Quantum Tech Integration : The Quantum Frontier Report: The Race to Lead, the Compounding Advantage, and the Investor Imperative

QUANTUM TECHNOLOGY INTEGRATION SERIES

The Quantum Frontier Report

The Race to Lead, the Compounding Advantage, and the Investor Imperative

The next generation of quantum systems is arriving now. When they do, everything changes. The companies that demonstrate a credible, integrated path to fault tolerance — combining hardware progress with complete platform infrastructure — will set the pace for the quantum industry and for the critical sectors of the broader technology economy that depend on it. This report determines who is best positioned to lead that era. The window to act is measured in quarters, not years.

This report is independent analysis prepared for an institutional and technical readership. It is not investment advice, and it is not a recommendation to buy or sell any security.

Conflict-of-interest disclosure: the author holds disclosed long equity positions spanning IonQ, IBM, Infleqtion, Horizon Quantum Holdings, D-Wave, Amazon, and Microsoft. These positions predate this research. They span a leading trapped-ion player, the superconducting incumbent, the annealing/gate-model dual-platform vendor, a neutral-atom sensing-and-computing company, and two of the three hyperscaler cloud layers. The breadth is deliberate context: the holdings represent exposure to the structure of the industry rather than a directional bet on any single thesis. The methodology below is built to let evidence, not holdings, drive every ranking — through within-tier scoring, disclosure-candor normalization, an independent-benchmark mandate, third-party framework anchoring, and a standing rule that every time-sensitive fact is re-verified against primary sources.

Source discipline: peer-reviewed results (Nature, Physical Review Letters/X) are weighted above regulatory filings (SEC registration statements, which carry legal accountability), which are weighted above company roadmaps and white papers (which are marketing artifacts until independently demonstrated), which are weighted above trade press. Throughout, claims that have cleared independent peer review are distinguished from vendor and preprint claims that have not. A company performing well on a credible benchmark is treated as the benchmark working, not as evidence of bias; skepticism is reserved for genuine gaps in verification and applied symmetrically.

Contents

How to Use This Report 4

Why This Report Matters 6

Executive Summary 8

The Next Generation: What Is Coming and What It Unlocks 11

The Logical-Qubit Inflection 15

Methodology & Scoring Framework 17

Master Comparison: The Twelve at a Glance 21

Company Profiles 22

Trapped-Ion: IonQ, Quantinuum 22

Superconducting: IBM, Google, Rigetti, D-Wave 28

Neutral-Atom: QuEra, Atom Computing, Pasqal, Infleqtion 33

Photonic: PsiQuantum, Xanadu 36

Context Players 39

Cross-Cutting Comparison 41

A Tale of Two Trapped-Ion QEC Strategies 46

Vertical & Ecosystem Impact 48

Sovereign & Geopolitical Axis 51

Cryptography & the Crypto-Break Gap 52

Competitive Leadership 54

The Compounding Advantage: Why the Distance Widens 58

Weighted Scorecard: All Twelve Companies 62

Sensitivity Analysis: Rankings Across Weight Profiles 64

Decision Framework: Translating Evidence into Action 65

What These Advances Mean for the World 68

The Verdict: Who Leads, Why It Matters, and What Comes Next 72

Conclusion: The Series Verdict on IonQ's Position 76

Risk Factors & Methodological Notes 78

Selected Sources & Evidence Base 78

How to Use This Report

This report is dense by design. The evidence base spans peer-reviewed physics, SEC filings, and primary technical papers — and the reader deserves to know which is which. The timed reading paths below let you extract what you need without reading all seventy-six pages. The role-based guide that follows maps reader type to the most decision-relevant sections.

If You Have…

Five Minutes. Read the Executive Summary (p.8) and the Weighted Scorecard (p.62). The scorecard shows every company’s seven-axis scores and blended ratings at a glance. The executive summary states the three structural findings and where the field stands. If you leave with one number: IonQ 9.2 near-term, IBM 7.9, gap of 1.2 points on a methodology that has been stress-tested against three prior reports in this series.

Fifteen Minutes. Add the Competitive Leadership section (p.54), the Sensitivity Analysis (p.64), and the IonQ and Infleqtion company profiles (p.20 and p.33). Competitive Leadership explains where IonQ’s lead is structural versus contested, and where IBM, Google, and Quantinuum hold genuine advantages. The sensitivity analysis shows whether the rankings hold under different weight assumptions — they do. The IonQ and Infleqtion profiles are the two companies where the scoring most diverges from market consensus and where the evidence base is most specific.

Thirty Minutes. Add the full company profiles section (p.15–25), the QEC comparison table (p.41), and A Tale of Two Trapped-Ion QEC Strategies (p.46). The profiles give you the Watch For line and change-of-verdict trigger for each company — the single most actionable content in the report for anyone managing positions or procurement timelines. The QEC table is the cleanest cross-vendor comparison of what has actually been demonstrated versus claimed. The trapped-ion comparison explains why IonQ’s qLDPC code efficiency advantage (4.5:1 versus Steane’s 7:1) is analytically significant and why it appears in the Specs score.

One Hour. Add the Methodology & Scoring Framework (p.17), the Master Comparison table (p.21), the Cross-Cutting Comparison (p.41), the Sovereign & Geopolitical Axis (p.51), What These Advances Mean for the World (p.68), and the full Conclusion (p.72). The methodology section is where every weight is defended and every axis defined; a reader who disagrees with any score should start there. The sovereign section is where the report’s geopolitical findings are concentrated and where the USTC Λ=1.40(6) result is placed in context. The civilizational impact section explains why these hardware advances matter beyond the investment and procurement horizons. The conclusion delivers the final verdict on each company.

