters to the full stack: The four-pillar frame is the first formal statement of how IonQ wants to be valued, and it is the reason a comparison to any single-modality quantum company is now the wrong comparison. It also creates an obligation the company has not yet met: four pillars invite four sets of segment economics, and only revenue was given. Singh later provided the reporting structure but not the margins.
Q3. You put out six press releases in one morning. Was that a strategy or a coincidence of timing? [PUT TO THE RECORD]
He treated it as a designed event, not an accident of the calendar: “For those of you who haven’t seen it, this is a big news day for us. We put out six press releases.” He then walked the audience through what he was proudest of — quintupled year-on-year revenue, the Walking Cat fault-tolerant architecture published in April, three DARPA programs (HARQ, It’s About Time, QBI), and progress on interconnect and networking.
Why it matters to the full stack: Six simultaneous releases across product launch, guidance, cryptanalysis research, a commercial security agreement, a foundry business launch and a competitor partnership is a deliberate demonstration that the pillars fire together. It is also a communications risk: a market can only absorb so much in one morning, and the cryptanalysis result — the most technically significant item — competed for attention with five others. Section 1 argues the market did not fully price it.
Q4. How close is Q-Day, and why should a chief information security officer care this year rather than later? [PUT TO THE RECORD]
He put it inside the planning horizon of the person who has to act: Q-Day is “coming sooner than people think, and it’s coming inside the planning horizon of a CISO, whether or not you are in a government position, federal position, or an enterprise one.” He then made the argument that the company that breaks the safe is the one you want building it — “you want a safe cracker to be the one who builds your safe” — and noted IonQ had delivered news on the safe-breaking front that morning.
Why it matters to the full stack: This is the commercial logic connecting the cryptanalysis paper to the security business, and it is the cleanest example on the day of one pillar generating demand for another. It is also the most self-serving argument any executive made, which is precisely why the restraint in the underlying paper matters: IonQ published what its attack cannot touch alongside what it can, which is the opposite of what a pure fear-selling strategy would do.
Q5. What is the accelerating factor you did not expect? [PUT TO THE RECORD]
He named classical AI, and did it in unusually direct terms, arguing that the combination of quantum computers and classical large language models in the wrong hands is what makes the security problem harder. He referenced lab incidents at AI companies and made a specific claim about a frontier model making progress against recommended post-quantum algorithms.
Why it matters to the full stack: This is not an offhand aside, and the deck changes how it should be read. Slide 9 makes the claim formally, under the heading that classical AI is weakening modern post-quantum algorithms, and states that AI has found previously unknown flaws in leading NIST PQC candidates. That wording is what reconciles it with IonQ’s own release stating that ML-DSA and SLH-DSA are unaffected: candidates that were evaluated and not selected are a different set from the standards that were. The company is therefore not contradicting itself, and this report withdraws the suggestion that it was. What remains is that no citation, paper or incident is given for the claim anywhere in the deck or the releases, and a formal slide assertion without a source is still an assertion without a source. Treat it as unevidenced rather than as inconsistent.
Q6. How should the market read the Superion naming and the qubit ladder behind it? [PUT TO THE RECORD]
He launched the line as a product family rather than a device, and committed to the ladder in public: “we’ll go from two, fifty-six to ten K to twenty K, two hundred K.” He argued that each generation will be more powerful “per unit dollar, per unit energy consumption, per unit space,” in some cases by orders of magnitude, and framed the goal as mass market — an access point every company and government in the world can invest in.
Why it matters to the full stack: Naming a product family rather than a machine is a commitment device: it forces the company to ship successors under the same brand or explain why it did not. The per-unit-dollar and per-unit-energy framing is the one that matters commercially, because it converts a physics roadmap into a total-cost-of-ownership argument that a procurement officer can act on. The releases put a figure on it — a projected reduction of more than 300 times in cost per qubit across the roadmap in the press release, which the deck states as 330 times.
Q7. What is your defence of the say-do ratio, and what did you commit to next? [PUT TO THE RECORD]
He tied credibility to a thirty-year record of delivery and to the last five or six years as a public company, closing the argument with the assertion that IonQ does what it says it will do. He then set two forward markers on the record: a return to the New York Stock Exchange for World Quantum Day on April 14, 2027, and the next analyst day one year out.
Why it matters to the full stack: Booking a date is the cheapest and most durable accountability mechanism available to a management team, and it is the reason the series maintains the IonQ Delivery Ledger. Two named future events give the market two scheduled opportunities to test the say-do claim against outcomes. The relevant open item is the one the ladder created: the 256-qubit device demonstration now has less runway than it did before the day began.
Q8. Who was in the room, and why did you spend stage time saying so? [PUT TO THE RECORD]
He inventoried the audience deliberately — distinguished university professors, the company’s founder, founders of public companies with larger market capitalizations than IonQ’s, a hereditary prince from Europe, a former ambassador, a former governor, and a serving Italian government official. He also noted that the event was oversubscribed and that people had to be turned away.
Why it matters to the full stack: Convening power is an asset that does not appear on a balance sheet and cannot be bought quickly. For a company whose commercial future depends on sovereign programs, national laboratories and regulated enterprises, the ability to assemble that particular room is itself a competitive position. It is soft evidence, and this report treats it as soft — but the national security panel two hours later converted it into something considerably harder.
12. Dr. Chris Ballance
President, Quantum Computing, IonQ · Presentation 18:36–31:08; live answers at 1:08:42, 1:10:04, 1:20:02, 1:23:45
Standing: Co-founder of Oxford Ionics, acquired by IonQ in September 2025 for approximately $1.075 billion. Now owns the entire computing roadmap of the acquiring company. Took four of the seven analyst questions.
Owns in the stack: Qubit control, chip architecture, error correction, compilation, the Superion scaling ladder, and the phrase — silicon to solution — that gives this report its spine.
Layers evidenced: L1 sovereign silicon, L2 qubit control and the device, L3 packaging and tiling, L4 architecture and error correction, L5 classical control and compilation
Q1. What did ten years and five generations of shipped machines actually teach IonQ? [PUT TO THE RECORD]
He answered with three lessons rather than a list of achievements. First, the platform must be radically simplified to reach orders of magnitude more qubits and more deployments. Second, the company must learn faster — “we need to go from learning over product cycles of years to product cycles of months.” Third, it must demonstrate scale generation over generation. He anchored the credibility of those lessons in operational reality: “There’s now not an hour goes by without a job running on the IonQ quantum computer fleet.”
Why it matters to the full stack: This is the most important framing choice any technologist made on the day, because it converts a decade of shipping into an argument about learning rate rather than about installed base. The competitive claim underneath it is that operating a fleet in production teaches things a laboratory cannot — supply chain, uptime, field failure modes — and that those lessons are what dictate the next architecture. It is also the setup for the SkyWater argument, since cycle time is the variable a captive fab moves.
Q2. What is the single biggest architectural change in Superion, and why that change? [PUT TO THE RECORD]
Lasers out, electronics in. He was explicit that laser control is “a great technology for the zero-to-one phase, getting the first hardware built in the lab” but “challenging to mass manufacture.” Superion replaces lasers for qubit control with electronics built into the chip, removing what he called the whole middle layer of laser control and connecting classical control electronics directly to the quantum chip — which lets IonQ build the chips “in standard semiconductor ways.”
Why it matters to the full stack: This is the decision that makes every other part of the stack cohere. Electronic qubit control is what allows a trapped-ion machine to be built on a semiconductor line rather than assembled around an optical table, which is what makes a captive foundry valuable, which is what makes cost per qubit a controllable variable, which is what makes the mass-market claim in de Masi’s remarks something other than aspiration. Remove this one choice and the acquisition logic of the last fifteen months substantially weakens.
Q3. Electronic control gets you one good cell. How do you get from one cell to hundreds? [PUT TO THE RECORD]
He named the second core technology and reached for a consumer analogy to size the gap: in the same way there is a great deal of technology between a single LED and a television display, there is a great deal between a single unit cell and an array of them. IonQ calls its answer quantum multiplexed I/O, and he said the company has built “an incredible, unique patent portfolio of IP around how we do this.” The unit cells in question have achieved better than four nines on two-qubit gate fidelity.
Why it matters to the full stack: The LED-to-display analogy is doing real work: it says the remaining problem is engineering the scale-out of a proven cell, not proving the cell. The four-nines figure is the strongest single performance number disclosed on the day — but it is a unit-cell figure, not a 256-qubit device figure, and Bolton’s question later established that IonQ would not yet state when the device-level number arrives. That gap is the sharpest genuine disclosure limit in the event — though it should be read as a component-validated architecture awaiting an assembled-device number, not as an unvalidated one.
Q4. Does the Superion 256 exist, and where is it in its life? [PUT TO THE RECORD]
He showed the chip and gave three staged facts. IonQ is “already producing these devices at scale at SkyWater by the wafer.” Those devices are being built into packaged quantum processors sitting at the core of Superion 256 machines. And then the announcement: “we now have ion qubits in these first prototype systems,” with first customer-site deployments expected in early 2027.
Why it matters to the full stack: Wafer-scale production, packaged processors and loaded ions are three separate milestones, and stating them together compresses a normal multi-year sequence into a single slide. The early-2027 deployment date is the most falsifiable commitment IonQ made all day and now anchors the Delivery Ledger. The release adds a commercial fact the room did not hear: IonQ sold its first 256-qubit system in the first quarter of 2026, before the platform had a name — disclosed in the Q1 release as a sale anchored by a secure quantum network and a broad IP-generation partnership spanning computing, networking, sensing and security, and reported at the time as the University of Cambridge, with commissioning expected by the end of the second quarter of 2027.
Q5. You claimed to have compressed innovation cycles from years to months. What is the evidence? [PUT TO THE RECORD]
He offered a countable one rather than a narrative: “Over the last six months, we’ve built more ion-trap chips, ion-trap quantum processors than we have in the history of the company to date.” He attributed it to deliberate operational change — how the company is structured, how technology is developed, how teams are run — and to running multiple generations in parallel.
Why it matters to the full stack: A six-month period out-producing a company’s entire prior history is the kind of claim that is either true or catastrophic to make in public, and it is checkable in a way that most technology assertions are not. It also converts the SkyWater acquisition from a supply-security story into a learning-rate story. Sonderman put the same claim in fab terms minutes later: twelve times more wafer lots over six months than the prior foundry delivered.
Q6. How does the roadmap get from 256 to hundreds of thousands, concretely? [PUT TO THE RECORD]
He gave the ladder and the mechanism at each rung. The Walking Cat error-correction architecture and the compute fabric shown in the 256 system carry through to the ten-thousand-qubit machine, with the major addition being integrated cryogenic CMOS to implement quantum multiplexing — already being de-risked at SkyWater, with first wafers on display in the room. The next scaling node comes from tiling two dies together in two-and-a-half-D integration, and hundreds of thousands of qubits from tiling more dies and raising qubit density per die.
Why it matters to the full stack: Each rung has a named physical mechanism and a named place where the de-risking is happening. That is what separates this from the roadmap slides the sector has produced for a decade. The strategic consequence is that the scaling path runs through packaging and integration rather than through new physics — which is exactly why owning a fab with an advanced packaging operation in Florida is a roadmap asset rather than a cost centre.
Q7. At what number of physical qubits does photonic quantum interconnect become important in your roadmap? [ASKED FROM THE FLOOR]
McPeak of Rosenblatt asked it, and Ballance declined the premise that photonics is a scaling dependency: on the core compute roadmap IonQ can “scale out well beyond millions of qubits” on a five-year view, and “right now, we don’t believe we need photonic interconnects in the near term for that.” He then reframed photonics as demand-driven rather than architecture-driven — the customers who want it are not the ones who want a single large machine, but those building fleets and networks — and called it “a tremendously valuable thing to sell alongside our core quantum compute products.”
Why it matters to the full stack: This is the most commercially revealing answer of the analyst session and it cuts against the company’s own promotional interest. Bhaskar had a kilohertz interconnect milestone to announce twenty minutes later, and Ballance still separated it from the critical path. Two consequences follow: the compute roadmap carries no photonics dependency risk, and the networking business is a discretionary revenue line sold to a distinct buyer. Investors who model photonics as a gate on the compute roadmap are modelling it wrong.
Q8. Why did IonQ choose SkyWater over an established industrial foundry? [ASKED FROM THE FLOOR]
Asked to explain the cycle-time compression, Ballance answered with culture and then produced a dated worked example. Late in the prior year IonQ was running two foundries. A major industrial foundry was pushing hardware delivery out by months over a chemical supply problem. SkyWater had the identical problem. IonQ spoke to the SkyWater team on a Tuesday; they said their supplier’s one-month quote was nonsense — in “a slightly fruitier word,” as he put it. By Thursday they had found a supplier in Korea holding one in stock and horse-traded for it. By Sunday it was installed in a machine and yielding results. The other partner was still quoting a two-month slip on a six-month timeline.
Why it matters to the full stack: Tuesday to Sunday against six months plus two is the most concrete due-diligence artifact disclosed on the day, and it explains a $2 billion acquisition better than any strategic rationale could. It also reframes what IonQ bought: not capacity, which can be contracted, but a decision-making tempo that cannot. Singh corroborated it independently from the buyer’s seat, describing an alternative foundry that needed permission from the mothership and simply moved too slowly.
Q9. What worries you about tiling — the technology that has to work for the whole ladder? [ASKED FROM THE FLOOR]
Ellis of B. Riley asked what competencies tiling requires. Ballance began by refusing the framing that it is a physics problem: “it’s a point of pride, it’s not a quantum problem, it’s an integration problem.” The requirements are die bonding within tight tolerances and designing the electrostatic transfer that physically pushes qubits across the junction. IonQ has run thorough studies against industry-standard tolerances and is, in his words, “comfortably within that” — not pushing the limit of any single thing. He then named what does concern him: “the stuff that worries me on that isn’t any one technology risk, it’s operational excellence.” How to push up yields. How to make the packaging boring.
Why it matters to the full stack: This is the single answer that best supports the thesis that the risk profile changed character on September 8. Yields and boring packaging are questions with known methods, known timelines and known failure modes. A physics uncertainty has none of those. Note also what he conceded: comfortably within industry tolerances is a claim about a study, not about a demonstrated tiled device — and he was candid that integration and iteration is the work happening now, ahead of production.
13. Tom Sonderman
Chief Executive, SkyWater — and following the July 31, 2026 close, leader of IonQ’s foundry operations · Presentation 31:11–36:37; live answers at 1:11:46, 1:13:58, 1:17:18, 1:21:05
Standing: Semiconductor operating executive with a prior career at AMD and GlobalFoundries. Presented the manufacturing plan for the whole company thirty-nine days after his company was acquired.
Owns in the stack: Wafer manufacturing, process node, the technology-as-a-service model, IP compartmentalization, advanced packaging, and the merchant supply position.
Layers evidenced: L1 sovereign silicon and L3 packaging, tiling and integration — plus the merchant supply position argued in Section 6.12
Q1. What actually changed when SkyWater and IonQ combined? [PUT TO THE RECORD]
He located the change at a specific boundary in the industry’s development: “we’ve moved out of the physics demonstration realm to now manufacturing at scale in a highly developed operation.” That operation delivers both the speed needed for cycles of learning against Ballance’s roadmap and the scale needed when volume arrives.
Why it matters to the full stack: Naming the physics-to-manufacturing transition from the manufacturing side is more persuasive than hearing it from the quantum side, because a fab executive has no professional incentive to declare the physics settled. It is also the sentence that most directly supports Section 1’s claim that the nature of the risk moved.
Q2. How much quantum production actually runs through the fab, and how much of it is IonQ’s? [PUT TO THE RECORD]
The disclosure that changes the model: “Today we have thousands, literally thousands of quantum wafers running in our fab. A third of those are tied to IonQ.” He drew the conclusion himself — that it means IonQ is designing for manufacturability, and that when it is time to run at scale, the operation will be ready.
Why it matters to the full stack: Two thirds of quantum production at the only quantum foundry of scale is running for someone other than IonQ. That single ratio converts the fab from a captive cost centre into a merchant supply business with a structurally different revenue profile: it grows when the industry grows, including when IonQ’s competitors succeed. This is the observation that justifies treating merchant supply as a distinct ecosystem rather than a footnote to the computing business.
Q3. Why was SkyWater able to do this when other specialty foundries were not? [PUT TO THE RECORD]
He gave the origin story as a deliberate strategic inversion. SkyWater was “literally created to solve” the lab-to-fab problem, through a technology foundry concept and a technology-as-a-service business model built to bridge “what’s known as the valley of death.” Where specialty foundries chose standardization because they could not keep up with TSMC, “we took the opposite approach. We embraced customization.” And the first customer of the technology-as-a-service model was a quantum company.
Why it matters to the full stack: This is the answer to the obvious objection that a larger foundry could simply decide to serve quantum customers. Standardization and customization are opposed operating philosophies, expressed in tooling, staffing, scheduling and pricing, and a fab optimized for one does not pivot to the other on a strategic decision. It is also the reason the competitive lead time Mobley asked about is longer than a capital-expenditure schedule would suggest.
Q4. Why does a foundry executive talk about national security? [PUT TO THE RECORD]
He made it a design principle rather than a market segment: what SkyWater put in place was part of a strategy “to not only create the environment, but create the security that protects those innovations.” Then the framing sentence — “Quantum is sovereign infrastructure” — and the dual obligation that follows, to protect customer IP and to protect national security for work tied to the defence of the country.
Why it matters to the full stack: Sovereign infrastructure is a category with different buyers, different procurement cycles and different substitutability than commercial semiconductors. If the framing holds, the trusted-fab position is not a margin advantage but a qualification barrier, and Singh’s claim that competitors are targeting 2029 becomes a statement about how long that barrier lasts rather than a boast.
Q5. What is the goal of the speed advantage, stated plainly? [PUT TO THE RECORD]
He said it twice in one breath and left no ambiguity: thousands of wafers allow parallel work and rapid learning, and what that will continue to do is “move the projections that Chris has provided to the left. That is our goal: to continue to move the roadmap to the left, and we believe that we are well on our way to doing that.”
Why it matters to the full stack: A public commitment to pull a published roadmap forward is an unusual thing for a manufacturing executive to make in front of analysts, because it is measurable against dates the company has already printed. It is also the mechanism behind the whole vertical-integration argument: the value of owning the fab is denominated in schedule, not in cost of goods.
Q6. How many quantum customers does the fab have, and what does that make SkyWater? [PUT TO THE RECORD]
Nine, as of the Qolab announcement that morning. And he drew the larger conclusion rather than leaving it implied: “So we aren’t just driving IonQ’s roadmap. We’re literally creating an industry, a quantum foundry industry, that will be foundational through the rest of this decade and beyond.” He closed on the line that all roads for quantum lead through SkyWater.
Why it matters to the full stack: Nine customers is the difference between a supplier and a piece of industry infrastructure. Infrastructure positions produce revenue that is uncorrelated with the owner’s own technology bets — the single most valuable property any part of IonQ acquired in the last fifteen months, precisely because every other pillar is a directional wager on trapped ions.