Full Due Diligence. Read the report cover to cover, including the evidence notes in each company’s addenda, the Watch For and Change of Verdict lines, the Selected Sources section (p.78), and the Risk Factors (p.78). Cross-reference every score against its stated evidence basis. The seven-axis framework is designed to be recomputed: all dimension scores and weights are explicit so you can substitute your own assumptions and see how the rankings move. For primary source verification: IonQ’s qLDPC result is arXiv:2606.06455 (June 4, 2026); the 99.99% fidelity result is arXiv:2510.17286; the Q1 2026 revenue and RPO figures are the SEC 8-K filed May 6, 2026; the SkyWater stockholder approval is the Form 8-K filed May 8, 2026.

By Reader Role

Your role	Start here	Key measures to track	Most decision-relevant section
Institutional investor	Executive Summary (p.8), Competitive Leadership (p.54)	Near-term and long-term weighted scores; revenue and RPO figures; SkyWater close date	Competitive Leadership & Scoring (p.54); Risk Factors (p.78)
Enterprise / procurement	Why This Report Matters (p.5), Master Comparison (p.15)	Distance-to-delivery scores; cloud access column; Watch For lines per vendor	Vertical & Ecosystem Impact (p.48); Decision Framework (p.65)
CTO / technical lead	Methodology (p.11), QEC Comparison Table (p.32)	Evidence tier tags (PR / PS / VA); logical-qubit type (V/ED); code distance and suppression factor	Company Profiles (p.20); Two Trapped-Ion Strategies (p.46)
Policy / government	Sovereign & Geopolitical Axis (p.51), Full-Stack Grid (p.39)	Supply-chain ownership; CHIPS Act positions; sovereign program context; IonQ’s SkyWater foundry (DoD-trusted, US-based, only quantum company with owned foundry pending close); defense program positions (DARPA HARQ, MDA SHIELD IDIQ, SDA HALO $39M); multi-continent QKD network deployments as US sovereignty infrastructure	Sovereign & Geopolitical Axis (p.51); Full-Stack Coverage Grid (p.39)

Why This Report Matters

Everything changes with the next generation of quantum systems. That is not a prediction about a distant future — it is a statement about what is happening right now, in 2026, with the systems being delivered and the platforms being built on these pages. The error-correction threshold has been crossed on multiple independent platforms. Architectural blueprints for fault-tolerant machines have been published with engineering specificity that did not exist eighteen months ago. The first sixth-generation commercial quantum systems have been contracted to paying customers. Foundries are being built specifically for quantum chips. A Chinese state program has crossed the fault-tolerance threshold outside the United States for the first time. These are not signals of an approaching era. They are the opening moves of one that has already begun.

The central question this report exists to answer is: who gets there first, and who integrates into a complete platform when they do? Those are two separate conditions, and both matter. A company that reaches fault-tolerant scale without the networking, security, sensing, software, and cloud infrastructure to make that capability commercially deployable does not lead the quantum commercialization era. A company that builds a broad platform without the hardware to anchor it does not lead it either. The company closest to achieving both — hardware progress toward fault tolerance anchored in a complete commercial platform — captures something far larger than the quantum industry itself. The companies that build and own the quantum platform layer will set the pace for drug discovery, materials science, financial modeling, national security, and cryptographic infrastructure for decades. That is the prize. This report determines who is positioned to claim it.

The urgency is not about the technology being imminent in some abstract sense. It is about the specific commercial dynamics of platform transitions. The customer who buys a 256-qubit system in 2026 is building expertise, integration architecture, and institutional knowledge on that platform before any competitor has shipped a comparable unit. The government that signs a foundry partnership today controls the supply chain when defense applications mature. The investor who understands the compounding dynamic this quarter has a positioning window that will not reopen once mainstream consensus catches up. Platform leaders in technology transitions do not get displaced by better hardware alone — they get displaced by platforms that are simultaneously better in hardware and better in everything built around it. The race to build that complete platform is the race this report analyzes, and it is being decided now.

The central finding of this report is specific: one company has assembled a position across quantum hardware, networking, sensing, security, and sovereign manufacturing that no competitor currently replicates, and the evidence for that position is now strong enough to state it directly rather than hedge it into irrelevance. IonQ leads the seven-axis scoring framework at 9.2 near-term and 9.2 blended — ahead of IBM by 1.2 points, Google by 2.1, and Quantinuum by 2.3. Those gaps are not rounding errors. They reflect five years of public delivery history, the world-record two-qubit gate fidelity at 99.99%, the first peer-reviewed qLDPC breakeven on trapped-ion hardware, a foundry acquisition in final regulatory close that accelerates chip iteration from nine months to two months per cycle, live quantum networking deployments on three continents, and commercial revenue growing at 755% year-on-year from a base of $130 million in 2025. The finding is stated here rather than buried in a scorecard because the reader deserves to know what they are reading toward.

What the static scoring does not capture is the compounding dynamic that makes this report’s timing significant. The 1.2-point lead over IBM today understates the potential competitive distance of 2028–2030 if IonQ’s specific milestone sequence delivers. Each element of the platform compounds the next: a foundry that iterates chips 4.5× faster means more hardware generations per year; each delivered 256-qubit system creates customers with lock-in before any competitor has shipped a comparable unit; each government contract deepens security clearances and program relationships that cannot be replicated by writing a larger check; each networking node deployed makes the next one easier and the alternative path longer for challengers. The mechanism is not linear. At the AQ10,000 threshold — reachable on a networked multi-chip system if the 2028 milestone delivers — several pharmaceutical, materials, and financial applications cross from intractable to tractable simultaneously, and the customer for that capability has already bought the prior generation. That is not an incremental lead. It is a categorical one. This report models that dynamic explicitly, in a dedicated section of the Competitive Leadership analysis.

Stating a leader does not end the analysis; it starts it. The field is genuinely contested in specific dimensions, and the report is explicit about where IonQ does not lead. On demonstrated peer-reviewed logical-qubit counts today, Google and QuEra are ahead. On delivered fault-tolerant algorithm execution, Quantinuum executed first. IBM commands the largest open-source quantum software ecosystem by a substantial margin. Infleqtion is the only public company simultaneously generating revenue from both quantum computing and precision sensing. None of these positions is trivial, and each is documented with primary-source evidence.

The stakes for getting this right have escalated in 2026 faster than most readers can track. A Chinese state program has crossed the fault-tolerance threshold for the first time outside the United States, with a suppression factor of Λ=1.40(6) on a distance-7 surface code. The U.S. government committed $2 billion in CHIPS Act incentives to nine quantum companies in May 2026 and announced plans to take equity stakes in each (the open period for companies to still seek participation is open, and it is feasible that IonQ and SkyWater will seek funding, after regulatory requirements, once the final determination of their merger is completed). SkyWater’s stockholder-approved acquisition by IonQ is pending only regulatory sign-off. Quantinuum is now publicly trading on the Nasdaq. IBM announced America’s first pure-play quantum foundry under a letter of intent. These are structural events that change the competitive map. A reader who has not processed them is operating on a stale picture.

The methodology is built to be checked. Every logical-qubit figure is tagged by whether it is verified error-corrected, error-detected, or merely physical. Every forward claim is labeled. Seven scoring axes with stated weights and per-dimension scores are published so that any reader can recompute under their own assumptions. The source discipline places peer-reviewed physics above regulatory filings, regulatory filings above roadmaps, and roadmaps above trade press, applied symmetrically to every vendor regardless of who benefits.

Executive Summary

The central fact of quantum hardware in 2026 is that the industry has stopped competing on raw qubit count and started competing on error-corrected logical qubits — measured by code distance, by logical error rate per cycle, by how many operations a machine can run before it fails, and — increasingly — by application-level time-to-solution, the metric IonQ's benchmarking framework foregrounds because it captures what actually matters to a customer: how fast a machine reaches a correct answer at a fixed quality threshold, not how many qubits it has. A processor with sixteen hundred physical qubits and a processor running dozens of error-corrected logical qubits are not on the same axis, and conflating them is the most common error in popular coverage of the field. This report is organized around that distinction.

Twelve commercial and emerging organizations anchor the analysis, spanning four hardware modalities — trapped-ion, superconducting, neutral-atom, and photonic. They are not commensurable, so they are never ranked on a single list. Instead they are sorted into two tiers — commercial and emerging — and scored only within tier. Leadership is assessed within the commercial Tier 1 cohort; the gaps cited throughout — IonQ leading IBM by 1.2 points, Google by 2.1, Quantinuum by 2.3 — are within-tier comparisons establishing unambiguous commercial and platform leadership. Scoring uses seven axes whose weights are stated, defended, and published so that any reader can recompute under their own assumptions. Sovereign research programs, including the Chinese and European programs discussed in the geopolitical section, are treated as context rather than cohort, because they operate on fundamentally different terms than companies with customers and shareholders.

On the question every reader asks — who is in the lead — the honest answer is that the leader changes with the time horizon and with the definition of leadership, and the report delivers a verdict for each rather than collapsing them into one misleading number. Measured on what has actually been demonstrated and peer-reviewed today, the contest sits among a small group with genuine error-correction results: Google holds the strongest surface-code suppression (distance-7, Λ=2.14, Nature), QuEra holds the verified logical-qubit count record (96 below threshold, Nature), Quantinuum holds the first fault-tolerant algorithm execution (12 logical qubits, Helios), and IonQ — as of its June 4, 2026 peer-reviewed paper — holds the first qLDPC breakeven on trapped-ion hardware, the most code-efficient path demonstrated to date. Each of these is a different kind of result, which is precisely why a single demonstrated-leadership ranking is misleading: a below-threshold memory result, a logical-qubit count, a fault-tolerant algorithm run, and a code-efficiency breakeven are not the same achievement, and none of them on its own captures application-level time-to-solution — the metric IonQ's framework foregrounds and the one that determines commercial value. Measured on position through 2028, the field remains contested on individual demonstrated-hardware dimensions but is not contested on commercial platform position: IonQ leads the seven-axis scoring framework by 1.2 points over IBM and by a larger margin over every other competitor, under every sensitivity profile. Measured on the decade, leadership turns on which constraint proves binding — manufacturing access, capital, or the physics of a given modality — and the report commits to a primary view while exposing the assumption it rests on.

The second major finding is that hardware capability is no longer the whole contest. A parallel competition is underway across the full stack — networking, security, sensing, space infrastructure, cloud reach, and software — and it is being fought through two distinct strategies: owned vertical integration, exemplified by IonQ's acquisition campaign, and the software-platform-plus-partnership model, exemplified by IBM's Qiskit ecosystem and its networking partnership with Cisco. Neither is strictly dominant, and the report scores full-stack breadth on its own axis precisely so that it does not silently inflate a hardware ranking.