Q7. How do the other eight quantum customers get comfortable that you will not handicap their roadmaps — and how much lead time do you have on IBM and GlobalFoundries? [ASKED FROM THE FLOOR]
Mobley of StoneX asked the hardest commercial question of the day and Sonderman answered both halves without hedging. On conflict: protecting IP is foundational to the foundry model, so “we compartmentalize the information, we don’t cross-contaminate,” and every customer should be “a hundred percent confident that their roadmap won’t be impeded by the needs of IonQ.” He added a rising-tide argument — moving fast with the largest quantum customer populates those capabilities across the others. On competition he was blunt and specific: Anduril bought HRL; IBM sold its microelectronics business to GlobalFoundries in 2015, so if they want back in, so be it; GlobalFoundries, where he previously worked, has capabilities but “they aren’t an innovation engine. They are a scaled specialty foundry. They love standardization. We embrace customization.”
Why it matters to the full stack: This answer had to work or the merchant supply thesis fails, and it is the moment the day was most exposed. He answered with process rather than assurance — compartmentalization is an auditable practice, not a promise — and then declined to be evasive about named competitors, including his own former employer. The rising-tide claim is the load-bearing one and it is also the one no outside party has yet corroborated; the eight other customers have not spoken publicly. Martinis and Ho, thirty minutes later, went as far as any third party could toward doing so.
Q8. What lithography node does this run on, and does the node matter? [ASKED FROM THE FLOOR]
Choksi of Northland asked, having called the cycle-time reduction astounding. Sonderman gave the number and then dismissed the premise of the question in the same breath: “we do this on 90-nanometer technology. You hear others talk about, oh, you need 28 and below, and some people even talking about two nanometer. That’s ridiculous.” The fabrication IonQ needs is done in a mature 200-millimetre environment that adapts well to Ballance’s innovations, and he expects cost per qubit from SkyWater and IonQ to stay industry leading.
Why it matters to the full stack: Ninety nanometres on 200-millimetre wafers is the cheapest, most mature and most available process environment in the disclosure, and it decouples IonQ’s roadmap from leading-edge capacity constraints, geopolitical exposure around advanced nodes, and the capital intensity that dominates the rest of the semiconductor industry. If the ladder to hundreds of thousands of qubits runs on a mature node, the cost curve is a manufacturing-learning curve rather than a node-migration curve. This is one of the most underrated facts of the entire day.
Q9. Do you have enough capacity, and what is Fab 25 for? [ASKED FROM THE FLOOR]
On capacity he was direct — Minnesota and Florida together have more than enough for the compute roadmap and the broader platform, because “these don’t require the same scale as like a smartphone or even AI data centers. But they require lots of iterations.” He reached for his own history at AMD, where three nodes ran in parallel against Intel; IonQ is running five device generations in parallel. On Bolton’s Fab 25 question he described the Texas fab as part of the foundry ecosystem, giving purchasing power and supply-chain optimization, with quantum work sitting in Minnesota and Florida. Singh added the buyer’s framing: Fab 25 is optionality that would require retooling, and so far has not been needed.
Why it matters to the full stack: Capacity being a non-issue removes a capital-expenditure question the market would otherwise have to model, and the five-generations-in-parallel figure quantifies the learning-rate advantage in a way an analyst can compare against industry norms. The Fab 25 answer is a small but real piece of honesty: the company did not claim a strategic quantum purpose for an asset that currently does not have one.
14. Inder Singh
Chief Financial Officer and Chief Operating Officer, IonQ · Opening 0:00–3:08; presentation 47:01–1:08:00; live answers at 1:12:19, 1:18:35, 1:22:07, 1:28:56
Standing: Former Wall Street analyst; joined the board before moving into the company in September 2025. Previously at Arm and Cisco. Presented the guidance bridge and the new segment structure.
Owns in the stack: Guidance, segment reporting, capital allocation, acquisition diligence from the buyer’s side, pricing strategy and the balance sheet.
Layers evidenced: L10 distribution, trust and governance — with a segment structure spanning L1 through L9
Q1. How do you get to the raised full-year number, line by line? [PUT TO THE RECORD]
He built the bridge in public rather than stating a total. Street consensus for SkyWater as of July 30 — the day before closing — was $608 million. SkyWater had already delivered $368 million of that through the first two quarters plus an estimate for July, which he flagged as unaudited. That leaves roughly $240 million across the five months IonQ will own the business this year, of which about $70 million is IonQ’s own spend and must be eliminated at the corporate level. Add the reaffirmed $280–290 million for IonQ, subtract the $70 million, and the range is $450–460 million.
Why it matters to the full stack: Guidance built by visible arithmetic rather than asserted as a total is a governance signal, and it lets an analyst re-run the calculation with different assumptions rather than accepting a number. The elimination line matters practically too: it quantifies how much of SkyWater’s revenue was already IonQ’s money, which is the single most common modelling error available on a vertical acquisition.
Q2. What did an experienced analyst look for before joining, and does IonQ have it? [PUT TO THE RECORD]
He described his own former checklist — technology, roadmap, investment dollars, staying power, positioning — and then said the item that actually decided things was management: “Do I believe, do I trust they can execute? Do they have a track record of execution?” He returned to it later as the test of the day: do they say something and then do it, and do they beat what they said they would do.
Why it matters to the full stack: The say-do framing is not decoration; it is the standard the company invited the market to hold it to, and it is why this series maintains a Delivery Ledger rather than a wish list. A chief financial officer who publicly nominates execution track record as the primary valuation input is accepting that a missed commitment costs more than a missed quarter.
Q3. How should the market think about the total addressable market, and where does SkyWater sit in it? [PUT TO THE RECORD]
He was careful to disown authorship: “It’s not my TAM number, but a hundred and forty-eight billion dollar TAM according to third parties” — and noted that virtually every other quantum company competes inside that single number. The published deck names the source and the basis, which the spoken remarks did not: McKinsey’s Quantum Technology Monitor 2026, an estimate of the internal quantum market — investments plus estimated vendor revenue plus estimated internal funding at large technology companies — for 2040, under optimistic growth scenarios. That is a very different object from a current addressable market, and the deck says so in its own footnote. Then the pointed omission: “I haven’t got a TAM up there for SkyWater. It’s new. It’s a huge TAM.”
Why it matters to the full stack: Declining to size the foundry opportunity while sizing everything else is the more credible choice, and this examination treats it as such rather than filling the gap with an estimate of its own. The structural point survives without a number: the quantum foundry market did not previously exist as a category, and the company that created it does not yet know how big it is. Sizing it would have been the easiest and least defensible slide of the day, and he did not make it.
Q4. What is the new segment structure and why change it now? [PUT TO THE RECORD]
Four buckets. On the IonQ side, quantum hardware — machines sold on premises, increasingly in hybrid deployments alongside a GPU in an AI factory — and quantum services, which holds applications and cloud access. He volunteered that services is growing faster than the computing business and is more profitable than many things IonQ does. On the SkyWater side, SEMI semiconductor foundry, principally the Texas operation, and quantum foundry and advanced technologies. He explicitly simplified SkyWater’s prior categories so that there is “a limited number of things that you all can follow that make sense to you, and you can hold us accountable for.”
Why it matters to the full stack: This is the reporting architecture that will govern every future earnings model, and it was designed for accountability rather than flattery. The most commercially significant sentence is the aside about quantum services: a faster-growing, higher-margin software and cloud line inside a company the market prices as a hardware manufacturer is a mix-shift argument nobody on the sell side pursued.
Q5. Why did IonQ pick SkyWater rather than an established foundry — from the buyer’s side of the table? [ASKED FROM THE FLOOR]
He corroborated Ballance’s account from the diligence seat. Other foundries were trying to be quantum foundries, and IonQ talked to one he declined to name. They could not move fast enough. They did not quite understand what was needed. They had no learning curve on quantum. They were, in his words, blown away by the volume numbers IonQ was describing for future years — and then the detail that decided it: “they needed permission from the mothership, and they just move too slowly.”
Why it matters to the full stack: Two executives from different companies, different disciplines and different sides of the transaction independently describing the same failure mode is the strongest form of corroboration available at a staged event. It also identifies precisely what a large incumbent cannot replicate quickly: not tooling, but decision rights close enough to the work that a supply problem can be solved between a Tuesday and a Sunday.
Q6. How do you think about the other quantum companies in your own fab? [ASKED FROM THE FLOOR]
He rejected the competitive framing outright: “I don’t think of the others as competitors that are in that foundry. They’re part of the industry and the ecosystem that’s going to exist in the future. They’re already our customers in some ways.” He then made the reciprocal claim — that accelerating IonQ’s own roadmap and learning curve helps the other players as well — and confirmed that IonQ brings the capital SkyWater needs to grow the quantum foundry.
Why it matters to the full stack: This is the same rising-tide argument Sonderman made, now stated by the officer who controls the capital that would enforce or violate it. Consistency between the operator and the funder on a question where their incentives could diverge is meaningful. It remains a claim about intent, and the eight other customers still have not spoken — which is why the Qolab agreement, signed rather than stated, carries more weight than either executive’s assurance.
Q7. What is the pricing strategy for the 256, and how should we model fiscal 2027? [ASKED FROM THE FLOOR]
Vijay of Mizuho asked both. Singh answered the first obliquely and the second not at all. On pricing he offered the shape without the number, describing “a candy jar of opportunity” — IonQ is lowering its own cost from an already low base, and over the next few years must decide how much of that margin to keep and how much to share with customers. He argued that by the time IonQ reaches a customer it is usually not a competitive situation, because the customer has already chosen a modality, so the conversation becomes value and use cases rather than price. Then the claim underneath it: “There isn’t another company with a 256 chip out there, period.” He closed by asking whether he had successfully dodged the question. Fiscal 2027 was never addressed, by him or by anyone else.
Why it matters to the full stack: The pricing answer is more informative than it appears — lowering cost from the lowest base while deciding how much to pass through is a description of a business expecting pricing power, and the forward-deployed engineering model he described is how a company converts a price conversation into a value conversation. The fiscal 2027 non-answer is the largest single gap the event left, and it is a regression: an investor day is the natural venue for a forward framework, and the market got the current year only.
Q8. What is the balance sheet for? [PUT TO THE RECORD]
He set an explicit calibration rather than a maximum: the intent is to keep “the strongest balance sheet required for success — not too strong, certainly not too weak.” He noted IonQ has spent more on quantum computing alone than entire governments have, roughly $1 billion on Oxford Ionics and nearly $2 billion on SkyWater, and said the goal is multi-year cash availability so that funding innovation is not a recurring question. On profitability he was candid about the limits of a one-month-old integration: EBITDA colour will come when the auditors have been through the numbers.
Why it matters to the full stack: Not too strong is an unusual thing for a pre-profit company to say, and it implies discipline about dilution rather than opportunism about it. The EBITDA deferral is legitimate this quarter and will not be next: the company has now defined four segments and raised guidance, and the next earnings call is the natural place for the margin structure underneath them. Until then the $120.3 million adjusted EBITDA loss in the second quarter stands as the most recent evidence, against roughly $2 billion of cash after the SkyWater close.
15. Jordan Shapiro
President, Quantum Platform, IonQ · Presentation 36:47–46:47; live answer at 1:25:01
Standing: Runs the platform businesses outside core computing: security, networking, sensing and space. Delivered the first public disclosure that IonQ runs its own satellite data through its own quantum computers.
Owns in the stack: Quantum security, quantum networking, positioning navigation and timing, remote sensing, and the convergence argument that ties the pillars together.
Layers evidenced: L7 networking and security, L8 sensing and PNT, L9 applications and solutions
Q1. Why is Q-Day closer than it was, and what is the number? [PUT TO THE RECORD]
He put two independent trends together. The qubit count required to threaten critical infrastructure is falling precipitously — “Today we announced 19,397 qubits is our requirement at IonQ” — and he credited quantum hardware and quantum software working together for bringing that threshold down. Simultaneously, governments are pulling migration deadlines in: the June US executive order moved the American timeline from 2035 to 2030.
Why it matters to the full stack: The two trends move toward each other, which is what makes the commercial argument work: the threat arrives sooner and the compliance deadline arrives sooner, and the gap between a CISO’s budget cycle and both is now uncomfortably small. It is worth noting the structural oddity in IonQ’s position: the same company is lowering the qubit threshold through research and selling the mitigation, which is precisely why the restraint in the underlying paper matters more than the headline.
Q2. What is the exposure today, before any large quantum computer exists? [PUT TO THE RECORD]
He named harvest now, decrypt later as a present-tense problem rather than a future one — adversaries collecting encrypted data today in the hope of decrypting it later with quantum computers. His conclusion was unambiguous: leaders need to transition now, and if they are not transitioning now, they are already too late.
Why it matters to the full stack: This is the argument that makes quantum security a current-year budget line rather than a 2030 problem, and it is the mechanism by which IonQ’s security business can generate revenue years before its computing roadmap reaches the threatening scale. It is also the industry’s standard argument rather than a proprietary one, and it should be assessed as such.
Q3. What is actually in the security product, as opposed to the narrative? [PUT TO THE RECORD]
He gave four layers. Quantum security posture management gives an enterprise visibility into where its infrastructure is vulnerable. A CNSA 2.0 and NIST-compliant solution updates that infrastructure, either by injecting post-quantum cryptography throughout or by deploying quantum key distribution — a physical solution spanning a network, using the same underlying quantum mechanics, especially where it matters most. Then orchestration and key management, and evolution of the architecture over time against what he called a changing standard.
Why it matters to the full stack: Four layers with an assessment product at the front is a full commercial stack rather than a device sale, and the assessment layer is the commercially clever part: it creates the enterprise’s own inventory of vulnerability, which is the document that funds everything that follows. The changing-standard framing also converts the product from a one-time migration into a maintained subscription — a materially different revenue profile.
Q4. Is anyone actually buying this? [PUT TO THE RECORD]
He announced the Congruity360 agreement for enterprise rollout of quantum security in the United States and called it, to IonQ’s knowledge, the largest deal of its kind in the country — spanning enterprises, healthcare systems, financial institutions, government organizations and telcos. He cited Singtel and SK Broadband as existing deployments, and claimed the position as not just the most robust platform but the most deployed.
Why it matters to the full stack: Deployment references convert a threat narrative into a revenue line, and the release adds the figure the room did not hear: $8.18 million, covering Clavis quantum key distribution devices and Solteris network appliances. It is a modest number against a $450–460 million year, and its significance is as a template rather than as revenue — an enterprise rollout channel that can be repeated.
Q5. What is the next step in quantum networking, and why does it matter beyond networking? [PUT TO THE RECORD]
He traced the sequence: IonQ was first to interconnect two quantum computers in an enterprise setting, with the Air Force Research Laboratory, and then provided AFRL a quantum network for entanglement distribution allowing experiments across different modalities and devices. The next step is to take a network that is fiber-based today and use IonQ’s optical communication technology to get off the fiber and operate in free space — which, in his words, “unlocks the capability of quantum connections that are airborne and in the future space-based.”
Why it matters to the full stack: Free-space optical is the single technical step that connects the networking business to the space assets, and it is the closest anyone came on stage to articulating what the Capella and Skyloom acquisitions are ultimately for. This series has argued a space-based quantum data center thesis on assembled capability rather than announced intent; this sentence is the strongest on-record support for the assembled-capability half, and it remains short of a stated intention.
Q6. You showed a quantum computer processing your own satellite data. What exactly was demonstrated? [PUT TO THE RECORD]
The first public disclosure of its kind. IonQ captures synthetic aperture radar data from across the globe daily. He showed a hybrid quantum-classical workflow with that data — an IonQ quantum computer placed in the loop, processing imagery from IonQ’s own SAR platform. The worked example was Miramar Air Base, with a structure changing between two images and a ground-truth reference. Of the three models tested, the one run on the IonQ quantum computer was, in his account, closest to ground truth and free of the noise and fuzziness of the classical models. He closed with the ownership point: IonQ is the only quantum computing company with that level of access to SAR data.
Why it matters to the full stack: This is the most vivid demonstration on the day of what owning the whole stack produces, because neither half of it is purchasable. A satellite operator cannot run this without a quantum computer; a quantum company cannot run it without daily proprietary imagery. What it does not yet establish is durability — one worked example on one base is a demonstration, not a benchmark, and IonQ did not name the subsidiary operating the platform or quantify the improvement.
Q7. What does the sensing business actually consist of, and how good is it? [PUT TO THE RECORD]
He listed the devices — quantum optical clocks, time-transfer devices, atomic gravimeters measuring specific gravity at any point on Earth, gyroscopes measuring inertial movement — all built on the same underlying physics, aimed at positioning, navigation and timing. The performance claim was specific: IonQ’s clocks are a thousand times more accurate than the best-in-class classical equivalent and lead NIST’s entire ensemble of clocks in that form factor. And they are deployed, not laboratory-bound: at sea, in space, on the X-37B space plane, on land and in the air.
Why it matters to the full stack: Sensing is the only pillar generating revenue from physics that is already finished, which makes it the near-term cash contributor while computing scales. The form-factor qualifier on the NIST claim is doing real work and should be read carefully — best in class within a form factor is a narrower statement than best in class. Deployment on the X-37B is the harder evidence, because that is a customer decision rather than a laboratory result.
Q8. What validates any of this beyond IonQ’s own assessment? [PUT TO THE RECORD]
He pointed at government customers across every pillar: DARPA QBI for quantum computing, a $58 million agreement for the It’s About Time program in sensing, and a recent contract with the National Reconnaissance Office on the space side. His conclusion was the convergence claim — that IonQ leads in each individual technology, but more importantly that the technologies are converging into solutions only IonQ can deliver.
Why it matters to the full stack: Government contracts are third-party validation with money attached and a technical evaluation behind them, which makes them a stronger evidence class than customer testimonials. The convergence claim is the one to hold the company to, because it is the entire justification for the conglomerate structure: if the pillars do not produce joint solutions, IonQ is four businesses under one roof rather than one platform, and the valuation arguments differ sharply between those two readings.
Q9. Are customers actually buying across the platform, or one pillar at a time? [ASKED FROM THE FLOOR]
Ellis of B. Riley asked how customers are engaging and whether it varies by geography or follows the executive orders. Shapiro led with the thesis — “the story of IonQ is that we are greater than the sum of our parts” — and said customers are asking not for one part of the platform but for integrated solutions, virtually globally and across virtually every industry. He gave two worked examples. Sovereign nations want everything from advanced PNT for GPS-denied environments through to economy-changing compute. In financial services, IonQ discusses quantum computing applications and then, in his phrase, hops over to the security team to talk to the CISO. In oil and gas it is quantum chemistry, PNT for finding new wells, and security for critical infrastructure.
Why it matters to the full stack: Cross-selling within a single account is the observable proof of a platform, and Shapiro named the specific motion — walking from one buying centre to another inside the same customer. What is missing is quantification: no attach rate, no multi-product customer count, no revenue split by pillar. The claim is credible and unmeasured, and it is exactly the metric this series will look for on the next earnings call.