Third, a wave of end-to-end architectural blueprints arrived in 2026 — most prominently IonQ's Walking Cat specification for trapped ions, alongside competing neutral-atom and photonic designs — shifting the conversation from qubit-count roadmaps to concrete engineering specifications grounded in quantum low-density parity-check codes. These blueprints are scored on engineering realism: whether they rest on capabilities already demonstrated or on capabilities still hypothetical.

Finally, the sovereign dimension has become decision-relevant rather than incidental. A Chinese state program has crossed the fault-tolerance threshold with Λ=1.40(6) on a distance-7 surface code — the first non-U.S. result below threshold, though not yet at parity with the leading Western suppression factor of Λ=2.14. European programs are building sovereign capability through national supercomputing integration, and supply-chain chokepoints — cryogenics, lasers, photonic components, and foundry access — now shape competitive position as much as physics does. For an institutional buyer, sovereign risk is a procurement variable, not a footnote.

Figure 11. Two Different Races

Each company plotted by demonstrated peer-reviewed QEC capability (horizontal) against FY2025 revenue (vertical, log scale). The image makes the report's central tension visible in one view: the demonstrated-capability leaders Google and QuEra sit in the high-capability, pre-revenue quadrant, while IonQ occupies the high-capability, high-revenue quadrant with IBM. Capability and commercial position are different races, and a reader who conflates them will misjudge the field. Capability axis is a qualitative analyst placement; revenue figures are from SEC filings and disclosures where available.

IonQ: Differentiators at a Glance

Six structural advantages that distinguish IonQ from every other company in this cohort. Each is anchored to a primary source. The combination is the basis for the report’s platform leadership verdict.

Highest demonstrated gate fidelity. 99.99% two-qubit gate fidelity (world record, no ground-state cooling), on the EQC prototypes underpinning the shipping 256-qubit product line. Primary source: arXiv:2510.17286, October 2025, peer-reviewed.

Peer-reviewed qLDPC breakeven + best-in-class code efficiency. Nine QEC codes on one device without hardware reconfiguration. The [[18,4,3]] qLDPC code (4 logical in 18 physical) achieves a logical error rate the paper states as up to 9× better than the prior superconducting qLDPC demonstration; decomposed per basis, that is ~4.3× better on X and ~8.5× better on Z (X: 2.01% vs ~8.67%; Z: 1.08% vs ~9.15% per cycle). First qLDPC breakeven (logical lifetime 3.95s vs 3.3s physical, overlapping error bars). Code efficiency 4.5:1 physical-to-logical versus Steane code 7:1. Primary source: arXiv:2606.06455, June 4, 2026, peer-reviewed.

Commercial revenue scale — first and only above $100M GAAP. $130M FY2025 revenue (+202% YoY); $64.7M Q1 2026 (+755% YoY, 30% above guidance midpoint); RPOs $470M (+554% YoY); five consecutive quarterly guidance beats. 6th-generation 256-qubit systems contracted to three named customers: University of Chicago ($9.2M Q1 2026 revenue; SEC 10-Q), University of Cambridge (Q1 2026 8-K), Horizon Quantum Holdings (April 2026). Primary source: IonQ Q1 2026 SEC 8-K.

SkyWater foundry acquisition compresses the hardware iteration cycle. Stockholder-approved May 8, 2026; regulatory close pending. Design-to-first-samples: 9 months → 2 months (4.5× faster). 200,000-qubit chip functional testing: 2030 → 2028. No other commercial quantum company has a comparable owned semiconductor foundry position. Primary source: SEC Form 425, January 2026.

Owned vertical integration across five pillars — revenue-generating today. Compute + networking (Switzerland, Romania, Florida deployed; Skyloom optical terminals) + security (ID Quantique, ~300 patents) + sensing (Vector Atomic revenue generating Q1 2026; CFO confirmed) + space (SDA HALO $39M, Capella Space). DARPA HARQ, MDA SHIELD IDIQ, and SDA HALO are signed contracts, not pipeline. Full-stack score 9.5 — highest in the 12-company cohort.

Structural cost advantage: 33× lower BOM vs. superconducting at scale. Third-party validated bill-of-materials for a 2M-qubit-scale system: <$30M versus >$1B for dilution-refrigerator superconducting systems. No cryogenics, no helium-3, no specialist facility required. EQC integrates qubit control on standard semiconductor chips. Primary source: Q3 2025 IonQ SEC 8-K.

The Next Generation: What’s Coming and What It Unlocks

The twelve companies on these pages are not building research instruments. They are building commercial systems, and the next generation of those systems is closer than most readers understand. The table below consolidates every major upcoming system across the cohort — the analyst’s forward-looking map, with evidence tiers applied. Vendor-stated timelines are labeled as such; SEC-filed commitments carry higher weight; peer-reviewed blueprints higher still.

Evidence tiers: PR = peer-reviewed; SEC = SEC filing or stockholder-approved; VS = vendor-stated roadmap. Timelines reflect stated targets; milestone slippage is addressed in the Scenario Analysis.

Company	Next major system	Expected window	Capability unlocked	Evidence tier
IonQ	256-qubit EQC system (6th gen, barium-133 + ytterbium)	2026 (delivering now)	High-performance NISQ-era commercial system, not a fault-tolerant system. Validates the world-record gate fidelity and qLDPC code quality that fault tolerance requires. AQ64+ delivered. The Walking Cat fault-tolerant architecture requires a minimum of ~2,500 physical qubits; fault-tolerant scale begins at the 200,000-qubit SkyWater milestone. SkyWater foundry close compresses next iteration to 2 months per cycle.	SEC (Form 425, May 2026); PR (arXiv:2510.17286, arXiv:2606.06455; arXiv:2604.19481 Walking Cat)
IonQ	200,000-qubit chip functional test (SkyWater foundry)	2028 (accelerated from 2030)	Chip-scale verification milestone, not a deployed commercial system. If the Walking Cat architecture performs as designed at this qubit count, a networked multi-chip configuration could support approximately 8,000 logical qubits — the AQ1,000 threshold range at which fault-tolerant utility for industrial chemistry and optimization becomes plausible. DARPA HARQ chip-to-chip networking targeted the same window. Both are forward projections from demonstrated small-device performance; this is the report’s highest-variance claim.	SEC (Form 425 committed); VS (SkyWater roadmap, Walking Cat blueprint arXiv:2604.19481)
Quantinuum	Sol system (next-generation trapped-ion)	2027 (targeted)	First system toward Apollo 2029 fault-tolerant target; logical-qubit count advance beyond Helios’ 48 error-corrected. First delivery record test at commercial-utility scale.	VS (S-1 roadmap); PR foundation (Helios, arXiv:2511.05465)
Quantinuum	Apollo fault-tolerant system	2029 (targeted)	Targeted fault-tolerant capability at commercial scale — Quantinuum’s stated path to validating its $14B valuation thesis. Requires Sol on schedule in 2027. Neither Sol nor Apollo has independent milestone verification; both are S-1 roadmap commitments.	VS (S-1 roadmap)
IBM	Kookaburra (bivariate-bicycle qLDPC)	2026 (targeted)	First IBM hardware demonstration of gross-code qLDPC on real hardware. The Tour de Gross paper projects ~90% overhead reduction versus surface codes — this is a theoretical result to be validated on Kookaburra, not yet demonstrated performance. Most important near-term IBM milestone: if the gross-code advantage holds on real hardware, it substantially strengthens IBM’s fault-tolerant roadmap credibility.	VS (IBM roadmap); PR foundation (Tour de Gross, arXiv:2506.03094 — theoretical; hardware undemonstrated)
IBM	Starling (200 logical qubits, 100M operations)	2029 (targeted)	IBM’s first stated large-scale fault-tolerant machine. $10B+ commitment. If Kookaburra delivers, Starling timeline is credible.	VS (IBM roadmap); SEC (capital commitment)
Google	Distance-9+ surface code extension	2026–27 (research timeline)	Confirmation that below-threshold suppression scaling holds beyond distance-7. This is a QEC memory result — demonstrating logical qubit stability, not logical gate operations. The distinction between memory-lifetime results and gate-operation results is the same discipline this report applies throughout. Next step in Google’s six-milestone roadmap.	VS (Google roadmap); PR foundation (Nature, Willow — distance-7 memory result)
QuEra	3rd-gen system (~10,000 physical, 100 logical)	2026–27 (targeted)	Advances from 96 Clifford logical qubits toward universal fault-tolerant operation; magic-state distillation for non-Clifford gates is the critical undemonstrated milestone.	VS (QuEra roadmap)
Atom Computing	Magne system (50 logical / 1,225 physical, Microsoft)	2026 (reported deployed)	First deployment of Microsoft-partnered neutral-atom system; 24–28 error-corrected logical qubits demonstrated via Microsoft QEC collaboration (joint company announcement, not yet peer-reviewed). Customer site in QuNorth Denmark. The 50-logical-qubit target is the system design specification, not yet the demonstrated result.	Joint company announcement (Microsoft/Atom, Jan 2026); not yet peer-reviewed
D-Wave	Gate-model 17Q (dual-rail superconducting)	2026 (targeted)	D-Wave’s first gate-model system; proof-of-concept for its dual-rail qubit approach and target suppression factor of 10×. Unproven architecture.	VS (D-Wave roadmap, June 2026)
PsiQuantum	First operational deployment site	2026–27 (targeted)	PsiQuantum’s all-or-nothing architecture means this is the first system-level result of any kind — no intermediate logical-qubit milestones precede it. If the platform performs as designed, it would represent photonic fault-tolerant computing at scale. The absence of intermediate milestones makes this the highest-uncertainty entry in the table: there is no experimental ladder to extrapolate from.	VS (PsiQuantum roadmap); PR foundation (Nature Comms, 2023 — fusion-based model only)
Infleqtion	Sqale 30+ logical qubits	2026 (targeted)	Next computing milestone on the neutral-atom platform; alongside sensing revenue already generating in Q1 2026.	VS (company roadmap)

The Near Horizon: Why 2026–2028 Is the Window That Decides Everything

A technology transition does not become obvious to the market until after the window to position well has closed. That is precisely where quantum computing stands today. The systems in the table above are not hypothetical — several are delivering now, others are months away, and the milestone sequence that runs from the current generation to AQ1,000 plays out before 2029 under the base case. What happens commercially during that window is not a gradual escalation. It is a step function, and the steps are imminent.