16. Mihir Bhaskar
SVP and General Manager, Quantum Technologies, SkyWater · Answer at 1:27:36; presentation 1:31:00–1:37:17; framing at 1:39:29 and 1:42:03
Standing: Co-founder of Lightsynq, acquired by IonQ in May 2025. Now leads the merchant quantum foundry business alongside Sonderman — a second acquired founder given a business to run rather than a laboratory to keep.
Owns in the stack: Photonic interconnect, quantum memory, the SP90 and SC250 foundry platforms, integrated photonics, and the scale-out architecture.
Layers evidenced: L6 interconnect and photonic integration, L7 networking — and L1 through the merchant platforms SP90 and SC250
Q1. What is the actual obstacle to connecting quantum computers together? [PUT TO THE RECORD]
He named it plainly and quantified the gap: “the elephant in the room when it comes to quantum interconnects is that quantum interconnects are very slow.” Historically the field has benchmarked interconnect in hertz. The requirement to connect quantum computers usefully is kilohertz. Three orders of magnitude, stated by the person announcing that IonQ had closed it.
Why it matters to the full stack: Stating the size of the problem before announcing the result is a credibility choice, and it is what distinguishes a milestone from a press release. It also establishes why data-center-scale quantum computing has remained rhetorical across the industry: the interconnect was too slow to sustain distributed computation, so modular architectures could be drawn but not built.
Q2. What did IonQ achieve, and is it incremental? [PUT TO THE RECORD]
A photonic-based quantum interconnect above a kilohertz. He was emphatic that it is not a matter of degree: “this isn’t just an incremental advance, this is really a step change.” The consequence he drew is that quantum systems can now be connected fast enough to sustain distributed computation — so the scale-out problem becomes addressable, and the object being built is not a laboratory technology but something that enables quantum data centers.
[UNDISC] One caveat the room did not hear. The deck footnotes the kilohertz figure as an IonQ finding from a forthcoming study, with details to be made available upon publication, benchmarked against the previous ion-ion record in O’Reilly et al. (2024). The result is therefore announced but not yet independently checkable, and this report treats it as a company claim pending publication. That said, the relevant base rate is good: IonQ published the Walking Cat architecture on arXiv in April 2026, published its 99.99% two-qubit fidelity result on arXiv in October 2025, publishes reproducible benchmark repositories, and took four best-paper awards at IEEE Quantum Week. A company with that record announcing ahead of publication is a timing question rather than a credibility one.
Why it matters to the full stack: Crossing an order-of-magnitude threshold that gates an entire architectural category is a different kind of result from improving a number. It converts data-center-scale quantum computing from a slide into an engineering programme. It is also worth noting what Ballance said about it: valuable, sold alongside the core products, and not on the critical path of the compute roadmap for five years. Both men are right, and the tension between them is informative rather than contradictory.
Q3. How did you get there — optimization or architecture? [PUT TO THE RECORD]
Architecture, and he said so explicitly: “we didn’t just reach this result by turning some knobs and making incremental advances.” IonQ took a fundamentally new approach through the acquisition and integration of Lightsynq quantum memory, connecting its trapped-ion computing system to that memory in a heterogeneous architecture — a device optimized for computing and a device optimized for networking, working together to produce a system better than the sum of its parts.
Why it matters to the full stack: This is the clearest single instance of an acquisition producing a technical result that neither party could have produced alone, and it arrived roughly fifteen months after the Lightsynq close. For a market assessing whether IonQ’s acquisition programme is accumulation or integration, a dated capability that required both companies is the strongest available evidence. It is also the founder of the acquired company announcing it, which is its own signal about retention.
Q4. Why does heterogeneity matter as a concept, not just as a result? [PUT TO THE RECORD]
He reached for the machine in front of the audience. Everyone in the room takes for granted that CPUs, GPUs and networking cards are designed for different things and work together. But in quantum systems, he said, “we have not yet, until this moment, successfully combined two different types of systems and gotten improved performance.” He framed it as marking maturity in building heterogeneous technologies, noted it is the subject of the DARPA HARQ program in which IonQ is a performer, and claimed applicability beyond trapped ions — to atoms, superconductors and more.
Why it matters to the full stack: The claim that the interconnect approach works across modalities is commercially larger than the milestone itself, because it makes the networking business sellable to companies whose computing technology competes with IonQ’s. Combined with the foundry, it is the second business line inside IonQ whose addressable market includes its rivals. That is the structural shape Section 6.12 calls merchant supply, and it is the reason it deserves treatment as an ecosystem in its own right.
Q5. What are the foundry platforms you launched, and why only two? [PUT TO THE RECORD]
He launched what he described as the industry’s first dedicated quantum foundry platforms — scalable, reproducible platforms creating an open ecosystem for companies building atom, ion, superconducting, photonic or spin-based processors. And he explained the number with a physics argument rather than a product-management one: “There are only two physics ways to route quantum signals. It’s light and it’s superconducting electronics, just like there’s copper and fiber in the data center.”
Why it matters to the full stack: Two platforms derived from two physical possibilities is a durable product architecture rather than a roadmap that will fragment as customers arrive. It also defines the ceiling of the merchant opportunity in a useful way: if every quantum modality routes signals by light or by superconducting electronics, then SP90 and SC250 between them address the whole industry rather than a segment of it. The releases add the surrounding portfolio — advanced packaging, diamond and lithium niobate.
Q6. How does Nexus Photonics fit into the roadmap? [ASKED FROM THE FLOOR]
Vijay of Mizuho asked and de Masi handed it to Bhaskar as the next presenter. He described it as “another one of these one-plus-one-equals-thirty stories with SkyWater”: SkyWater has the ability to manufacture; Nexus brings a proven pedigree and intellectual property in the design and integration of integrated photonic solutions. Together they can take photonic solutions required across many modalities — ions, atoms, photonics, networks, sensors — and move the complexity of hand-assembled systems onto a chip. He was specific about where it lands first: quantum sensing devices, where clocks and inertial navigation sensors are currently built from discrete components with wires still coming out of them.
Why it matters to the full stack: Sensing being the first beneficiary matters commercially, because sensing is the pillar already generating revenue from finished physics. Moving those devices from hand assembly to chip-scale integration attacks size, weight, power, cost, yield and reliability simultaneously — which is the difference between a specialist instrument and a deployable product. This is the clearest example of an acquisition improving a business other than the one it appears to belong to.
Q7. Why should anyone believe the foundry platforms will attract the industry? [PUT TO THE RECORD]
He declined to argue it and pointed instead at where practitioners are choosing to build: “you should ask yourself when you think about the quantum foundry, where are the leaders and luminaries in the field choosing to build their quantum technologies?” He named Professor John Bowers, who pioneered putting laser systems on a chip and was in the audience, and Dr. John Martinis. Both, he said, are at SkyWater.
Why it matters to the full stack: Revealed preference is a stronger argument than capability description, and it is the correct one to make when the audience cannot evaluate process technology. The Bowers reference is doing quiet work, and it is more than an appeal to reputation: the heterogeneous bonding approach to laser-on-silicon integration developed in his UCSB laboratory was commercialized by Intel, Juniper and — directly relevant here — Nexus Photonics, which he co-founded and IonQ acquired. That relationship should temper how the reference is read: Bhaskar was citing a luminary who is also, through the acquisition, a colleague. Optical interconnects derived from that lineage are what let today’s AI data centers scale out, and the implicit parallel is that quantum will scale the same way. That is an argument for the interconnect roadmap made without asserting it.
Q8. What is the summary claim about where scaling runs? [PUT TO THE RECORD]
He closed by collapsing both halves of his presentation into one sentence — that whether through scale-out solutions or the scale-up semiconductor ecosystem, “all of the roads to scaling quantum are running through the team in this room.” It is Sonderman’s all-roads-lead-through-SkyWater claim, extended from wafers to the whole scaling problem.
Why it matters to the full stack: Two executives from two acquired companies independently converging on the same claim, in the same session, is either coordinated messaging or genuine strategic alignment, and the distinction matters. The evidence for alignment rather than script is that they arrived at it from different directions — Sonderman from wafer counts and customer numbers, Bhaskar from interconnect physics and platform architecture.
17. Chad Sakac — Go-to-Market
Chad Sakac — SVP, Quantum Field Engineering, IonQ · Presentation 1:42:36–1:59:52
Standing: An engineer who spends his time with customers, by his own description, following the scientists on stage. Delivered the only segment of the day built around what customers ask for rather than what IonQ has built.
Owns in the stack: Applications, customer engagement economics, the life sciences programme, benchmarking and third-party validation, and the forward-deployed engineering model.
Layers evidenced: L9 applications and solutions, L10 distribution and go-to-market
Q1. What phase is quantum computing actually in? [PUT TO THE RECORD]
He answered with aviation. Leonardo da Vinci had the dream and nothing came of it — invention. December 17, 1903 produced twelve seconds of flight on an aircraft built by two men from the period equivalent of duct tape and bubble gum, and he compared that to the early days of quantum computing, where he said some competitors are still trying to get out of that phase. Humans then iterated: productionize, manufacture, make it safe and reliable and deployable. Boeing emerged. And today airlines are not about airplanes at all but about how societies function. He called it building the machine that makes the machine — the organization, the people, the process — and placed IonQ in the iterate and scale phase.
Why it matters to the full stack: The machine that makes the machine is the most useful sentence in the entire go-to-market segment, because it names the actual asset: an organization, not a device. It also reframes the acquisition programme as organizational construction rather than technology accumulation. And it plants a competitive claim without naming anyone — that some of the field is still at twelve seconds.
Q2. What do customers actually ask about? [PUT TO THE RECORD]
Not what the industry benchmarks. He said customers very rarely ask about two-qubit gate fidelity or coherence time; they ask how long it will take to reach a solution valuable for their workload, and what the economics are. Time to solution, economics to solution. He characterized IonQ’s advantage as a sniper against a shotgun — sometimes comparable, sometimes ten, a hundred, a thousand or ten thousand times better.
[FACT] The deck defines the metric rather than leaving it loose, which matters because time-to-solution is easy to game. IonQ’s stated definition: the clock starts at job submission and stops when a verified, qualifying answer is returned or the last shot is completed, with compilation, quantum execution, error mitigation and all classical co-processing counted toward the total. On that definition the deck claims up to a ten-thousand-fold improvement on enterprise-relevant workloads, a hundredfold on MaxCut optimization against other quantum modalities, and twentyfold on a hybrid GPU/QPU computational chemistry workflow.
Why it matters to the full stack: This is the commercial translation layer the sector generally lacks, and it explains a strategic choice that runs through the whole day: IonQ optimizes for delivered outcomes rather than for leaderboard metrics. It also creates an uncomfortable adjacency with the disclosure gap on Superion fidelity — a company that says customers do not ask about fidelity has a softer commercial reason to publish it, even though the investment community very much does ask.
Q3. Why should anyone believe the speedup claims? [PUT TO THE RECORD]
Because they can be checked. He said IonQ published the repository so anyone could duplicate the results, and published third-party validation — the company showed up with receipts. He named one of the workloads where IonQ is roughly a thousand to ten thousand times faster: the quantum Fourier transform, which appears inside Shor’s algorithm. Quantum phase estimation, he noted, shows up in chemistry and life sciences.
Why it matters to the full stack: Publishing a reproducible repository is a materially different evidentiary posture from publishing a claim, and it is the same discipline visible in the cryptanalysis paper — arithmetic exposed for outside checking rather than asserted. The link he draws is also structurally important: the workload where IonQ is fastest sits inside the algorithm Ballance compiled, which is why the two announcements reinforce each other rather than merely coexisting.
Q4. What is the state of the AstraZeneca work and the protein folding programme? [PUT TO THE RECORD]
The AstraZeneca work continues, now through three iterations. On protein folding he gave the scale honestly by anchoring it against real molecules: oxytocin has nine amino acids and GLP-1 drugs run roughly thirty to fifty. IonQ has worked on twelve and then fourteen amino acids using Forte Enterprise and Tempo hardware, and updated work in August took Tempo from fourteen to sixteen.
Why it matters to the full stack: Twelve to fourteen to sixteen is a small ladder and he did not dress it up, which is what makes the disclosure useful. Placing it against oxytocin at nine and GLP-1s at thirty to fifty lets a reader see both that the work has passed a real therapeutic peptide and that it is not yet at the commercially dominant drug class. This is the honest framing that account-level pharmaceutical work depends on, and it is more useful to a business development conversation than a larger unqualified number would be.
Q5. Is there an application that needs a machine IonQ has not yet built? [PUT TO THE RECORD]
Yes, and he named it precisely. Cytochrome P450 is not solvable exactly by classical methods today and represents roughly two to three billion dollars a year of research. A 2025 Google and Caltech paper places the problem at a scale beyond what a Superion 20K could do. IonQ, he said, could do it with a 100K — or with five 20K systems in parallel — pulling the workload forward by about two years, worth hundreds of millions in net present value. He was careful to add that networking is not a prerequisite, but that pulling work forward has immense economic value.
Why it matters to the full stack: This is the only place on the day where an application was mapped to a specific machine class, a specific competing paper and a specific economic value, and it is the most useful single disclosure for anyone doing pharmaceutical account work. It also carries a caveat the company did not state: the two routes are not equivalent in maturity. A single 100K system and five networked 20K systems are different engineering programmes with different dependencies, and the second reintroduces exactly the interconnect that Ballance placed off the critical path.
Q6. What external recognition does the research programme have? [PUT TO THE RECORD]
He gave the arithmetic of IEEE Quantum Week. Eight hundred and fifty-seven papers were submitted and twenty were selected as best papers. IonQ submitted thirteen, of which ten entered, and four of those ten won best in show — which he characterized as a 3.7-sigma event. The deck states the same result slightly differently, as ten accepted papers and four best-paper wins out of twenty-seven, and this report notes the discrepancy in the denominator without resolving it; the four wins are consistent across both, and the named tracks are quantum applications, end-to-end hybrid case studies, and quantum-GenAI co-design. He cited one example: quantum algorithms improving large language model behaviour, with twenty-four percent better accuracy, lower time to solution and better energy efficiency. He also noted a paper with NVIDIA and Oak Ridge on using large language models to improve quantum algorithms.
Why it matters to the full stack: Four of twenty best papers from one company is peer-reviewed third-party validation of research quality, which is a different and harder evidence class than customer references or internal benchmarks. The NVIDIA and Oak Ridge collaboration matters separately: it places IonQ inside the classical high-performance computing ecosystem rather than adjacent to it, which is the practical precondition for the hybrid deployments Singh described alongside GPUs.
Q7. How big is the commercial organization, and is it growing? [PUT TO THE RECORD]
Roughly seventy people worldwide across forward-deployed engineering, customer solutions and applications R&D, almost entirely PhDs — a team he said more than doubled in the first half of the year. His operating philosophy was that activity beats almost everything, and he gave a worked example: an oil and gas specialist hired who in his first month met two of the largest US energy companies, both of which already had quantum science teams.
Why it matters to the full stack: Doubling a PhD-heavy customer-facing organization in six months is a leading indicator that shows up in revenue later, and it is the operational substance behind Singh’s claim that IonQ converts price conversations into value conversations. The oil and gas detail carries a market-structure implication worth noting: if the largest energy companies already have quantum science teams, the sales motion is not evangelism but displacement of whatever those teams are currently using.
Q8. Do customers come back? [PUT TO THE RECORD]
He gave the repeat-purchase pattern directly: customers who bought Aria and Forte bought Tempo, and want Superion cloud access. He also described EPB’s expansion — started in networking, expanded into computing, with a system being commissioned and an expansion into quantum memories in Chattanooga. And he noted that IonQ supports all seven major open-source stacks because customers want choice of language, and offers cloud, on-premises and sovereign deployment because they want choice of location.
Why it matters to the full stack: A generational upgrade path with named customers moving Aria to Forte to Tempo to Superion is the closest thing the sector has to recurring hardware revenue, and it is the strongest available evidence that the installed base is a compounding asset rather than a series of one-off sales. The language and location optionality is the unglamorous part that determines whether the platform is adoptable inside a regulated enterprise.
18. Qolab — Dr. John Martinis and Alan Ho
Martinis: co-founder and Chief Technology Officer, Qolab; 2025 Nobel Laureate in Physics (pre-recorded, 1:37:17–1:39:27). Ho: Chief Executive, Qolab (1:40:01–1:42:03)
Standing: A competitor. Qolab builds superconducting qubits — a rival modality to IonQ’s trapped ions. Neither man is employed by IonQ, holds an IonQ board seat, nor has any disclosed IonQ relationship other than the manufacturing agreement announced that morning.
Owns in the stack: Independent verification of the foundry thesis — the only pillar of the day validated by someone with a commercial reason to say the opposite.
Layers evidenced: L1 sovereign silicon — verified from outside, by a customer building on a competing modality
Q1. What phase is quantum computing in, from a rival modality’s perspective? [PUT TO THE RECORD]
Martinis described the transition rather than the destination: most of the science is understood, and the present task is figuring out how to go from that to a manufacturable system that works better when scaled up. He was explicit about the precondition — that transition happens if scientists like himself start collaborating with people in the manufacturing space, especially semiconductor fabrication.
Why it matters to the full stack: A physicist who co-developed qubit technology in the 1980s and won a Nobel Prize for the underlying work stating that the remaining problem is manufacturing rather than science is the most authoritative external corroboration available for the central claim of the day. He arrives at it from superconducting circuits, not trapped ions, which means the conclusion is about the field rather than about IonQ’s architecture.
Q2. Why SkyWater specifically? [PUT TO THE RECORD]
Because of what they are used to doing. Martinis said Qolab is happy working with SkyWater because they are accustomed to specialized fabrication and to developing new processes to overcome materials and processing demands. Ho gave the commercial version: SkyWater provides predictable unit economics for manufacturing, “which is required for us to underwrite, investors to underwrite manufacturing at scale.”
Why it matters to the full stack: Ho’s answer is the one that matters to an investor, because it identifies what a merchant foundry actually sells: not wafers, but the financeability of a manufacturing plan. A quantum company that cannot forecast unit economics cannot raise capital against production. If SkyWater is the only place in the industry that supplies that predictability, its pricing power is considerably more durable than a capacity-based analysis would suggest.
Q3. What does a competitor think of a competitor buying the fab it depends on? [PUT TO THE RECORD]
Martinis endorsed it without qualification. He said Qolab is very excited that IonQ acquired SkyWater and is helping with extra focus on quantum computing, that it is a very good partnership and acquisition, and that it helps SkyWater build new capabilities in fabrication to properly scale up quantum computers and build them properly.
Why it matters to the full stack: This is the answer that resolves Mobley’s conflict-of-interest question, and it does so from the only position that could — a customer with a competing technology and no obligation to IonQ. Sonderman’s compartmentalization commitment is a promise; Martinis’s endorsement plus a signed multiyear agreement is behaviour. Nothing else on the day carries this weight, precisely because the speaker had every commercial reason to say the opposite.
Q4. What is Qolab actually committing to? [PUT TO THE RECORD]
Ho said Qolab is proud to be the first customer for the SC250 process. The releases add the structure the room did not hear: a multiyear agreement moving Quantum System-in-Package devices from custom development onto standardized SC250 wafer services, with SkyWater investing in dedicated equipment in Minnesota to support it.
Why it matters to the full stack: Multiyear, standardized and supported by dedicated capital expenditure is the profile of an anchor customer rather than an experiment, and it is what converts SC250 from a launched platform into a validated one on day one. The migration from custom development to standardized wafer services is also the exact commercial transition that makes a foundry business scalable — custom work does not compound, process platforms do.
Q5. What is the historical analogy being made? [PUT TO THE RECORD]
Ho reached back to processors in the 1960s — a jumble of wires and discrete components — and to the innovation credited to Nobel laureate Jack Kilby that resolved it. The implication he left is that quantum systems are at the pre-integration stage of that same arc.
Why it matters to the full stack: It is the same analogy Bhaskar used for photonic integration, arrived at independently by a competitor. When two companies with different technologies and no common interest describe the industry’s position using the same historical frame, the frame is more likely to be describing something real than to be corporate messaging.
Q6. Why does the location of the fab matter? [PUT TO THE RECORD]
Ho made an argument nobody at IonQ made. People talk about Silicon Valley and not about the Silicon Heartland. Qolab does its design, packaging and cryogenic testing in Madison, Wisconsin — also in the Midwest — and SkyWater’s quantum manufacturing is in Minnesota. His conclusion: “having a quantum manufacturing ecosystem in the Midwest is a resilient quantum manufacturing ecosystem for America.”
Why it matters to the full stack: Geographic concentration of design, fabrication, packaging and test inside a single region is a supply-chain resilience argument with direct national-security relevance, and it came from a competitor rather than from the company that owns the fab. It strengthens Sonderman’s sovereign-infrastructure framing with an independent voice, which is exactly the kind of corroboration a policy audience weighs more heavily than a vendor claim.
Q7. What does Martinis say he is actually trying to build with SkyWater? [PUT TO THE RECORD]
Not a machine. He said he looks forward to working with the SkyWater team not simply to build a better and bigger quantum computer, but to “lay the industrial foundation for a new era of computing.”
Why it matters to the full stack: Industrial foundation is the same concept Sonderman reached for with creating an industry and Bhaskar with all roads to scaling. Three people — two of them from acquired companies, one of them a competitor with a Nobel Prize — independently describing the fab in infrastructure terms rather than supplier terms is the strongest cumulative case on the day for treating merchant supply as a distinct ecosystem with its own economics.
19. John Lokada — ServiceNow
Futures Director, ServiceNow · Panel remarks 2:01:46, 2:10:25, 2:19:00
Standing: Customer and prospective partner. No disclosed IonQ relationship beyond an exploration engagement. Speaks for an enterprise software platform with its own security acquisition programme.
Owns in the stack: Enterprise workflow, the security-plus-optimization convergence, and agent orchestration as a quantum workload.
Layers evidenced: L7 networking and security, L9 applications — stated from the demand side
Q1. Which part of the IonQ platform is relevant to ServiceNow? [PUT TO THE RECORD]
Two lanes, and he refused to keep them separate. Security first — he called the opportunity around quantum key distribution fascinating, and noted ServiceNow has been leading with security, having acquired Armis and Veza. Then optimization. And then the synthesis: “I think there no longer can be two different lanes of security and optimization. I think secure optimization is the really interesting path.”
Why it matters to the full stack: A customer independently arriving at the convergence argument that Shapiro made from the stage is worth more than the argument itself. It also identifies a buying pattern IonQ can act on: an enterprise that has just spent heavily on security acquisitions has a budget line and an executive sponsor already in place, which is the shortest path into an account.
Q2. What is the actual workload? [PUT TO THE RECORD]
Agent orchestration at a scale that does not yet exist. He said several billion agents will come online over the next several years and asked how anyone orchestrates that — how to leverage probabilistic intelligence to deliver genuinely deterministic workflows.
Why it matters to the full stack: This is a demand signal from outside the quantum industry’s usual application list of chemistry, logistics and finance, and it is anchored to a growth curve the enterprise software market already believes in. If agent orchestration becomes a quantum optimization workload, the addressable market is enterprise software rather than research computing — a different order of magnitude and a different buyer. It is early and unproven, and this examination treats it as a signal rather than a pipeline.
Q3. How fast does a quantum answer need to be? [PUT TO THE RECORD]
He pushed back on the industry’s own framing. Quantum will not help every problem, and people sometimes look at quantum as a very fast solution when in fact “several days is fine for the right strategic solution that you could deliver.”
Why it matters to the full stack: A customer relaxing the latency requirement widens the near-term addressable set considerably, because it admits workloads where a quantum system is slower in wall-clock terms but better in answer quality. It aligns precisely with Chad’s time-to-solution and economics-to-solution framing, and it is the sort of thing that only carries weight coming from the buyer rather than the seller.
Q4. Is post-quantum cryptography sufficient? [PUT TO THE RECORD]
No. He said that as organizations become future-ready they need to look beyond the letters PQC — “and PQC is largely classical anyway” — and to understand how to lead with security while building the plumbing that integrates between today and tomorrow, with GPUs where they make sense and QPUs where they are needed.
Why it matters to the full stack: This is a customer making the argument for quantum key distribution rather than IonQ making it, which matters because QKD is the more contested half of IonQ’s security portfolio and the half that requires physical infrastructure. An enterprise buyer volunteering that software-only post-quantum cryptography is insufficient is the demand-side case for the hardware business, and Congruity360 is the commercial expression of it.
Q5. What is the framing you want the industry to adopt? [PUT TO THE RECORD]
He offered a redefinition of hybrid work — not home and office, but working across classical and quantum systems — and argued that too many people keep quantum and AI mutually exclusive when the convergence is what actually needs delivering. He also said no company can become future-proof, but that quantum plus AI can make one future ready.
Why it matters to the full stack: Future-ready rather than future-proof is a procurement-grade distinction, because it lowers the bar an enterprise must clear to justify spending now. It is also the most honest thing any customer said: it concedes that the technology is not finished while arguing that waiting is not the correct response to that.
Q6. What is the missing term in this discussion? [PUT TO THE RECORD]
Operational resiliency. He said it is a term not used much in these conversations and that longevity is about resiliency rather than compliance — and framed the whole thing as a partner ecosystem where coming together is mutually beneficial.
Why it matters to the full stack: Compliance spending is deadline-driven and ends when the deadline passes; resiliency spending is continuous. Which of the two frames the market adopts determines whether IonQ’s security business has a 2030 cliff or an annuity, and this is the single most consequential unresolved question in the security pillar. A customer arguing for the annuity framing is the more useful evidence, though it remains one customer’s view.
Q7. Where does AI stop? [PUT TO THE RECORD]
He put the question rather than answering it, and credited Rabinowitz for the underlying point — that you need to know where AI stops and where complexity becomes far greater, and that it requires thinking outside the box. He wrapped it in an Aerosmith line about needing to know how to lose before you can win.
Why it matters to the full stack: The commercially relevant version of this question is whether there is a class of enterprise problem where classical AI plateaus, because that is the entry point for quantum in an enterprise software workflow. Rabinowitz supplied a rigorous answer to it from the same panel, in the specific case of protein folding and local minima. Together the two make a coherent argument that neither makes alone.
20. Robert Long — EPB
President of Strategic Initiatives, EPB of Chattanooga · Panel remarks 2:02:32, 2:08:09, 2:19:26
Standing: Customer. A municipal power utility operating the first nationally available quantum network and a commercially available quantum computer. A decade of quantum operating history predating any IonQ relationship.
Owns in the stack: The only customer on the day describing a fully operational quantum deployment with its own staff writing its own algorithms.
Layers evidenced: L7 networking, L9 applications, L10 distribution — the only live full-stack deployment on the record
Q1. What does EPB actually have running? [PUT TO THE RECORD]
A full fiber network in the Chattanooga area, a quantum network in place for about ten years, and an IonQ Forte Enterprise system he expected to be fully commissioned within the quarter. He claimed both firsts — the first nationally available quantum network and the first commercially available quantum computer — and described making the systems available to the community and nationally, focused on energy grid optimization and energy security, using what he called pretty much the full stack of IonQ services.
Why it matters to the full stack: Full stack of IonQ services from a customer’s mouth is the platform thesis stated by the buyer rather than the seller, and this is the only account on the day where that phrase is supported by a live deployment. The commissioning timing is also the most immediately checkable commitment made by anyone: a customer-stated date inside the current quarter, which now sits in the Delivery Ledger as an open item.
Q2. Does EPB depend on IonQ to use the machine? [PUT TO THE RECORD]
No, and this is the answer that separates EPB from a reference account. He said EPB has eight fellows via a grant, uses the IonQ training platform to train them, and — the operative clause — “we are writing our own algorithms and running our own programs to service the energy needs of the industry.”
Why it matters to the full stack: Customer-developed algorithms are the difference between a pilot and an installed capability. A deployment where the customer’s own staff produce the workloads generates utilization that does not depend on IonQ’s services organization, which is both a better margin structure for IonQ and a much stronger signal of durability. It is the single best piece of evidence on the day that IonQ machines get used rather than showcased.
Q3. How long has this actually been going on? [PUT TO THE RECORD]
A decade. He described partnering ten years ago with Oak Ridge National Laboratory, Los Alamos National Laboratory and Qubitekk — now inside the IonQ family — and successfully testing quantum key distribution across a twenty-one kilometre span on EPB’s own fiber network.
Why it matters to the full stack: The Qubitekk detail is quietly important: EPB’s original quantum partner was acquired by IonQ in late 2024, which means this relationship predates IonQ and was inherited rather than sold. For a market assessing whether the acquisition programme bought assets or relationships, a customer with ten years of operating history who then expanded into computing is evidence for the latter.
Q4. What is EPB building beyond its own operations? [PUT TO THE RECORD]
A regional ecosystem. IonQ has landed an office and an R&D facility in Chattanooga. Vanderbilt has announced a satellite office bringing 250 faculty or staff over three to five years. There are partnerships with UT Chattanooga and other entities. He described a supportive city, county and state government wanting to invest behind quantum, and summarized the strategy as first mover — “momentum gets momentum.”
Why it matters to the full stack: An anchor customer that recruits a research university, a municipal government and a state to the same technology creates switching costs no contract can produce. It is also the template IonQ would want to replicate: a regional cluster where the customer does the ecosystem-building. Two hundred and fifty Vanderbilt staff over three to five years is a slower-moving and more consequential commitment than most commercial announcements made that day.
Q5. Why did a power utility do this at all? [PUT TO THE RECORD]
The workload is its own. Energy grid optimization and energy security are the stated focus, and EPB is servicing the energy needs of the industry rather than experimenting for its own sake. He also noted expansion into quantum memories in Chattanooga.
Why it matters to the full stack: Grid optimization is a combinatorial problem with immediate economic value and a regulated operator with a mandate to solve it — which is a materially better near-term fit than most publicised quantum use cases. The expansion into quantum memories also puts Lightsynq technology in front of a commercial customer, which is a commercialization path for an acquisition the market has so far understood only as an interconnect research asset.
Q6. What does the sector need from government? [PUT TO THE RECORD]
Coordination more than money. He said the future will need an unbelievable amount of teamwork and connectivity across organizations, and that government support and coordination of those efforts “will absolutely pull forward the technology and continue to push the US to a leadership position in quantum.”
Why it matters to the full stack: Coming from a municipal utility rather than a vendor, this is a demand-side argument for the executive-order environment that Shapiro and Cardillo discussed from the supply side. It also identifies the specific policy good a customer values — coordination — which is cheaper to supply than subsidy and therefore more likely to be supplied.
21. Dr. Matthew Rabinowitz — Natera and MyOme
Founder and Executive Chairman, Natera; founder, MyOme · Panel remarks 2:03:45, 2:12:54
Standing: Customer and collaborator. Serial entrepreneur across aeronautics, AI and biotechnology. No disclosed IonQ employment or board relationship. Brought the only quantitative result on the day produced outside IonQ.
Owns in the stack: The scientific case for why quantum matters to AI, and the only externally produced experimental number presented at the event.
Layers evidenced: L9 applications and solutions — carrying the only externally produced quantitative result of the day
Q1. Why is a genomics entrepreneur interested in quantum computing? [PUT TO THE RECORD]
Because of an unresolved question in AI. He said there are problems where it is still not understood whether the limit is the available data or the architectures and the ability to find the global minimum within them. That is the question quantum might answer, and he framed his interest as being ahead of the possibility rather than reacting to it.
Why it matters to the full stack: This is a fundamentally different investment case from the sector’s usual chemistry-simulation argument. If quantum computing addresses an architectural limitation in machine learning rather than only a simulation problem in chemistry, the addressable market is the AI industry rather than the pharmaceutical one. It is speculative, and he presented it as such, but it is the most economically consequential idea anyone put on that stage.
Q2. What is the actual computational problem, in terms a non-specialist can hold? [PUT TO THE RECORD]
He built it deliberately. Take a hundred amino acids, where the first two can rotate relative to one another and whatever they do is relevant to every other configuration including the hundredth. Assume four permutations between each pair, as he said IonQ did in one of its papers, and there are four to the hundredth power hypotheses to evaluate. Computationally non-tractable.
Why it matters to the full stack: Four to the hundredth power, stated by an outside scientist rather than by the vendor, is the clearest articulation on the day of why classical scaling does not solve this class of problem. It also supplies the quantitative backing for Chad’s protein-folding ladder — and quietly explains why moving from fourteen to sixteen amino acids is a harder achievement than the small numbers suggest.
Q3. Nature solves this problem constantly. What does that tell us? [PUT TO THE RECORD]
He drew the observation out fully. Classical systems like AlphaFold solve this kind of problem but do not solve it robustly. Yet in nature, a protein made in a cell from a DNA sequence always finds the global minimum — by what he described as a combination of quantum tunneling and jiggling — and never gets stuck in a local minimum the way today’s AI algorithms do.
Why it matters to the full stack: This is the most elegant argument made at the event and it belongs to a customer. It reframes quantum computing not as a faster classical computer but as a machine that solves optimization the way physical reality does. Whether that translates into engineered advantage is unproven, and he did not claim otherwise — but it explains why a sophisticated buyer would fund the exploration before the proof exists.
Q4. What is the healthcare application, and how large is it? [PUT TO THE RECORD]
He listed protein binding, neoantigen prediction for personal cancer vaccines, predicting patient response to therapies, and prognostic applications. On MyOme he was specific about scale: models built with leaders in the field suggest more than two hundred billion dollars of savings to the United States healthcare system by predicting susceptibility to common and rare diseases early enough to treat them cheaply, and to delay or stop onset.
Why it matters to the full stack: A two-hundred-billion-dollar figure is a modelled projection from a private company and should be held as such rather than repeated as a market size. Its analytical value is in direction rather than magnitude: it identifies prediction and prevention, not treatment, as where the economic value concentrates — which is the segment where noisy multi-modal data most rewards better optimization, and therefore where quantum has the most plausible entry.
Q5. What data does Natera bring that makes this tractable? [PUT TO THE RECORD]
He described the capability underneath it — seeing single molecules from a tumour in the blood, which lets Natera catch cancer recurrence months or years before clinical symptoms, assess therapy response better than imaging, and determine after surgery whether all cancer has been cleared and who will benefit from chemotherapy. The result is curated clinical data with AI models, upfront tumour and RNA sequence, and outcomes. He named the hard target: less than one percent of the mutations in a tumour will be good neoantigens.
Why it matters to the full stack: The needle-in-a-haystack ratio is what makes this a quantum-relevant problem rather than a data-volume problem — the difficulty is selection under noise, not scale. It also describes an asset IonQ cannot buy and a competitor cannot replicate: curated longitudinal oncology outcome data. Whichever quantum company gets access to it has an application advantage independent of hardware.
Q6. Is there a result yet? [PUT TO THE RECORD]
Yes, with a clearly stated limit. Following an IonQ paper published in May on fine-tuning a large language model to classify text sentiment, his team emulated a similar architecture: an approximately eight-hundred-length vector embedding from a standardized large language model, then a quantum output layer as the fine-tuning head, tested first with six qubits and then with fourteen. Model power went from 0.70 AUC to 0.83. He stated plainly, without being asked, that this was a simulation of the quantum hardware rather than a run on it.
Why it matters to the full stack: This is the only quantitative experimental result presented on the day by someone who does not work for IonQ, and the unprompted simulation caveat is what makes it usable. Zero-seven-zero to zero-eight-three is a large move in a metric clinicians take seriously. The unresolved question is whether it survives the transition from simulation to hardware, and that is precisely the question a Superion-class machine exists to answer — which makes this the most direct link on the day between the hardware roadmap and a paying commercial outcome.
Q7. Why bring this to a quantum company now rather than later? [PUT TO THE RECORD]
He answered with a story about being early and losing it. In 2010 he held a patent on training multi-layer nonlinear dynamic systems with embedded memory. Stanford let it expire over a thousand-dollar annual fee and an inability to reach him — the year before Google Brain began working intensively on training multi-layer neural architectures with memory. He said he never dreamed at the time what deep learning would become, and that he is now trying to be ahead of these possibilities.
Why it matters to the full stack: This is why a sophisticated technologist engages before the economics are proven, and it is a more honest account of the buying decision than any procurement rationale. It also tells IonQ something about its own market: the early adopters are not buying computation, they are buying option value on an architectural shift, which is a different sales motion and a different customer profile than the one a hardware roadmap implies.
22. Alessio Butti — Italy
Undersecretary of State for Technological Innovation; directed the adoption of Italy’s first national quantum strategy · Keynote 2:21:37–2:27:25
Standing: A serving official of a G7 government, speaking on its behalf. His office is Sottosegretario di Stato — Undersecretary of State to the Presidency of the Council of Ministers — with the technological innovation portfolio. That is a delegated government office, not a cabinet ministry, and this report does not call him a minister. No commercial relationship with IonQ. The only sovereign voice on the programme.
Owns in the stack: The sovereign demand case — what a national government wants from a quantum supplier, stated by the government rather than inferred by the supplier.
Layers evidenced: L10 distribution, trust and governance — sovereign demand stated by the buyer
Q1. What is Italy’s ambition, and is it funded by policy? [PUT TO THE RECORD]
He stated it as national intent: Italy intends to become one of Europe’s leading centres of quantum technologies, and under his direction Italy adopted its first national strategy for quantum technology, issued last July at the same time as Europe’s. He was specific that the strategy goes beyond computing — it encompasses quantum communication and security, sensing, research, skills and the development of a strong industrial ecosystem.
Why it matters to the full stack: A national strategy spanning computing, communications, security and sensing is a procurement frame that maps almost exactly onto IonQ’s four pillars, which is the structural reason a platform company has an advantage in sovereign sales that a single-product company does not. A sovereign buyer purchasing to a national strategy buys the whole frame or none of it.