In 2026 and 2027, the commercial reality begins to separate from the research story. One distinction is essential before going further: the 256-qubit systems delivering now are not fault-tolerant systems. They are high-performance NISQ-era commercial platforms that validate the physical-qubit quality fault tolerance will require and establish the customer relationships and platform ecosystems that will define who leads when fault-tolerant scale arrives. The fault-tolerant era — for IonQ, on the Walking Cat architecture — begins at approximately 2,500 physical qubits at minimum and scales meaningfully at 10,000 and above. That is the 2028 SkyWater milestone, not the 2026 system. Companies with 256-qubit systems delivering to paying customers today are establishing the integration relationships, the software workflows, the institutional knowledge, and the supply agreements that define who gets the call when fault-tolerant systems arrive. A pharmaceutical company that pilots quantum chemistry workflows on a 256-qubit system this year is not just running an experiment. It is building a team, a library of circuits, and a vendor relationship. When fault-tolerant systems arrive and molecular simulation crosses from interesting to decisive, that team and that vendor relationship are already in place. The company that waits for fault-tolerant scale to begin evaluation starts two to three years behind.

In 2028, if the IonQ-SkyWater foundry milestone delivers and chip-to-chip networking is demonstrated on the DARPA HARQ timeline, the quantum industry crosses a threshold that has no analog in the prior decade of quantum development. A networked multi-chip system operating above AQ1,000 enables industrial-scale optimization problems that classical heuristics cannot match on cost or quality. It enables quantum chemistry simulations that are directly relevant to specific pharmaceutical targets. It enables financial risk calculations that currently require Monte Carlo approximations. These are not speculative applications. They are problems that quantum algorithms are theoretically designed to address, and AQ1,000 is the threshold at which the hardware is sufficient to attempt them on real industrial instances. The operator who has a platform relationship in place when that threshold is crossed has an advantage that a competitor cannot replicate by writing a check.

For the broader technology industry, the implications compound beyond quantum computing itself. The company that controls the quantum networking layer controls the infrastructure over which quantum-secured communications travel — the successor to the SSL certificate, the VPN, and the hardware security module. The company that controls the quantum sensing layer holds the position in GPS-denied navigation and subsurface mapping that GPS itself holds in conventional navigation. The company that controls the foundry for quantum chips holds the sovereign manufacturing position that TSMC holds for classical semiconductors — with the national-security implications that position carries. These are not quantum industry dynamics. They are technology industry dynamics, and they are being established right now, in 2026, with the contracts, acquisitions, and platform integrations documented in this report.

The investor looking at this field in June 2026 is not early. Early was 2020. But the investor is not late either — late is when the platform leader is obvious to everyone and the pricing reflects it. June 2026 is the moment when the evidence is strong enough to reach a defensible conclusion and the mainstream consensus has not yet caught up to what the primary sources say. That is the window this report is written for.

The Logical-Qubit Inflection

For most of the past decade, quantum hardware was marketed on qubit count, the way early microprocessors were marketed on clock speed: a single number, easy to compare, and increasingly disconnected from real performance. That era has ended. The binding problem in quantum computing is that physical qubits are fragile and accumulate errors faster than useful computations can complete. The only known route to large-scale utility is quantum error correction, which encodes one robust logical qubit across many noisy physical qubits and suppresses the error rate — but only if the physical error rate sits below a critical threshold.

In late 2024 and through 2025, that threshold was crossed in the laboratory for the first time on multiple platforms, and the consequence is that the entire field re-anchored its language around logical qubits during 2025 and 2026. The meaningful questions are now precise: how many logical qubits, encoded at what code distance, achieving what logical error rate per cycle, and sustaining how many operations before a logical failure occurs.

Three tiers of claim must be kept distinct, because vendors and press routinely blur them. A verified or fault-tolerant logical qubit is error-corrected with below-threshold suppression and no post-selection. An error-detected or post-selected logical qubit discards the runs in which an error is caught, which inflates headline counts but does not scale the same way — acceptance rates can collapse on the deepest circuits. A physical-qubit count is just that, and the choice to quote it is itself a signal of maturity. This report tags every figure by tier, and that single discipline resolves most of the apparent contradictions in cross-vendor comparison.

Figure 1. Demonstrated Logical Qubits Today

Verified/error-corrected results (blue) versus error-detected/post-selected results (amber), which are not directly comparable. QuEra's 96 and Google's distance-7 logical memory are peer-reviewed in Nature; Quantinuum's 48 error-corrected qubits are the figure comparable to QuEra's, not its 94 error-detected qubits. Google's entry is a single distance-7 logical memory, not a 96-qubit-style count.

Methodology & Scoring Framework

The framework below was stress-tested against a series of structural criticisms before any company was scored. The fixes are reflected throughout: companies are tiered before they are ranked, track record is normalized for disclosure candor, vendor claims are offset against independent benchmarks wherever they exist, logical-qubit figures are qualified by the underlying physics, speculative roadmaps are down-weighted, a measured cost-and-energy axis is introduced, and the full-stack and architecture dimensions are anchored to third-party taxonomy or else explicitly labeled as author-constructed.

Cohort Tiering — Applied Before Scoring

The twelve commercial and emerging organizations are not commensurable. A century-old firm with a ten-billion-dollar commitment, a venture-stage startup whose demonstrations outrun its revenue, and a photonic pure-play that cannot be scored on logical-qubit count by design cannot be placed on one ranked list without producing a number that looks comparable and is not. They are therefore sorted into three tiers and scored only within tier.

Tier	Members	Basis
Tier 1 — Commercial / revenue-bearing	IonQ, Quantinuum, IBM, Google, Rigetti, D-Wave, Infleqtion	Public or revenue-generating, with delivery obligations to real customers. Scored on the full seven-axis framework.
Tier 2 — Emerging / pre-scale-commercial	QuEra, Atom Computing, Pasqal, PsiQuantum, Xanadu	Venture or SPAC-stage; demonstrations exceed revenue. Scored on the seven-axis framework with maturity caveats.
Tier 3 — Sovereign / research (context players)	USTC (China), planqc (Germany)	Not profiled as companies — treated as context. Assessed on scientific milestones and national-strategy relevance in the sovereign section rather than ranked against commercial vendors.