Q2. Does Italy want to buy quantum technology or build it? [PUT TO THE RECORD]
Build it. He was unambiguous: “We don’t want Italy simply to use quantum technology developed elsewhere. We want to help create them.” He framed international cooperation as the route — the United States bringing technological leadership and entrepreneurial strength, Italy bringing scientific excellence, engineering talent, universities and industrial capability.
Why it matters to the full stack: This distinguishes a sovereign customer from a large commercial one and explains what IonQ is actually selling into these relationships: local capability, not delivered machines. It also validates the on-premises and sovereign deployment optionality Chad described — a government that wants to co-create will not accept cloud access to a machine in another jurisdiction.
Q3. What is Italy’s relationship with IonQ specifically? [PUT TO THE RECORD]
He welcomed IonQ Italia, led by Marco Pistoia, as bringing global quantum expertise to Italy and investing in talent, research and universities, and said the decision to establish it reflects the collaborative approach Italy wants. He noted IonQ is a founding member of the Q alliance, which connects companies, universities and research institutions and provides researchers access to advanced quantum compute systems. He also referenced an MOU signed the previous year and a return engagement at the Como Lake Digital Innovation Forum in Cernobbio the following month.
Why it matters to the full stack: Founding membership in a national alliance and a named local operating entity are structural positions rather than contracts, and they are difficult for a competitor to displace because they were built through relationship rather than procurement. The release announced the following morning added the commercial layer the room did not hear: a national go-to-market partnership between IonQ Italia and Lutech.
Q4. What does Italy want the ecosystem to look like? [PUT TO THE RECORD]
Open, competitive and technology-neutral, in line with the European Union. He also insisted that technology alone is not enough — investment in people, skills and trust, and connecting research to real economic and social needs.
Why it matters to the full stack: Technology neutrality is a constraint on IonQ as much as an opportunity: it means the Italian position cannot be locked, and that competitors retain access. The honest reading of the Italian relationship is therefore a strong incumbency built on early presence, not an exclusive one. That is a materially different asset from a sovereign contract, and it should not be modelled as one.
Q5. How will Italy judge whether quantum worked? [PUT TO THE RECORD]
By outcomes rather than by capability. He said quantum technology will be measured by the problems it solves, the productivity it creates, the knowledge it expands and the opportunities it gives citizens.
Why it matters to the full stack: A sovereign buyer measuring on solved problems rather than delivered qubits is aligned with exactly the time-to-solution framing Chad described from the commercial side, and it favours a supplier with an applications organization over one with a better device. It also sets the evaluation criterion under which IonQ Italia will eventually be judged — and it is a criterion no quantum company can currently satisfy at scale.
Q6. Why did the renaissance framing land? [PUT TO THE RECORD]
Because he had reached it independently. He told de Masi he was pleased when he first heard him mention the quantum renaissance in Rome the previous year, because it was a concept he had been discussing himself. He extended it: Italy has an extraordinary scientific tradition, and the Renaissance was not about looking backward but about moving forward.
Why it matters to the full stack: Shared vocabulary between a vendor and a serving government official is a marker of relationship depth rather than a rhetorical flourish, and it is the kind of thing that takes years to build and cannot be purchased. In sovereign business, where the decision cycle is measured in governments rather than quarters, that is a durable form of advantage — and also a fragile one, since it attaches to individuals.
23. Gen. John W. Raymond
IonQ board member · Panel remarks 2:30:47, 2:37:39, 2:44:40
Standing: Thirty-five and a half years in the Air Force by his own account; led the establishment of the United States Space Force in December 2019 and served as its first Chief of Space Operations, holding four-star rank in both services. Sits on IonQ’s board — his standing is external, his relationship is not.
Owns in the stack: The institution-building analogy, and the translation of research into warfighter capability.
Layers evidenced: L10 distribution, trust and governance — institutional execution and capability transition
Q1. What does 2019 have to do with 2026? [PUT TO THE RECORD]
He drew the parallel himself. Reflecting on the environment when the Space Force was established, he described a nation becoming much more reliant on space and “a sense of national urgency associated with moving at speed.”
Why it matters to the full stack: Speed under national urgency is the exact condition under which institutions get built badly or well, and the person making the comparison ran the last successful instance of it. It frames IonQ’s fifteen-month acquisition programme as an institution-building exercise rather than a buying spree — which is a generous frame, and one worth naming as generous given his board seat.
Q2. What attracted you to IonQ? [PUT TO THE RECORD]
The team, and specifically its integrity. He said it was a team that operated with the same integrity he saw in the Space Force and in military service, and that was as mission focused as they had been in uniform. It felt, in his words, like a good fit.
Why it matters to the full stack: Integrity and mission focus are the criteria a national security customer actually applies, and they are not assessable from outside an organization. His answer is testimony rather than evidence, and this examination marks it as such — but the willingness of someone with his record to attach his name is itself a costly signal, because reputational capital of that kind is spent once.
Q3. Is there an organizational parallel between Space Force and IonQ? [PUT TO THE RECORD]
A precise one, and it is the most analytically useful thing he said. Establishing the Space Force meant taking a majority from the Air Force but also bringing people from the Army, the Navy and the Marines, plus people from off the street, from colleges and from the academies — and having to “integrate these people all together into one high-functioning team that our nation required.” He said he sees that at IonQ.
Why it matters to the full stack: This is the closest thing on the day to an expert assessment of the integration question, delivered by someone who has run the hardest version of it. Ten companies absorbed in fifteen months is a personnel integration problem before it is a technology one, and the failure mode is cultural rather than technical. His comparison is not proof — he is a director — but it identifies the correct variable, and Section 3 tests it against what the stage actually showed.
Q4. What does IonQ do that a national security customer cannot get elsewhere? [PUT TO THE RECORD]
Translation. He deflected the technical credit with characteristic self-deprecation — he is not the smart guy, which is why they sat him at that end of the panel — but was direct about the capability: taking the brains of the people the audience had heard from and translating them “into real capabilities and put them in the hands of the warfighter is nationally critical. That’s what IonQ does really, really well.”
Why it matters to the full stack: Translation from research to fielded capability is where most defence technology programmes fail, and the gap it names is the same one Muller described from the IARPA side and Sonderman described as the valley of death in semiconductors. Three people from three unrelated domains identifying the same failure mode, and locating IonQ on the right side of it, is a pattern worth more than any one of the three statements.
24. Robert Cardillo
IonQ director and Executive Chairman of IonQ Federal · Panel remarks 2:33:00, 2:39:32, 2:43:39
Standing: Almost four decades in the US intelligence community across seven presidents; sixth Director of the National Geospatial-Intelligence Agency, 2014 to 2019; four years providing intelligence in the Oval Office. Joined IonQ’s board in 2024 and became Executive Chairman of IonQ Federal in September 2025 — a relationship this report states plainly wherever his remarks are used.
Owns in the stack: Intelligence tradecraft applied to the SAR result, and the governance account of how the cryptanalysis paper was handled.
Layers evidenced: L9 applications and solutions, L10 trust and governance
Q1. What is the difference between strategic surprise and strategic failure? [PUT TO THE RECORD]
A distinction he said he learned across four decades: strategic surprise does not come from failing to expect what is coming over the next horizon. “Sometimes you know what’s coming, and yet you don’t prepare. And so I would actually call that strategic failure.”
Why it matters to the full stack: This is the intellectual frame for the entire quantum security business, delivered by someone who spent a career on the consequences. It reclassifies the post-quantum migration problem from a forecasting question to a preparation question — which removes the most common objection an enterprise makes, that nobody knows when Q-Day arrives. Under his framing, the arrival date is not the decision-relevant variable.
Q2. Why this company rather than another? [PUT TO THE RECORD]
Assembly, and then application. He said IonQ “was the only company that I saw bringing together all of the pieces that you’ve seen today,” crediting the founding work and the compute roadmap as the foundation. But what excites him is downstream: the applications, the customers and the missions — because, as he put it, one thing he learned as an intelligence professional is that what they do does not matter unless it affects a better decision for those they serve.
Why it matters to the full stack: Coming from the former director of an intelligence agency, the assembly claim is a competitive assessment made by someone who has evaluated a great many vendors. The decision-outcome framing is also the most rigorous statement of what quantum has to deliver commercially: not capability, but a changed decision. That is a harder bar than the industry usually sets for itself.
Q3. What does the SAR demonstration look like to someone who ran that mission? [PUT TO THE RECORD]
He identified it immediately as central to the business he used to run inside government, where synthetic aperture radar was applied against adversaries to build understanding of their capabilities and intentions. Then he described what IonQ’s applications do with it: taking an overwhelming volume of data — and he stressed the word purposefully — and performing the synthesis between noise and signal that is becoming steadily harder. He gave it a name: “a coherence-from-chaos service. Finding the right info at the right time to make that better decision.”
Why it matters to the full stack: Coherence from chaos is the most useful description of IonQ’s applications value produced by anyone at the event, and it generalizes well beyond intelligence — Rabinowitz’s neoantigen selection, EPB’s grid optimization and Lokada’s agent orchestration are all instances of it. It also converts Shapiro’s Miramar demonstration from a technical curiosity into a mission capability, assessed by someone qualified to make that call. His board relationship means it is corroboration from an insider, not from an independent party, and this examination weights it accordingly.
Q4. Did the last twelve months change the company? [PUT TO THE RECORD]
He said the teammates invited over the past year, many of them on stage that day, have not just complemented the company but “truly scaled” it — and connected that directly to policy, noting that the executive order Raymond mentioned demands exactly that: a scaled ecosystem.
Why it matters to the full stack: Scaled rather than complemented is a precise distinction and the correct one for assessing an acquisition programme. It also identifies a policy tailwind with a specific mechanism: if the executive order requires a scaled ecosystem, then scale itself becomes a qualification criterion, and the fifteen-month programme is a response to a stated federal requirement rather than opportunism.
Q5. How did IonQ handle publishing a result that describes breaking encryption? [PUT TO THE RECORD]
This is the governance disclosure of the day and he volunteered it. He credited the whole team for the conversations that ensued, noting that Rick Muller led many of them with the senior-most levels of the US government, for two stated reasons: “we didn’t want to surprise our partners in government,” and “we wanted to be responsible going forward on the safe-cracking part of it.” He called that the essence of being a good partner and said IonQ had raised its reputational credibility with its most important customers.
Why it matters to the full stack: Pre-briefing the government before publishing offensive cryptanalysis is a governance choice with real commercial cost — it slows disclosure and cedes control of timing — and IonQ made it anyway. For a company whose growth depends on federal and sovereign customers, reputational credibility with those buyers is a balance sheet item that does not appear on the balance sheet. It is also the strongest available answer to the objection that IonQ is commercialising fear: a company doing that does not give the customer advance copies.
25. Rick Muller
IonQ · Panel remarks 2:35:04, 2:41:49
Standing: Roughly two decades running the quantum programme at Sandia National Laboratories, then Director of IARPA. Holds a doctorate in ion trapping — the only speaker with both the science and the government transition experience. Now at IonQ.
Owns in the stack: The government buyer’s actual requirements, and the technical case for ions as a deployment technology rather than a laboratory one.
Layers evidenced: L2 qubit control and the device, L9 applications, L10 trust — federal requirements stated by a former buyer
Q1. What does the government actually want? [PUT TO THE RECORD]
Not what the industry sells. He put it in one sentence: “the government doesn’t want a qubit. They want mission solutions to their hardest problems.” What IonQ brings, in his account, is an amazing set of technologies and a broad platform of capabilities that can be combined to address those challenges.
Why it matters to the full stack: Delivered by a former IARPA Director, this is the clearest statement of federal procurement logic anyone offered, and it is the strategic justification for the platform structure. A government buyer assembling a mission solution from four vendors carries the integration risk itself; a platform vendor absorbs it. That is the specific reason breadth is worth more in this market than in a commercial one.
Q2. Is technology sufficient? [PUT TO THE RECORD]
No, and he was pointed about his own limits. He can tell you whether a qubit will work or whether a clock will work, but it takes people who understand the mission to understand what those solutions are worth. He said one of the things IonQ builds is the team — and joked that for the second year running he had been completely upstaged by the people sharing the stage with him.
Why it matters to the full stack: Pairing scientific capability with mission understanding is the standard failure point in defence technology transition, and he is describing it from the side of the desk that decided which programmes made it. His conclusion that this constitutes a very solid business model is an assessment from a former customer, which is a stronger evidence class than a vendor’s claim about its own model.
Q3. Why ions, in this context specifically? [PUT TO THE RECORD]
He gave Chris Monroe’s original argument and then extended it operationally. Ions are naturally identical qubits and naturally quantum particles that can be controlled classically, and because of that IonQ can build low-cost, low-power solutions that scale rapidly. The consequence he drew is about deployment rather than performance: what that gives you is “not just some anonymous quantum service that you log in to some anonymous host,” but a solution that can be deployed locally or remotely and is genuinely your own.
Why it matters to the full stack: This is the sovereignty argument stated in physics terms, and it is the connective tissue between the trapped-ion modality and the sovereign business Butti described from the buyer’s side. Low power and rapid scaling are what make local deployment feasible; local deployment is what a government that wants to build rather than rent will accept. Singh made the same point commercially when he noted IonQ machines plug into the wall rather than requiring a dedicated power plant.
Q4. What is the oldest thing IonQ has, and does it still matter? [PUT TO THE RECORD]
He answered both halves at once: the ion trap is the oldest technology IonQ has, and he thinks it is uniquely capable of delivering high-trust solutions to sovereign nations, to companies, and ultimately to individuals who need to be able to use that capability.
Why it matters to the full stack: High trust is a defensible market position rather than a technical one, and it is a category where an incumbent with a decade of deployment history and a domestic trusted fab has advantages that a superior device cannot immediately overcome. It also closes the loop with Sonderman’s security-from-design-to-delivery claim: trust is manufactured through the supply chain, not asserted in a datasheet.
Q5. What was your role in the cryptanalysis disclosure? [PUT TO THE RECORD]
He did not describe it himself. Cardillo did, crediting Muller with leading many of the conversations with the senior-most levels of the US government before the result was published.
Why it matters to the full stack: The relevant point is what the assignment required: someone who could reach that level of government, be received credibly, and be trusted to handle an offensive cryptanalysis result responsibly. Very few people at any company possess all three, and this is the clearest instance on the day of a specific hire producing a specific institutional capability. It is also a reminder that the pre-briefing was not merely a policy — it required a person who could execute it.
26. The Analysts — What Seven Questioners Chose to Ask
Seven people had the floor. Six were named, one was not. What they asked is evidence in its own right, because a sell-side analyst spends a scarce question on the thing they believe is least settled. This section examines the question set as a body rather than the questioners individually.
Q1. What did the analysts not ask about? [PUT TO THE RECORD]
The physics. Across seven questions covering photonic interconnect sequencing, cycle time, lithography node, foundry conflict of interest, competitive lead time, scale-out phases, median gate fidelity, crosstalk, Fab 25, tiling competencies, customer engagement patterns, the Nexus acquisition fit, 256 pricing and fiscal 2027 modelling, not one questioner asked whether trapped-ion quantum computing works.
Why it matters to the full stack: This is the most significant fact about the analyst session and it is visible only in aggregate. A year ago the sector’s question set was dominated by feasibility. On September 8 it was dominated by tolerances, schedules, capacity, competitive positioning and revenue modelling — the question set of an industrial business. The analysts recategorized IonQ before most of the market did, and Section 1.1 treats this as the strongest available third-party evidence that the risk moved.
Q2. Which question was the hardest, and did it land? [PUT TO THE RECORD]
Mobley of StoneX asking how the other eight quantum customers get comfortable that SkyWater will not handicap their roadmaps, and how much lead time exists against IBM and GlobalFoundries. It went to the structural contradiction at the centre of the merchant supply thesis. Sonderman answered with process — compartmentalization and no cross-contamination — and then named competitors directly, including his former employer.
Why it matters to the full stack: The question was correctly aimed and the answer was better than it had to be, but the decisive corroboration did not come from the stage answer. It came thirty minutes later from Martinis and Ho, who are among the eight and had signed a multiyear agreement that morning. An analyst asking the right question and an unrelated speaker answering it independently is the most robust evidentiary sequence available at a managed event.
Q3. Which question exposed a genuine disclosure gap? [PUT TO THE RECORD]
Bolton of Needham asking when IonQ would have enough data to discuss median gate fidelity across the 256 qubits, how confident the company is in hitting the fidelity target, and what had been done about crosstalk. Ballance answered around the question — describing short-loop de-riskers, decoupled element-by-element learning, confidence in ultimate integrated performance, and integration work leading to production and early 2027 deployments. No date for device-level fidelity data was given, and no fidelity figure was disclosed.
Why it matters to the full stack: This is the sharpest limit on the day and this examination names it plainly. IonQ disclosed better than four nines on the unit cell but nothing at the device level, and would not say when it would. That is the difference between a validated component and a validated machine, and it is the metric on which the early-2027 deployment commitment will ultimately be judged. Any reader building a position should hold this as the principal open technical question.
Q4. Which question was not answered at all? [PUT TO THE RECORD]
Vijay of Mizuho asked three things: how Nexus Photonics fits, what the 256 is priced at, and how to model fiscal 2027 between core IonQ and SkyWater. Bhaskar answered Nexus thoroughly. Singh handled pricing obliquely and asked, with some humour, whether he had successfully dodged it. Fiscal 2027 was never addressed by anyone, and the session ended.
Why it matters to the full stack: The fiscal 2027 framework is the largest gap the event left, and it is a regression rather than a neutral omission: this series’ own prediction scorecard records the absence of forward-year detail as a clean miss. An investor day is the natural venue for a forward framework, and the analyst asked for it explicitly. Its absence is the single most reasonable criticism of the day, and it is the first question this series will put on the next earnings call.
Q5. Which question produced the most valuable answer? [PUT TO THE RECORD]
Choksi of Northland on cycle time and lithography node. It produced the ninety-nanometre disclosure and Sonderman’s dismissal of the leading-edge premise — that talk of twenty-eight nanometres and below, or two nanometres, is ridiculous for this application — alongside his expectation that cost per qubit stays industry leading.
Why it matters to the full stack: A mature 200-millimetre node at ninety nanometres decouples the entire roadmap from leading-edge capacity, geopolitical exposure around advanced nodes, and the capital intensity that governs the rest of the semiconductor industry. It received the least attention of any major disclosure on the day and is arguably the most important thing said about the cost structure. It is the clearest case of a good analyst question extracting a fact the prepared remarks would not have produced.
Q6. What does the unidentified analyst’s question tell us? [PUT TO THE RECORD]
The unnamed questioner prefaced their question by saying they had no doubts about scale-up or the applications, and then asked about the photonic interconnect scale-out roadmap, noting that phase four is very complicated, along with a question about manufacturing capability for shorter application cycles.
Why it matters to the full stack: The preface is the informative part. An analyst volunteering that scale-up and applications are no longer in doubt, and spending the question on scale-out sequencing instead, is conceding the core thesis in order to probe the periphery. Combined with the absence of any physics question across seven questioners, it indicates the recategorization is not one analyst’s view.
27. What the Examination Establishes
One hundred and twelve questions across sixteen examinations produce four findings that no single speaker’s remarks would support alone.
27.1 Independent Convergence
The most durable evidence at a managed event is not what any speaker says but what unrelated speakers say without coordination. Four convergences stand out.
- The valley of death between research and deployment was named independently by Sonderman in semiconductor terms, Muller from the IARPA side, Raymond in warfighter terms, and Chad through the aviation analogy — four domains, one failure mode, IonQ located on the same side of it each time.
- The foundry as industrial infrastructure rather than supplier was reached by Sonderman from wafer counts, Bhaskar from interconnect physics, Martinis from Nobel-level process knowledge, and Ho from unit economics and financeability.
- The SkyWater speed advantage was described in a dated worked example by Ballance from the engineering side and independently corroborated by Singh from the diligence side, describing the same failure mode at an unnamed alternative.
- Coherence from chaos — Cardillo’s phrase — describes Rabinowitz’s neoantigen selection, EPB’s grid optimization, Lokada’s agent orchestration and Shapiro’s SAR result equally well, though none of those four speakers used it.
Speakers who did not coordinate arrived at the same four conclusions from four different professional vocabularies. That is the hardest evidence a staged event can produce. |
27.2 What the Outsiders Established That IonQ Could Not
Six of the twenty-five voices hold no IonQ employment, board seat or advisory relationship: Martinis, Ho, Lokada, Long, Rabinowitz and Butti. Professor John Bowers, credited from the stage and present in the audience, is not among them — the published deck identifies him as a co-founder of Nexus Photonics, which IonQ acquired. He is a distinguished witness with a real relationship, and this report says so. Between them they supplied the merchant supply validation that resolves the conflict-of-interest question, the only externally produced quantitative result of the day, a live full-stack deployment with customer-written algorithms, the sovereign demand case, and an enterprise argument that post-quantum cryptography alone is insufficient.
Raymond, Cardillo and Muller brought the most impressive institutional standing in the room — the Space Force, the National Geospatial-Intelligence Agency and IARPA — but all three hold IonQ relationships. Their standing is external; their testimony is not independent. Part Two states that distinction wherever their remarks are used, and Section 5 does the same in Part One.
27.3 The Three Genuine Limits
- Device-level fidelity for the Superion 256 was not disclosed and no date for it was given, despite a direct question. Four nines on a unit cell is a component result, not a machine result.
- Fiscal 2027 received no framework of any kind, from anyone, despite being asked for explicitly in the final question of the session.
- Profitability was deferred. No combined EBITDA guidance, no segment margins, and no free cash flow discussion — legitimate one month after a close, and not legitimate at the next earnings call.
27.4 The Finding
Examined one at a time, the twenty-five voices do something Part One can only assert. They show a company where the acquired founders run the acquirer’s roadmaps, where the competitors are customers, where the customers write their own algorithms, and where the hardest question of the day was answered by someone who does not work there.
No single speaker proved the thesis. The absence of a weak link across twenty-five of them is the proof. |
What would change this view is set out in Section 9 and applies to Part Two unchanged. The three limits above are the places to watch: a device-level fidelity figure that disappoints, a fiscal 2027 framework that reveals slower commercial conversion than the platform narrative implies, or a merchant supply position that erodes once a competitor discovers that compartmentalization is easier to promise than to sustain.
Sources for Part Two: the full event record of IonQ’s 2026 Investor Day, New York Stock Exchange, September 8, 2026; IonQ’s six press releases of the same date; IonQ’s Form 10-Q and Form 10-K as filed with the Securities and Exchange Commission. Where the event record and the press releases differ on proper nouns, the releases govern. The speaker citation index at Appendix A and the categorized proceedings at Appendix B follow.
Appendix A — Speaker Citation Index
Every voice at the event, the layer or ecosystem each one evidences, and where each is cited in this report. Twenty-five voices; none omitted. Each entry states the speaker’s relationship to IonQ so a reader can weight the testimony accordingly.
Speaker | Affiliation | Evidences | Cited in |
Niccolò de Masi | IonQ — Chairman and CEO | Platform frame, Superion launch, security logic, merchant supply, Italy | 1, 2.1, 3.4, 4 (L4, L7, L10), 6.1, 6.3, 6.7, 6.10, 6.11, 7.1, 11 |
Inder Singh (rendered "Indra Singh") | IonQ — CFO and COO | Foundry procurement, segment structure, sensing rationale, land and expand, cloud economics | 1, 2.4, 3.3, 4 (L1, L3, L7, L8, L10), 6.1, 6.2, 6.5, 6.9, 6.11, 7.1, 8.2 |
Dr. Chris Ballance | IonQ — President, Quantum Computing | Superion architecture, QEC, Shor's compilation, tiling, cadence, hybrid stacks | 1, 2.1, 2.2, 2.4, 3.2–3.5, 4 (L2–L5, L9), 6.4, 6.9, 8.2, 9, 10, 11 |
Tom Sonderman | IonQ / SkyWater | Foundry model, wafer scale, node, parallelism, IP compartmentalization, competition | 1, 2.4, 3.2–3.6, 4 (L1, L3), 6.1, 6.2, 7, 8.2, 9, 10 |
Jordan Shapiro | IonQ — Quantum Platform | Q-Day, security stack, networking, SAR, sensing, sovereign demand | 1, 2.1, 2.3, 3.4, 4 (L6–L8), 6.2–6.6, 6.10, 7, 7.1 |
Mihir Bhaskar | IonQ / SkyWater Quantum Solutions — co-founder of Lightsynq; now leads the merchant foundry business | Kilohertz interconnect, foundry platforms, Nexus integration, chip-scale sensing | 1, 1.2, 2.2, 3.2–3.5, 4 (L1, L6), 6.1, 6.4, 6.5, 7, 7.1, 9, 10 |
Chad | IonQ — go-to-market | Time to solution, life sciences, P450, IEEE results, team scale, upgrade pattern | 1, 3.4, 4 (L9, L10), 6.4, 6.7, 6.8, 6.9, 6.12, 7, 8.2 |
Rick Muller | IonQ; formerly Sandia and IARPA | Government buying behavior, modality argument, sovereign trust | 3.7, 4 (L2), 6.2, 6.10, 9 |
Gen. John W. Raymond | IonQ board; founder, US Space Force | Institutional execution, warfighter translation | 3.7, 5, 6.2 |
Robert Cardillo | IonQ director; sixth Director of the NGA | SAR and intelligence analysis, government pre-briefing, governance layer | 2.3, 3.7, 4 (L10), 5, 6.2, 6.6, 6.7, 7, 10 |
Dr. John Martinis | Qolab — co-founder and CTO; Nobel laureate | Silicon and fabrication; competitor endorsement of the foundry | 5, 5.1, 7 |
Alan Ho | Qolab — CEO | Foundry unit economics; SC250 multiyear agreement; Midwest ecosystem | 5, 5.1, 6.1, 6.12, 7 |
Professor John Bowers | Co-founder, Nexus Photonics (acquired by IonQ); UCSB | Photonic integration lineage | 4 (L6), 5 |
Alessio Butti | Italy — Undersecretary of State for Technological Innovation | Sovereign strategy, national ecosystem, Q alliance | 5, 6.10, 6.12 |
Robert Long | EPB — President of Strategic Initiatives | Operating network and computer, own algorithms, regional ecosystem | 5, 5.2, 6.4, 6.8, 6.12 |
Dr. Matthew Rabinowitz | Natera and MyOme — Founder and Executive Chairman | Independent reproduction, protein folding, oncology and genomics applications | 5, 5.3, 6.7, 7, 9 |
John Lokada | ServiceNow — Futures Director | Enterprise workflow, secure optimization, agent orchestration | 5, 6.3, 6.9 |
John McPeak | Rosenblatt Securities | Interconnect sequencing question | 2.2, 5.4 |
Nihal Choksi | Northland Capital Markets | Cycle-time and node question — source of the eight-to-two-months figure | 2.4, 5.4, 8.2 |
Gary Mobley | StoneX | Foundry conflict-of-interest and competitive lead-time question | 4 (L1), 5.4 |
Quinn Bolton | Needham | 256 fidelity data, crosstalk, Fab 25 | 4 (L3), 5.4, 9 |
Craig Ellis | B. Riley Securities | Tiling competencies; cross-technology customer engagement | 4 (L3), 5.4 |
Vijay | Mizuho | Nexus fit, 256 pricing, FY2027 modeling | 4 (L6), 5.4, 9 |
Unidentified analyst | — | Photonic interconnect scale-out phases and manufacturing capacity | 5.4 |
Corporate film | IonQ | National-security and domestic-manufacturing framing | Referenced as context; no analytical claim rests on it |
Where the event transcript and IonQ's published releases disagree on a name, figure or process designation, this report follows the releases. Corrections applied in this revision: Qolab (not QLab), SC250 (not SD250), Lightsynq (not Lightsync), Mihir Bhaskar, Inder Singh, and a nine-to-two-month design cycle (not eight-to-two). Chris Monroe, IonQ's founder, was present in the audience and referenced by several speakers but did not deliver remarks.
Quantum computing companies do not talk like that. Manufacturers do. |
Appendix B — The Full Proceedings, Categorized
This appendix is the evidentiary substrate for the entire report. Every substantive statement made across the nearly three hours of proceedings is indexed here under one or more of twenty-three categories, cited by speaker and turn timestamp in the same form used throughout the body text — so any claim in Sections 1 through 10 can be traced to its source and read alongside everything else that was said on that subject.
Entries are ordered as they occurred in the event. A statement appearing under more than one category is intentional cross-referencing, not duplication: a customer deployment that is also a sovereign relationship and also a networking milestone is indexed three times, because a reader coming from any of those three directions needs to find it. The categories are this series’ analytical framework, not IonQ’s, and they are the direct successor to the seventeen-category index built for the Q2 2026 earnings call.
Index | Category |
I | Financial results, guidance and segment reporting |
II | Corporate structure, M&A and integration |
III | Quantum computing hardware and the Superion roadmap |
IV | Error correction, algorithms and the software stack |
V | Semiconductor manufacturing and the quantum foundry |
VI | Integrated photonics and interconnects |
VII | Quantum networking |
VIII | Quantum security and cryptography |
IX | Quantum sensing, PNT and timing |
X | Space, remote sensing and Earth intelligence |
XI | Government, defense and national security |
XII | Sovereign and international |
XIII | Life sciences, healthcare and pharmaceuticals |
XIV | Energy and utilities |
XV | Enterprise software, financial services and AI convergence |
XVI | Industrials, logistics, oil and gas |
XVII | Go-to-market, customers and commercial motion |
XVIII | Merchant supply and the industry-enablement business |
XIX | Talent, organization and leadership |
XX | Academic, workforce and regional ecosystem |
XXI | Competitive positioning claims |
XXII | Forward-looking statements and dated commitments |
XXIII | Analyst Q&A index, and transcription uncertainty flags |
I. Financial Results, Guidance and Segment Reporting
[Singh 47:01] FY2026 consolidated guidance revised to $450–460 million, issued by press release the same morning.
[Singh 47:01] IonQ standalone FY2026 guidance of $280–290 million reaffirmed; granular sub-guidance from the Q2 call explicitly not restated, with a pointer back to the Q2 transcript.
[Singh 47:01] SkyWater bridge disclosed in full: $608 million street consensus as of July 30, less $368 million delivered through July, leaves $240 million for the five-month stub; less $70 million of intercompany elimination.
[Singh 47:01] SkyWater interim figures flagged as unaudited, with purchase-price adjustments still outstanding and characterized as noise-level.
[Singh 47:01] No EBITDA guidance provided; CFO stated it would come once auditors had completed work, and that the company would focus on EBITDA margin over gross margin because it captures R&D.
[Singh 47:01] Stated intent to drive EBITDA margins higher over time; acknowledgment that R&D spend likely exceeds most peers.
[Singh 47:01] New four-bucket segment frame introduced: quantum hardware and quantum services on the IonQ side; SEMI semiconductor foundry and quantum foundry / advanced technologies on the SkyWater side.
[Singh 47:01] Quantum services described as growing faster than the computing business and more profitable than many other lines.
[Singh 47:01] Balance-sheet philosophy stated: the strongest balance sheet required for success, explicitly not too strong and not too weak; intent to fund multi-year innovation without near-term financing pressure.
[Singh 47:01] $148 billion quantum-computing TAM cited, explicitly attributed to third parties and explicitly described as covering nearly every other quantum company — with IonQ positioned in additional TAMs beyond it.
[Singh 47:01] Capital deployed characterized as approximately $1 billion on Oxford Ionics and nearly $2 billion on SkyWater, framed as exceeding what many governments have committed to quantum.
[de Masi 4:40] Revenue quintupled year on year; pre-SkyWater run-rate characterized as the better part of $300 million.
[de Masi 4:40] Q2 revenue characterized as roughly twice that of the rest of the public quantum sector combined.
[Singh 1:28:56] Superion pricing question explicitly declined; framed instead as a margin-retention-versus-sharing decision over coming years and a total-cost-of-ownership argument.
[UNDISC] FY2027 revenue framing was directly requested by Mizuho and not answered. This is the largest single disclosure gap of the day and the first item on the forward watch list.
II. Corporate Structure, M&A and Integration
[de Masi 4:40] SkyWater merger characterized as taking seven months to close and as historically significant; IonQ described as the largest quantum company in history and the first vertically integrated one.
[Singh 47:01] Close approval received July 31, days before the Q2 earnings call, on the previously guided timeline; guidance was withheld at that time as too recent.
[Singh 47:01] Combined entity described as operating under the single IonQ name, with Sonderman continuing to run SkyWater.
[Bhaskar 1:27:36] Nexus Photonics acquisition characterized as a one-plus-one-equals-thirty combination with SkyWater: Nexus brings design and integration IP for integrated photonics, SkyWater brings manufacturing.
[Bhaskar 1:27:36] First commercial impact of Nexus stated to be in quantum sensing devices — replacing hand-assembled discrete components with on-chip integration.
[Bhaskar 1:31:00] Lightsynq quantum memory, acquired the prior year, identified as the enabling technology behind the >1 kHz interconnect result.
[de Masi 4:40] Quantum networking and quantum memory business explicitly attributed to the Lightsynq acquisition.
[Long 2:08:09] Qubitekk identified as now inside the IonQ family, in the context of EPB's decade-old QKD work.
[Bhaskar 1:31:00] New role disclosed: Bhaskar stepping into a position working directly with Sonderman at SkyWater on the foundry platforms.
[de Masi 2:45:40] Commitment restated to existing SkyWater customers, covering both team retention and IP protections, described as absolute.
[INFER] The Nexus–SkyWater framing and Bhaskar's new dual role together suggest the photonics business is being organized as a foundry platform line rather than as a component supplier to IonQ's own systems. That is a materially different economic shape and is treated as a distinct ecosystem in Part III.
III. Quantum Computing Hardware and the Superion Roadmap
[de Masi 4:40] Superion product line launched; described as upgradable, cloud- and sovereign-accessible, and expected to lead on power per unit dollar, per unit energy and per unit space.
[de Masi 4:40] Generational ladder stated as 256 → 10K → 20K → 200K.
[Ballance 18:36] Three design constraints named as the origin of Superion: radical simplification, learning in months rather than years, and demonstrated generation-over-generation scale.
[Ballance 18:36] Electronic qubit control replaces laser control, removing the middle layer and connecting classical control electronics directly to the quantum chip; framed as both a manufacturability and a reliability gain.
[Ballance 18:36] Quantum multiplexed I/O introduced as the unit-cell-to-array scaling technology, with an LED-to-display-panel analogy; covered by a described unique patent portfolio.
[Ballance 18:36] Over four nines two-qubit gate fidelity achieved on the electronic-qubit-control unit cell.
[Ballance 18:36] First Superion 256 chip shown; wafer-scale production underway at SkyWater; packaged quantum processors being built.
[Ballance 18:36] Systems designed to be data-center deployable and mass manufacturable; physical infrastructure footprint built out over the past year.
[Ballance 18:36] Ion qubits now loaded in the first prototype systems — announced on stage as new.
[Ballance 18:36] First customer-site deployments expected in early 2027.
[Ballance 18:36] More ion-trap chips and processors built in the last six months than in the company's entire prior history.
[Ballance 18:36] Scaling path beyond 10K: integrated cryogenic CMOS for multiplexing, then two-die 2.5D tiling, then more dies plus higher per-die qubit density.
[Ballance 18:36] De-risking CMOS wafers already in production at SkyWater and displayed in the room.
[Ballance 1:20:02] Short-loop prototype de-riskers described as the mechanism for parallel learning; integration and production-scale validation named as the current work.
[Ballance 1:23:45] Tiling characterized explicitly as an integration problem rather than a quantum problem — die bonding tolerances and electrostatic qubit transfer across junctions, with studies showing IonQ comfortably inside industry-standard tolerances.
[Ballance 1:23:45] Named residual risk is operational excellence and yield, not any single technology risk; goal stated as making the packaging boring.
[Singh 1:28:56] Claim that no other company has a 256 chip, and nothing known at 10,000; logical-qubit advantage attributed to the ion-trap modality.
[Muller 2:41:49] Modality argument restated: ions are naturally identical quantum particles controllable classically, enabling low-cost, low-power, rapidly scalable systems deployable locally or remotely.
[Singh 47:01] Aria-to-Superion-10K qubit count characterized as a 400× increase; separate cost-per-qubit chart framed as a quantum Moore's Law.
[Singh 47:01] Machines described as not requiring dedicated grid infrastructure — plugging into the wall rather than needing a dedicated reactor.
[FACT] The ions-in-prototype-systems statement is the single most consequential hardware disclosure of the day. It converts the 256 program from a fabrication story into an operating-device story, and it is what makes the early-2027 deployment date a scoreable commitment rather than an aspiration.
IV. Error Correction, Algorithms and the Software Stack
[de Masi 4:40] Walking Cat described as the world's first shovel-ready fault-tolerant architecture, published April 2026.
[Ballance 18:36] Walking Cat characterized as the most detailed full-stack quantum computing architecture published, spanning hardware, QEC, compiler and application layers; unifies multiple QEC codes into one.
[Ballance 18:36] Walking Cat plus the 256 compute fabric stated to carry through to the 10,000-qubit system.
[Ballance 18:36] Customized quantum error-correction codes demonstrated on real hardware at breakeven performance, in the last six months.
[Ballance 18:36] Mega-quantum-operation real-time decoding stack published within the prior fortnight, tailored to the 10K device and running on efficient classical computation in parallel with quantum hardware.
[Ballance 18:36] First full end-to-end compilation of Shor's algorithm made public — described as not an estimate, with exact qubit counts and every physical operation modeled.
[Ballance 18:36] Over 30× improvement in CCZ magic-state factories disclosed as a byproduct of the Shor's work.
[Ballance 18:36] Time-to-solution speedups demonstrated on pure-play quantum algorithms, and on hybrid stacks combining QPUs, GPUs and CPUs.