The Seven Scoring Axes

Tier 1 and Tier 2 companies are scored on seven axes across two time horizons. Near-term (2026–2028) weights: Track Record 27%, Distance 22%, Published Specs 15%, Projected Power 7%, Cost & Energy 4%, Commercial Platform Completeness 15%, Full-Stack Integration 10%. Long-term (2029–2032) weights: Track Record 17%, Distance 22%, Specs 17%, Projected Power 21%, Cost & Energy 8%, CPC 10%, Full-Stack Integration 5%. The blended score combines 60% near-term and 40% long-term. The seventh axis — Full-Stack Integration (FSI) — captures owned versus partnered coverage across compute, networking, security, sensing, space, cloud, and software. Its inclusion in the blended score corrects a structural omission in earlier versions of this framework: a platform that owns foundries, operates live networking infrastructure, generates sensing revenue today, and holds satellite QKD capability was being scored identically on the hardware axes to one with hardware only. That equivalence was wrong. FSI is weighted most heavily near-term (10%) because platform breadth is the primary driver of revenue, government contracts, and customer lock-in through 2028. Its long-term weight falls to 5% because hardware capability becomes the dominant differentiator once fault tolerance is demonstrated at scale.

Axis	Weight	How it is scored
Track Record	Near 27% / Long 17%	Normalized for disclosure candor. A company that publicly discloses a missed milestone is not scored below one whose aggressive targets simply have not yet come due. Reported in two columns — demonstrated delivery and forward-claim aggressiveness. Down-weighted at long horizon because past delivery record becomes less predictive beyond five years.
Distance-to-delivery	Near 22% / Long 22%	Modality-adjusted. Incrementalist architectures show a continuous logical-qubit ramp; photonic architectures are structurally all-or-nothing at scale and are scored on component and loss milestones. Captures proximity to full useful delivery across the entire platform — not just the QEC roadmap. Absence of networking, foundry, or cloud distribution paths all constrain distance.
Published specs	Near 15% / Long 17%	Qualified by the deep-physics layer: every logical-qubit figure is tagged with code distance and logical error rate or lambda suppression factor. Neutral-atom figures carry the atom-reload and shuttling-overhead caveat; superconducting carries the refrigeration and wiring-density ceiling; photonic carries the optical-loss gating metric. Code efficiency (physical-to-logical ratio) is part of this axis.
Projected power	Near 7% / Long 21%	Down-weighted near-term because hockey-stick roadmaps correlate with marketing budget rather than capability. Weighted more heavily at the long horizon because fault-tolerant roadmap credibility becomes the dominant competitive variable. Where used, the axis requires a demonstrated volumetric figure — never an aspirational qubit count.
Cost & Energy-to-Solution	Near 4% / Long 8%	A measured axis adopting energy-to-solution and, where disclosed, cost and wall-clock time-to-solution at a fixed quality threshold. Structural physics constraints — dilution refrigerators at 25–50 kW per system, helium-3 supply chains, specialist cryogenic facilities — are scored as permanent penalties on this axis. IonQ’s third-party validated BOM of $30M versus $1B+ for superconducting is primary-source evidence. Absence of disclosure is itself scored as a transparency gap.
Commercial Platform Completeness	Near 15% / Long 10%	Cloud distribution breadth (AWS, Azure, GCP, own cloud each scored independently), revenue scale and growth rate against peers, quantum-dedicated IP density (not total corporate patents), and sovereign program participation (government contracts, trusted-foundry status, defense program wins). Its near-term weight of 15% is conservative relative to the Full-Stack Part III’s 25%; the long-term weight reduction reflects that cloud access becomes less differentiating as more vendors achieve multi-cloud distribution.
Full-Stack Integration	Near 10% / Long 5%	Scored from the Full-Stack Coverage Grid (Figure 5): owned or strongly-partnered coverage across compute, networking, security (QKD/QRNG), sensing, space, cloud, and software. Owned beats partnered; deployed beats announced; revenue-generating beats contractual. A company with live quantum networking on three continents, sensing revenue generating today, satellite QKD capability, and a DoD-trusted foundry in final regulatory close is scored materially higher than one with hardware only. The near-term weighting of 10% reflects that full-stack breadth is the primary driver of the compounding advantage described in the Competitive Leadership section — each delivered system, each government contract, and each networking node makes the competitive position harder to replicate. Long-term weight falls to 5% because hardware capability dominates once fault tolerance is demonstrated at scale.

The Independent-Benchmark Mandate

No specification or performance claim stands on vendor self-report alone where an independent check exists. The report actively seeks peer-reviewed results, standards-body application benchmarks, and academic re-analyses, and where none exist it says so rather than implying verification that is absent.

A word on the standards bodies is warranted, because it is easy to misread their role. Consortium-built benchmark suites are co-developed by competing members, and that broad, adversarial-by-construction participation is a strength, in the same way that the credibility of established machine-learning benchmarks comes precisely from rival firms agreeing on a shared method. Where results favor one architecture, the natural reading is that the benchmark is working — reporting a real performance difference — not that it is captured. A measurement is not discredited by who scores well on it. The only genuine limit, and it applies symmetrically to every vendor and every metric, is that a benchmark is not a substitute for independent peer-reviewed replication. The report therefore pairs benchmark results with peer-reviewed evidence where it exists and attaches no special suspicion to whichever company leads.