[Chad 1:42:36] Quantum Fourier transform cited as a workload where IonQ measures 1,000–10,000× advantage; noted as a component of Shor's algorithm.
[Chad 1:42:36] Quantum phase estimation identified as the recurring primitive in chemistry and life-sciences use cases.
[Chad 1:42:36] Results published as a public repository for reproduction, alongside third-party validation.
[Chad 1:42:36] Paper with NVIDIA and Oak Ridge on using large language models to improve quantum algorithms; GPU and QPU clusters operating together.
[Ballance 18:36] Full-stack work described as a virtuous flywheel: solutions tailored to hardware, hardware and software tailored to anticipated future solutions.
[ARG] The Shor's compilation and the ECC resource estimate are the same body of work presented to two different audiences — a scientific claim in the roadmap section and a market-creating claim in the security section. Treating them as one artifact rather than two announcements is the more accurate reading, and it explains why the government pre-briefing was necessary.
V. Semiconductor Manufacturing and the Quantum Foundry
[Sonderman 31:11] Partnership described as moving out of physics demonstration into manufacturing at scale, delivering both cycle speed and volume.
[Sonderman 31:11] Thousands of quantum wafers running in the fab, approximately one third tied to IonQ.
[Sonderman 31:11] Technology foundry concept and technology-as-a-service business model described as purpose-built to bridge the lab-to-fab valley of death.
[Sonderman 31:11] Positioning against conventional and specialty foundries: SkyWater embraces customization where others require standardization.
[Sonderman 31:11] Claim that SkyWater is the only quantum foundry, operating for over ten years, with security and IP protection built into the model.
[Sonderman 31:11] Quantum framed as sovereign infrastructure, with IP and national-security protection presented as foundational rather than incidental.
[Sonderman 31:11] Nine quantum customers at SkyWater as of the Qolab announcement; explicit framing that SkyWater is creating a quantum foundry industry, not only serving IonQ.
[Sonderman 31:11] Minnesota front-end and Florida advanced packaging named as the quantum footprint; heterogeneous integration named as the scaling mechanism.
[Sonderman 31:11] Stated goal of continuing to move IonQ's roadmap to the left.
[Sonderman 1:13:58] IP compartmentalization described in response to a direct question on conflict of interest; assurance that non-IonQ roadmaps will not be impeded; rising-tide framing for capability spillover.
[Sonderman 1:13:58] Competitive commentary: reference to Anduril's acquisition of HRL; IBM's 2015 sale of its microelectronics business to GlobalFoundries; GlobalFoundries characterized as a scaled specialty foundry rather than an innovation engine.
[Sonderman 1:11:46] 90-nanometer, 200-millimeter process identified as the node in use; explicit rejection of claims that 28nm or 2nm are required.
[Ballance 1:10:04] Foundry-culture anecdote: a supply-chain part quoted at a month by a major industrial foundry was sourced from Korea and installed within days at SkyWater, against a competing two-month slip on a six-month timeline.
[Singh 1:12:19] During the Oxford Ionics acquisition, other candidate quantum foundries were evaluated and found too slow, lacking a quantum learning curve, and requiring parent-company approval.
[Sonderman 1:17:18] Minnesota and Florida capacity described as more than sufficient for compute and platform requirements; scale requirement contrasted with smartphones and AI data centers — fewer parts, many more iterations.
[Sonderman 1:17:18] Five device generations running in parallel, compared with three process nodes in parallel during the speaker's AMD tenure.
[Sonderman 1:21:05] Fab 25 characterized as part of the foundry ecosystem, providing purchasing power and supply-chain optimization; quantum work remains Minnesota and Florida.
[Singh 1:22:07] Fab 25 framed as optionality requiring retooling if repurposed; no quantum work performed there to date.
[Singh 47:01] Trusted-fab argument stated end to end: security from design through engineering, manufacturing and delivery; SkyWater described as the only trusted fab of scale in the United States, with competitors targeting 2029.
[Singh 47:01] Multiple simultaneous prototypes named as a capability competitors lack, enabling faster fail-fast selection.
[Bhaskar 1:31:00] Launch of the industry's first dedicated quantum foundry platforms — scalable, reproducible, and explicitly open to atom, ion, superconducting, photonic and spin-based processor companies.
[FACT] Two thirds of the quantum wafers in SkyWater's fab are for parties other than IonQ. That single number is the strongest evidence offered all day that the foundry is a business in its own right rather than an internal cost center.
VI. Integrated Photonics and Interconnects
[Bhaskar 1:31:00] Two foundry platform elements named, on the argument that there are only two physical ways to route quantum signals — light and superconducting electronics — presented as the quantum analog of copper and fiber.
[Bhaskar 1:31:00] Integrated photonics platform anchored on Nexus Photonics IP; Professor John Bowers credited as the pioneer of lasers on chip and present in the audience.
[Bhaskar 1:31:00] Nexus described as having pioneered on-chip quantum lasers, enabling a fraction of the size, weight, power and cost with higher manufacturability, yield and reliability.
[Bhaskar 1:31:00] Superconducting platform anchored on the SC250 process, with Qolab as launch customer.
[Bhaskar 1:27:36] Sensing devices identified as the first application of Nexus integration — replacing discrete, hand-assembled, wire-connected components.
[Shapiro 1:28:48] A photonic integrated circuit for the sensing unit displayed in the showcase.
[Bhaskar 1:31:00] Photonic interconnect performance historically benchmarked in hertz; kilohertz identified as the requirement for sustaining distributed computation.
[Bhaskar 1:31:00] Interconnect above 1 kilohertz announced as met — characterized as a step change rather than an incremental advance.
[Bhaskar 1:31:00] Method described as heterogeneous architecture: trapped-ion compute device connected to a Lightsynq quantum memory optimized for networking.
[Bhaskar 1:31:00] Framed as the first successful combination of two different quantum system types yielding improved performance; analogy to CPUs, GPUs and network cards.
[Bhaskar 1:31:00] DARPA HARQ identified as the program this milestone directly addresses; applicability claimed across trapped ions, trapped atoms and superconductors.
[Ballance 1:08:42] Core compute roadmap stated to scale beyond millions of qubits without photonic interconnects on a five-year view; interconnects positioned as valuable to customers building fleets and networks rather than a single large machine.
[ARG] There is a real and useful tension between Bhaskar's framing and Ballance's. Bhaskar presents the interconnect as the unlock for data-center-scale quantum computing; Ballance says the core compute roadmap does not need it. Both can be true — the interconnect is a product for other people's architectures and a hedge for IonQ's own — but a reader should notice that IonQ has decoupled its scaling story from its networking story. That decoupling is strategically prudent and reduces single-path risk.
VII. Quantum Networking
[Shapiro 36:47] Prior milestone restated: first company to interconnect two quantum computers in an enterprise setting, with the Air Force Research Laboratory.
[Shapiro 36:47] AFRL provided with a quantum network for entanglement distribution supporting experimentation across different quantum computer modalities and devices.
[Shapiro 36:47] Next step stated: moving the fiber-based network into free space using IonQ optical communication technology, unlocking airborne and, in the future, space-based quantum connections.
[de Masi 4:40] Data-center-scale quantum computers named as the multi-year opportunity, delivered through the photonic unit and the networking and quantum memory business.
[Bhaskar 1:31:00] Large-scale quantum systems framed as a network of modular interconnected systems rather than a mainframe.
[Long 2:02:32] EPB describes the first nationally available quantum network alongside the first commercially available quantum computer, on a full fiber network in Chattanooga.
[Long 2:08:09] EPB's decade-old QKD work over a 21-kilometer span with Oak Ridge, Los Alamos and Qubitekk described as the origin of the current network.
[Chad 1:42:36] EPB described as having started in networking and expanded into computing, now expanding into quantum memories in Chattanooga.
[Singh 47:01] AFRL relationship described as bidirectional learning — IonQ built for them and learned from them how to network machines adjacently and at distance.
[de Masi 2:45:40] Multimodal networking claimed as a long-standing leadership position, attributed to Monroe and Bhaskar.
VIII. Quantum Security and Cryptography
[Shapiro 36:47] Q-Day requirement announced at 19,397 physical qubits to threaten critical infrastructure encryption, attributed to hardware and software advances working together.
[de Masi 4:40] Restated as roughly twenty thousand physical qubits to recover a 256-bit elliptic-curve key; described as done to a historic level of detail, without approximation.
[de Masi 4:40] Work described as done responsibly, in partnership with the US government.
[Cardillo 2:43:39] Government pre-briefing confirmed from the board side: senior-most levels engaged, led substantially by Muller, on the explicit grounds of not surprising government partners and behaving responsibly on the safe-cracking side.
[Shapiro 36:47] Second converging trend: government migration deadlines pulled in, with the June US executive order moving the US timeline from 2035 to 2030.
[Shapiro 36:47] Harvest-now-decrypt-later framed as present-tense exposure, not future risk.
[Shapiro 36:47] Full-stack security platform described: QSPM for visibility into quantum-vulnerable infrastructure; a CNSA 2.0 and NIST-compliant remediation path via PQC; QKD as a physical network-layer solution; key orchestration and architecture evolution over time.
[Shapiro 36:47] Congruity360 agreement announced for US enterprise rollout, described as the largest deal of its kind in the United States to IonQ's knowledge, spanning enterprises, healthcare systems, financial institutions, government organizations and telcos.
[Shapiro 36:47] Singtel and SK Broadband cited as existing deployments; claim made to be both the most robust and the most deployed commercial quantum security platform.
[de Masi 4:40] Safe-maker and safe-breaker framing; defense in depth across hardware and software; claim to the largest security practice in the sector.
[de Masi 4:40] Classical AI named as an accelerant of cryptographic risk, with reference to large-language-model progress against recommended post-quantum algorithms.
[Singh 47:01] Security described as originating in a responsibility argument — a company building encryption-breaking machines must also supply the defense.
[Lokada 2:01:46] ServiceNow describes security and optimization converging into secure optimization, with QKD as the notable opportunity; Armis and Veza acquisitions cited.
[Lokada 2:10:25] Argument that the industry must look beyond PQC, which is characterized as largely classical.
[UNDISC] The claim regarding large language models making progress against recommended post-quantum algorithms is the most consequential unverified statement of the day. It was made in passing, without a citation, and it materially strengthens the commercial case for the security business. It must be independently verified before it appears in any published work from this series.
IX. Quantum Sensing, PNT and Timing
[Shapiro 36:47] Portfolio described: quantum optical clocks stated to be the world's most accurate commercial clocks, plus time-transfer devices, atomic gravimeters and gyroscopes.
[Shapiro 36:47] Clocks claimed at 1,000× the accuracy of the best classical equivalent and leading NIST's clock ensemble in this form factor.
[Shapiro 36:47] Deployment environments named: at sea, in space, on the X-37B space plane, on land and in the air.
[Shapiro 36:47] $58 million agreement cited for the DARPA It's About Time program.
[de Masi 4:40] About Time named alongside HARQ and QBI as the three DARPA programs IonQ participates in.
[Singh 47:01] Sensing rationale framed around jam resistance, with reference to GPS jamming and aircraft being walked off course; noted as a national-security rather than general-enterprise product.
[Singh 47:01] A Vector Atomic clock on a space platform cited as a worked example of a cross-portfolio solution rather than a product sale.
[Shapiro 1:25:01] Sovereign demand described as spanning advanced PNT for GPS-denied environments through to computing.
[Shapiro 1:25:01] Oil and gas cited as using PNT for well discovery alongside quantum chemistry and security.
[Bhaskar 1:27:36] Sensing named as the first beneficiary of Nexus photonic integration, moving clocks and inertial sensors from discrete assemblies onto chips.
[de Masi 2:45:40] Atomic clocks named explicitly as a merchant-supply product sold to leading players in the friendly quantum ecosystem.
X. Space, Remote Sensing and Earth Intelligence
[Shapiro 36:47] Disclosed for the first time: quantum computers used with IonQ's own remote-sensing organization.
[Shapiro 36:47] IonQ described as capturing synthetic aperture radar data globally on a daily basis.
[Shapiro 36:47] First hybrid quantum-classical SAR workflow to IonQ's knowledge, with the quantum computer in the processing loop.
[Shapiro 36:47] Worked example shown: a structural change at Miramar Air Base, with the IonQ quantum model closest to ground truth and free of the noise present in classical model outputs.
[Shapiro 36:47] Claim to be the only quantum computing company with this level of SAR data access for running quantum models.
[Shapiro 36:47] Free-space optical operation named as the path to airborne and future space-based quantum connections.
[Shapiro 36:47] NRO contract cited on the space side alongside DARPA validations.
[Singh 47:01] Using quantum computing to improve SAR satellite imaging cited as a worked cross-portfolio solution.
[Cardillo 2:39:32] Board-level endorsement of the SAR result, with SAR identified as central to the speaker's prior government role; framed as a coherence-from-chaos service — synthesis of signal from overwhelming data volume.
[Singh 47:01] Imaging from space using AI named in the opening as one of three customer-facing solution categories, alongside security and computing.
[UNDISC] No subsidiary was named in connection with the SAR platform on stage. The capability is described as IonQ's own. Readers of this series will recognize the underlying asset, but this document does not assert the linkage because the event did not, and the distinction between what is stated and what is inferred is load-bearing here.
XI. Government, Defense and National Security
[de Masi 4:40] Three DARPA programs named: HARQ, About Time, and QBI.
[Shapiro 36:47] US government customers described as validating across the platform: DARPA QBI for computing, sensing awards including It's About Time, networks, and an NRO space contract.
[Shapiro 36:47] June US executive order cited as pulling the federal PQC migration timeline from 2035 to 2030.
[Promotional video 3:08] Recent US executive orders framed as establishing the need for a competitive quantum ecosystem; SkyWater and IonQ positioned as the domestic manufacturing answer.
[Sonderman 31:11] Quantum characterized as sovereign infrastructure; foundry security framed in national-security terms including classified work.
[Singh 47:01] Quantum foundry and advanced technologies segment described as including classified government work for cleared personnel.
[Singh 47:01] Adversary framed as nation-states rather than competing companies.
[Raymond 2:30:47] Board member background: approximately 35.5 years in the Air Force, transition to Space Force in 2019; founder of the Space Force; described as the only four-star to hold that rank in two services since 1947.
[Cardillo 2:33:00] Board member background: nearly four decades in the US intelligence community across seven presidents, ending as director of the National Geospatial-Intelligence Agency; a director since 2024 and Executive Chairman of IonQ Federal.
[Cardillo 2:33:00] Strategic-failure framing: surprise is not the failure to see what is coming, but the failure to prepare for what is already visible.
[Muller 2:35:04] Background: quantum program leadership at Sandia National Laboratories for roughly 21 years, then director of IARPA, joining IonQ approximately 12 to 18 months ago.
[Muller 2:35:04] Central government-market argument: the government does not want a qubit, it wants mission solutions to its hardest problems; technology alone is insufficient without mission understanding.
[Muller 2:41:49] Sovereignty and trust framing: ion-based systems enable locally or remotely deployed solutions that belong to the customer, rather than anonymous hosted quantum service.
[Raymond 2:44:40] Translation of research capability into warfighter capability named as nationally critical and as IonQ's particular strength.
[Cardillo 2:39:32] Executive order characterized as demanding a scaled ecosystem, with the past twelve months of team additions framed as the response.
[de Masi 2:45:40] Merchant supply positioned across the US, the Five Eyes and the allied world.
XII. Sovereign and International
[Butti 2:21:37] Italy's national quantum strategy, adopted under the speaker's direction and issued the prior July, concurrently with the European strategy; scope explicitly beyond computing to communication and security, sensing, research, skills and industrial ecosystem.
[Butti 2:21:37] Italy's National Quantum Forum described as convening government, researchers, universities and industry.
[Butti 2:21:37] Stated ambition for Italy to become a leading European center for quantum technologies, and explicitly to help create technologies rather than only use those developed elsewhere.
[Butti 2:21:37] IonQ Italia welcomed by name, led by Marco Pistoia, framed as investing in Italian talent, research and universities.
[Butti 2:21:37] Q alliance cited, with IonQ as a founding member, connecting companies, universities and research institutions and providing researchers access to advanced quantum compute systems.
[Butti 2:21:37] Quantum renaissance framing described as a shared concept, first discussed with de Masi in Rome the prior year.
[Butti 2:21:37] Stated Italian preference for an open, competitive and technology-neutral ecosystem in line with the European Union.
[Butti 2:21:37] Forward date: the Como Lake Digital Innovation Forum at Cernobbio the following month, referencing an MOU signed there the prior year.
[de Masi 2:27:33] IonQ Italia and Lutech partnership announced for Italian go-to-market across all IonQ quantum platforms; release stated to follow the next morning before market open.
[Shapiro 1:25:01] Sovereign demand described as global and cross-industry, spanning PNT through computing.
[de Masi 4:40] Cloud presence since 2020 across Amazon, Google and Microsoft; claim to be the largest cloud partner in the quantum space; global customer map presented.
[Chad 1:42:36] Customer choice framed on two axes: language and stack, and deployment location — cloud, on-premises or sovereign.
[Ho 1:40:01] Silicon Heartland framing: a Midwest quantum manufacturing ecosystem described as resilient for America, with Qolab design, packaging and cryogenic testing in Madison, Wisconsin.
XIII. Life Sciences, Healthcare and Pharmaceuticals
[de Masi 4:40] Life sciences singled out as near and dear to the company, spanning oncology, predictive medicine and personalized medicine; modeling of quantum-mechanical systems named as the modality's superpower.
[Chad 1:42:36] AstraZeneca work continuing, described as three iterations to date.
[Chad 1:42:36] Protein folding progressed from 12 to 14 amino acids on Forte Enterprise and Tempo, then to 16 in the August Synopsys update.
[Chad 1:42:36] Scale framing: oxytocin at nine amino acids; GLP-1s at roughly 30 to 50.
[Chad 1:42:36] Cytochrome P450 identified as not exactly solvable classically today, with $2–3 billion per year of associated research.
[Chad 1:42:36] P450 resource estimate against a 2025 Google/Caltech paper: beyond a Superion 20K; achievable with a 100K, or five 20K systems in parallel.
[Chad 1:42:36] Economic framing: pulling the workload forward roughly two years, worth hundreds of millions in net present value; networking described as not a prerequisite but valuable.
[Chad 1:42:36] CCRM healthcare work; quantum pattern-finding on fMRI data for early-onset brain cancer and dementia.
[Rabinowitz 2:03:45] Protein-folding intractability framed concretely: 100 amino acids with four permutations between each yields four-to-the-hundredth hypotheses, citing an IonQ paper's assumption.
[Rabinowitz 2:03:45] Argument that nature reaches the global minimum through quantum tunneling and thermal motion, where classical AI systems including AlphaFold get trapped in local minima.
[Rabinowitz 2:03:45] Named healthcare applications: protein binding, neoantigen prediction for personalized cancer vaccines, therapy response prediction, prognostics.
[Rabinowitz 2:12:54] Natera oncology data described: recurrence detection ahead of clinical symptoms, therapy response measurement outperforming imaging, adjuvant clearance assessment and chemotherapy benefit prediction, via single tumor molecules in blood.
[Rabinowitz 2:12:54] Neoantigen selection framed as the target problem, with under one percent of tumor mutations making good neoantigens.
[Rabinowitz 2:12:54] MyOme collaboration with IonQ on LLM fine-tuning; full LLM training on quantum architecture described as still ahead.
[Rabinowitz 2:12:54] Result: emulating a May 2026 IonQ paper on sentiment fine-tuning, an approximately 800-length embedding with a 6-qubit then 14-qubit quantum fine-tuning head improved AUC from 0.70 to 0.83 — explicitly a hardware simulation.
[Rabinowitz 2:12:54] MyOme's whole-genome plus EMR modeling described as potentially saving over $200 billion to the US healthcare system through earlier disease detection.
[Cardillo 2:33:00] Drug discovery named among the mission outcomes that drew the speaker to the company.
[ARG] This is the deepest life-sciences disclosure IonQ has made at any public event, and it lands directly on the two threads this series has been building in the pharmaceutical dossier program. The P450 resource estimate in particular converts an open technical question into a stated machine-class requirement. Part V treats this as the single highest-value new input for the dossier work.
XIV. Energy and Utilities
[Long 2:02:32] EPB's IonQ Forte Enterprise system expected to be fully commissioned around the current quarter.
[Long 2:02:32] Focus areas named: energy grid optimization and energy security, using substantially the full stack of IonQ services.
[Long 2:02:32] EPB writing and running its own algorithms and programs for industry energy needs, with eight grant-funded fellows trained on the IonQ platform.
[Long 2:08:09] Chattanooga ecosystem strategy described: IonQ office and R&D facility, Vanderbilt satellite office, UT Chattanooga partnerships, supportive city, county and state government.
[Long 2:08:09] Computer to be made available physically in the quantum center and via cloud.
[Chad 1:42:36] EPB path described as networking first, then computing, now expanding into quantum memories.
[de Masi 2:07:02] EPB characterized as among IonQ's largest, closest and most historic partners, running its own applications and algorithms.
[Shapiro 1:25:01] Oil and gas engagement spanning quantum chemistry, PNT for well discovery, and critical-infrastructure security.
[Chad 1:42:36] An oil-and-gas specialist hired who met two of the largest US energy companies in month one, both of which already had quantum science teams.
XV. Enterprise Software, Financial Services and AI Convergence
[Lokada 2:01:46] ServiceNow evaluating two lanes — security and optimization — converging on secure optimization; quantum plus AI framed as the strategic direction.
[Lokada 2:10:25] Hybrid work reframed as work spanning classical and quantum systems; explicit acknowledgment that quantum will not help every problem and that multi-day runtimes can be acceptable for the right strategic problem.
[Lokada 2:10:25] Several billion agents anticipated online over coming years, raising an orchestration problem; probabilistic intelligence leveraged for deterministic workflows.
[Lokada 2:19:00] Operational resiliency named as the missing frame in the industry conversation, positioned above compliance.
[Shapiro 1:25:01] Financial services engagement described as dual-track: quantum computing applications with one team, quantum security with the CISO.
[Singh 47:01] Quantum positioned as what comes after AI, and as already working with AI in hybrid solutions today.
[Singh 47:01] Hybrid deployments described as an AI factory with a GPU next to the quantum computer.
[Chad 1:42:36] Quantum algorithms improving LLM behavior: 24 percent better accuracy with lower time to solution and better energy efficiency.
[Chad 1:42:36] NVIDIA and Oak Ridge paper on LLMs improving quantum algorithms; GPU and QPU clusters working together.
[Rabinowitz 2:03:45] Open question posed on whether AI is limited by data or by architecture and global-minimum search — framed as the reason quantum matters to AI.
XVI. Industrials, Logistics, Oil and Gas
[de Masi 4:40] Application roadmap unlocks named: computational engineering, logistics, image change detection and pharmaceuticals.
[Chad 1:42:36] Einride logistics work cited, with the methodology carried into healthcare via CCRM.
[Chad 1:42:36] Synopsys work updated in August.
[Shapiro 1:25:01] Oil and gas described across three product lines simultaneously.
[de Masi 4:40] Materials science and chemistry named alongside life sciences as sectors with major implications from the Shor's-scale work.
XVII. Go-to-Market, Customers and Commercial Motion
[de Masi 4:40] On-premise system count, cloud computing hours delivered and global customer count presented on a map; customer satisfaction described as now part of the company lexicon alongside manufacturability, upgradability and accessibility.
[de Masi 4:40] Five-plus years in the cloud and five-plus years shipping systems; machines first operating in 2017; applications in production today.
[Ballance 18:36] Not an hour passes without a job running on the IonQ fleet.
[Singh 47:01] Land-and-expand strategy stated explicitly, with security increasingly the entry point rather than computing.
[Singh 47:01] One-stop-shop framing; the Cisco of quantum offered as the preferred analogy over the Amazon of quantum.
[Singh 1:28:56] Competitive dynamics characterized: by the time IonQ reaches the customer, modality is usually already chosen, so the conversation is value and use cases rather than price.
[Singh 1:28:56] Forward-deployed engineers and application developers used to move customers from the cost side to the revenue side of their equation.
[Singh 1:28:56] Meeting the customer where they are: QCaaS offered where a full system purchase is not wanted.
[Chad 1:42:36] Customers rarely ask about two-qubit gate fidelity or coherence time; they ask time to solution and economics to solution.
[Chad 1:42:36] Sniper-versus-shotgun positioning; results published to a public repository with third-party validation.
[Chad 1:42:36] Customer-engagement team more than doubled in the first half to about 70 people worldwide, almost entirely PhDs, across forward-deployed engineering, customer solutions and applications R&D.
[Chad 1:42:36] Upgrade pattern reported: customers who bought Aria and Forte bought Tempo and want Superion cloud access.
[Chad 1:42:36] Iterate-and-scale phase framing across both hardware and go-to-market; building the machine that makes the machine described as organizational rather than technical.
[Shapiro 1:25:01] Customers approaching IonQ for integrated multi-technology solutions rather than single products, globally and across industries.
[de Masi 2:45:40] Federal team described as as much a part of go-to-market as the enterprise organization.
[INFER] The upgrade pattern — Aria and Forte buyers taking Tempo and asking for Superion cloud access — is the closest thing offered on stage to a net-revenue-retention argument. It was stated qualitatively and without numbers, but it is the mechanism a platform thesis requires, and it should be pressed for quantification at the next earnings call.
XVIII. Merchant Supply and the Industry-Enablement Business
[Singh 47:01] Direct statement, described as rarely discussed: IonQ sells components to other quantum computing companies that already depend on it, and manufactures some of what they need in the SkyWater quantum foundry.
[Singh 47:01] Explicit statement of intent to continue; other foundry participants framed as industry and ecosystem rather than competitors, and in some cases as customers.
[Singh 47:01] IonQ's accelerating roadmap and learning curve framed as helping other players.
[Sonderman 31:11] Nine quantum foundry customers; framing that SkyWater is creating a quantum foundry industry rather than only serving IonQ.
[Sonderman 1:13:58] IP compartmentalization commitment given directly to the room, including non-IonQ customers in the audience.
[Bhaskar 1:31:00] Foundry platforms explicitly built as an open ecosystem across atom, ion, superconducting, photonic and spin-based modalities.
[Martinis 1:37:17] Competitor-modality endorsement: IonQ's acquisition of SkyWater described as helping SkyWater focus and build new capabilities to properly scale quantum computers.
[Ho 1:40:01] Qolab as launch SC250 customer; SkyWater credited with providing predictable unit economics necessary for investors to underwrite manufacturing at scale.
[de Masi 4:40] Ecosystem expansion framed as the objective, with access, price, power and space targets set so that every company and government can invest.
[de Masi 2:45:40] Merchant supply business described as grown decisively this year, with atomic clocks named as a supplied product, across the US, Five Eyes and allied world.
[ARG] This series has previously treated merchant supply as a feature of the semiconductor ecosystem rather than an ecosystem in its own right. On the evidence of this event, that is no longer the right treatment. Two thirds of SkyWater's quantum wafer volume is external, nine foundry customers exist, a Nobel-laureate-founded competitor appeared on stage to endorse the arrangement, and the CEO closed the day by naming merchant supply as a growth business. Part III proposes it as a twelfth ecosystem.
XIX. Talent, Organization and Leadership
[Singh 47:01] CFO joined the board when the company was at roughly $2 million in revenue with one machine and thirty people; moved from board into the company one year prior, one week before the first analyst day.
[Ballance 18:36] Big operational changes made to company structure, technology development and team organization to compress innovation cycles from years to months.
[Chad 1:42:36] Organizational scaling framed as building the machine that makes the machine — people and process, not hardware.
[Muller 2:35:04] Team-building argument: combining best-in-class science with people who understand mission requirements is what produces solutions, and is characterized as a solid business model.
[Raymond 2:37:39] Team named as the reason for joining; integrity and mission focus compared to military service; Space Force stand-up cited as the analogous integration challenge.
[Cardillo 2:39:32] Twelve months of team additions characterized as having scaled the company rather than merely complemented it.
[de Masi 4:40] Board members present in the room and acknowledged; founder present in the audience.
[de Masi 2:37:13] Muller identified as the only one of the four panelists holding a PhD in ion trapping.
[Bhaskar 1:31:00] New role working directly with Sonderman at SkyWater on the quantum foundry platforms.
[de Masi 2:45:40] Board described as supportive of an aggressive technical and commercial agenda, with more expected over the coming twelve months.
XX. Academic, Workforce and Regional Ecosystem
[Bhaskar 1:31:00] Professor John Bowers present and credited as the pioneer of on-chip lasers underpinning today's AI data-center optical interconnects.
[Bhaskar 1:39:29] Argument offered to investors: assess the foundry by where field luminaries choose to build — naming Bowers and Martinis, both at SkyWater.
[Ho 1:40:01] Midwest quantum manufacturing ecosystem framed as resilient for America; Qolab design, packaging and cryogenic testing in Madison, Wisconsin.
[Long 2:08:09] Chattanooga ecosystem: IonQ R&D office, Vanderbilt satellite office with 250 faculty or staff over three to five years, UT Chattanooga partnerships, state and local government backing.
[Long 2:02:32] Eight EPB fellows trained via the IonQ training platform, writing their own algorithms.
[Butti 2:21:37] Italian strategy explicitly includes research and skills; Q alliance provides researchers with access to advanced quantum compute systems.
[Butti 2:21:37] Explicit statement that technology alone is insufficient — investment in people, skills and trust is required.
[Chad 1:42:36] IEEE Quantum Week outcome: 857 papers submitted, 20 selected as best papers; IonQ submitted 13, 10 entered, four of the ten won.
[de Masi 4:40] Distinguished university professors present; hope expressed that third parties will build businesses and research on IonQ machines at arm's length.
XXI. Competitive Positioning Claims
Claims in this category are positioning statements made by IonQ or its partners on stage. They are indexed as claims, not endorsed as findings.
[de Masi 4:40] Largest quantum company in history; first vertically integrated one.
[de Masi 4:40] Roughly 2× the Q2 revenue of the rest of the public quantum sector combined.
[de Masi 4:40] Generally the biggest cloud partner in the quantum space.
[de Masi 4:40] First company to publish a complete fault-tolerant, shovel-ready blueprint.
[Sonderman 31:11] SkyWater is the only quantum foundry; a capability that does not exist anywhere else in the world.
[Singh 47:01] The only trusted fab of scale in the United States; the only quantum fab in the world; competitors targeting 2029.
[Singh 1:28:56] No other company has a 256 chip; nothing known at 10,000.
[Shapiro 36:47] Most robust and most deployed commercial quantum security platform.
[Shapiro 36:47] World's most accurate commercial clocks; leads NIST's clock ensemble in this form factor.
[Shapiro 36:47] Only quantum computing company with this level of SAR data access.
[Shapiro 36:47] First hybrid quantum-classical SAR workflow, to IonQ's knowledge.
[Bhaskar 1:31:00] World's only quantum platforms for semiconductor manufacturing; industry's first dedicated quantum foundry platforms.
[Bhaskar 1:31:00] First successful combination of two different quantum system types yielding improved performance.
[Ballance 18:36] Most detailed full-stack quantum computing architecture published; first full end-to-end compilation of Shor's algorithm.
[Long 2:02:32] First nationally available quantum network and first commercially available quantum computer, per EPB.
[Sonderman 1:13:58] GlobalFoundries characterized as a scaled specialty foundry rather than an innovation engine.
[Sonderman 1:11:46] Explicit rejection of the claim that advanced nodes are required for quantum fabrication.
[ARG] The competitive claims made this year are notably narrower and more checkable than a year ago. Almost every one is now framed around manufacturing or deployment rather than around benchmark performance. That is a defensible shift — manufacturing claims are auditable in a way that algorithmic-qubit claims never were — but it also means the claims can be tested against reality faster, and this series should test them.
XXII. Forward-Looking Statements and Dated Commitments
Every statement on stage that carries a date, a number, or an explicit future obligation. These are the scoreable items and should be carried into the Delivery Ledger.
Commitment | Date or condition stated | Speaker / cite |
First Superion systems begin customer-site deployments | Early 2027 | Ballance 18:36 |
FY2026 consolidated revenue | $450–460 million | Singh 47:01 |
IonQ standalone FY2026 revenue | $280–290 million | Singh 47:01 |
EPB Forte Enterprise fully commissioned | Approximately the current quarter | Long 2:02:32 |
EBITDA color | At the next earnings report, post-audit | Singh 47:01 |
IonQ Italia / Lutech release | Next morning, before market open | de Masi 2:27:33 |
Superion generational ladder | 256 → 10K → 20K → 200K, sequence stated without individual dates | de Masi 4:40 |
Core compute roadmap beyond millions of qubits without photonic interconnect | Five-year view | Ballance 1:08:42 |
Free-space, airborne and eventually space-based quantum connections | Next step; no date given | Shapiro 36:47 |
Vanderbilt Chattanooga satellite office | Three to five years, 250 faculty or staff | Long 2:08:09 |
Como Lake Digital Innovation Forum, Cernobbio | Next month | Butti 2:21:37 |
World Quantum Day at the NYSE | April 14, 2027 | de Masi 2:45:40 |
Next Analyst Day | One year from the event date | de Masi 2:45:40 |
Continued investment across every named category | Ongoing, with fail-fast discipline | Singh 47:01 |
[FACT] The early-2027 Superion deployment date is the most consequential new dated commitment created today, and it is the first one in the Ledger's history that is gated on manufacturing yield rather than on physics. That is a different risk profile and should be tracked as such.
XXIII. Analyst Q&A Index, and Transcription Uncertainty Flags
XXIII.a — Analyst Q&A by firm
Firm / analyst | Question | Answered by |
Rosenblatt Securities — John McPeak | At what physical qubit count does photonic quantum interconnect become important in the roadmap? | Ballance — not needed on the core compute roadmap on a five-year view; valuable to fleet and network customers |
Northland — Nihal Choksi | How was the cycle-time reduction from eight months to two months achieved, and at what lithography node? | Ballance on foundry culture and supply-chain hustle; Sonderman on 90nm / 200mm; Singh on rejected alternative foundries |
StoneX — Gary Mobley | How do the other eight quantum foundry customers view conflict of interest, and how much lead time exists versus IBM and GlobalFoundries QTS? | Sonderman — IP compartmentalization; rising-tide framing; competitive commentary on Anduril/HRL, IBM's 2015 divestiture and GlobalFoundries' standardization model |
Unidentified | Is there a scale-out roadmap for photonic interconnect phases three and four, and are the manufacturing capabilities in place? | Sonderman — Minnesota and Florida capacity sufficient; five generations in parallel; time-to-market framing |
Needham — Quinn Bolton | When will there be enough data on median gate fidelity across the 256 QPU, how confident is the fidelity target, what about crosstalk, and what are the plans for Fab 25? | Ballance on short loops and integration confidence; Sonderman and Singh on Fab 25 as ecosystem and optionality |
B. Riley — Craig Ellis | What technology competencies are needed for tiling at end-of-decade scale, and how are customers engaging across the three technology areas? | Ballance — integration not physics, tolerances comfortably within industry standard; Shapiro — integrated-solution demand, global and cross-industry |
Mizuho — Vijay | How does Nexus Photonics fit the IonQ and SkyWater roadmap; what is 256 pricing; how should fiscal 2027 be modeled between core IonQ and SkyWater? | Bhaskar on Nexus; Singh declined pricing and did not address FY2027 |
[UNDISC] The Mizuho question contained three parts and received two answers. FY2027 modeling was not addressed by any speaker. Whether this was an oversight in a compressed Q&A or a deliberate non-answer is not determinable from the transcript, but the result is the same: no FY2027 framework exists in the public record as of today.
XXIII.b — Names and figures, resolved against the published releases
The proceedings were captured from audio, and a number of proper nouns and figures were rendered imprecisely in that capture. Every one of them has been resolved against IonQ’s own published materials from the same day. Where the two disagree, the releases govern, and the corrections below have been applied throughout this document.
As captured from audio | Correct form | Basis |
"Indra Singh" | Inder Singh | IonQ’s Chief Financial and Operating Officer, named in the company’s own materials |
"Meher Bhaskar" | Mihir Bhaskar | Named in the SkyWater Quantum Solutions release as the leader of the merchant foundry business; co-founder of Lightsynq |
"QLab" | Qolab | The superconducting quantum computing company co-founded by John Martinis, based in Madison, Wisconsin |
"SD250" | SC250 | SkyWater’s superconducting platform; the photonics platform is SP90 |
"Lightsync" | Lightsynq | The acquired quantum-memory business behind the kilohertz interconnect |
Cycle time "eight months to two" | Nine months to two | The eight-month figure was supplied by an analyst from the floor; IonQ’s own release states nine to two, alongside six tapeouts in the first half of 2026 and twelve times more wafer lots over six months than at its previous foundry |
"$58 million" for It’s About Time | Carried as stated, pending confirmation | Prior series records carry a $28 million DARPA atomic-clock production contract. This is the one figure in the proceedings not resolved by the day’s releases, and it is the only item in this index still open |
"19,397 qubits" | Confirmed | Stated in the published resource estimate against secp256k1, alongside 1,457 logical qubits, approximately 39 million logical Toffoli gates, and roughly 25.7 days per attempt |
"Congruity360" | Confirmed | Named in the release, with the agreement valued at $8.18 million covering Clavis QKD devices and Solteris network appliances |
"Superion", "Walking Cat" | Confirmed | Both used consistently across the releases and prior series work |
[FACT] One item in this index remains open: the value of the DARPA It’s About Time agreement, where the audio and this series’ prior records disagree. Everything else has been resolved against primary company materials. No claim in the body of this report rests on the unresolved item.
Prepared September 8, 2026 for the Quantum Technology Integration Series. The author holds a long position in IonQ. Not investment advice.
No comments:
Post a Comment