Framework Origin and the Full-Stack Axis

The full-stack and system-architecture dimensions are newer to analyst framing than the core hardware axes, which raises a fair question of whether they were constructed to favor particular companies. They were not, and the report guards against the appearance as well as the fact. The compute-architecture layers are mapped to established vendor-neutral academic taxonomy — the canonical five-layer model from the peer-reviewed architecture literature, corroborated by independent stack definitions — so that the axis is anchored externally rather than invented. The honest boundary is that these academic stacks describe the compute stack only. The broader platform elements — networking, sensing, space, and security — have no clean external taxonomy, so the report labels that portion author-constructed and justifies each pillar from primary evidence rather than implying an endorsement that does not exist.

Master Comparison: The Twelve at a Glance

The table below is the report's single most useful reference: every commercial and emerging company on one set of axes. Figures are the best publicly documented as of mid-2026; logical-qubit entries are tagged V (verified/error-corrected), ED (error-detected/post-selected), or — (not yet demonstrated). Peer-reviewed results are noted in the profiles that follow.

Company	Modality	Qubits (physical)	Best 2Q fidelity	Logical qubits	Next milestone (date)	Cloud access
IonQ	Trapped-ion (two platforms)	256 (EQC/Yb product) / 40 (Ba research)	99.99% (EQC demo)	BREAKEVEN qLDPC: 4 logical/18 phys; ~4.3×(X)/~8.5×(Z) vs. SC demo (arXiv:2606.06455)	256-qubit integrated test (2026)	AWS, Azure, GCP, own
Quantinuum	Trapped-ion	98 (Helios)	~99.92%	48 V / up to 94 ED	Sol ~100 logical (2027)	Azure, own
IBM	Superconducting	~1,121 (fleet) / 120 (Nighthawk)	~99.5% (Heron R2)	roadmap (200 by 2029)	Kookaburra qLDPC module (2026)	IBM Quantum / Qiskit
Google	Superconducting	105 (Willow)	below-threshold (d7)	d7 surface, 0.14%/cyc	long-lived logical qubit	Early-access / NQCC
Rigetti	Superconducting	108 (Cepheus-1)	99.5% (Ankaa-3 fSim)	— (decoding demos)	99.5% on 108Q; >1,000Q (3-4 yr)	AWS, Azure, qBraid, own
D-Wave	Superconducting (gate+anneal)	17 (gate, 2026)	n/a (early)	10 by 2030 (target)	49-qubit gate system (2027)	Leap cloud
QuEra	Neutral-atom	448 (current)	Rydberg, high-rate	96 V (Nature)	100 logical / 10k phys (2026-27)	AWS Braket
Atom Computing	Neutral-atom	~1,180 (Phoenix)	Rydberg	24-28 (w/ Microsoft)	Magne 50 logical (deployed 2026)	Azure
Pasqal	Neutral-atom	~324 (Fresnel 2)	lags leaders	roadmap to 2029	Vela; 100s logical (2029)	Azure, GCP, CUDA-Q, Scaleway
Infleqtion	Neutral-atom	1,600 (Sqale)	99.73% CZ	12 (post-select)	30 logical (2026)	Oqtant, NVQLink
PsiQuantum	Photonic (fusion)	Omega chipset	loss-limited	— (by design)	wiring chips to scale	On-prem deployment sites
Xanadu	Photonic (GKP)	12 (Aurora demo)	loss-limited	GKP qubit (Nature)	FT ops (2028); 500 logical (2029-30)	Cloud + PennyLane

Company Profiles

Trapped-Ion

IonQ NYSE: IONQ · Tier 1 │ Near 9.2 Long 9.3 Blended 9.3

● PEER-REVIEWED — qLDPC breakeven (June 2026, peer-reviewed, arXiv:2606.06455); 99.99% fidelity without ground-state cooling (arXiv:2510.17286, peer-reviewed); $30M BOM at 2M-qubit scale vs $1B+ superconducting — third-party validated (Q3 2025 SEC 8-K); five consecutive quarterly guidance beats; RPOs $470M (Q1 2026 SEC 8-K). Track Record raised to 9.3 (from 9.2 in the prior series report): the June 4, 2026 peer-reviewed qLDPC breakeven removes the one demonstrated-error-correction gap competitors could previously cite — IonQ now holds a peer-reviewed error-correction result on real hardware alongside Google, QuEra, and Quantinuum, rather than resting on fidelity and roadmap alone. The adjustment is deliberately modest: the result strengthens the evidentiary basis for IonQ’s already category-leading Track Record and Specs scores rather than materially repricing them, and the blended score and the 1.2-point lead over IBM are essentially unchanged. The breakeven was demonstrated on a 40-ion barium-133 research platform with overlapping error bars against the physical layer, which is why it supports the existing high scores rather than driving them higher.

IonQ enters this report as the most complete full-stack quantum platform in the field — the only company in the cohort that simultaneously holds a leading hardware position, a deployed quantum networking business, a revenue-generating quantum sensing line, a quantum security product, and, pending regulatory close, owned semiconductor manufacturing through SkyWater. The strategic significance of the sixth-generation 256-qubit system is not that it is itself a fault-tolerant machine — it is not; it is a high-performance NISQ-era commercial system that validates the gate fidelity and qLDPC code quality fault tolerance will require. Its significance is that it anchors that full stack: it is the compute core around which the networking, sensing, security, and foundry layers compound. The platform thesis of this report is that the company which integrates demonstrated hardware progress into a complete commercial platform sets the pace for the era, and IonQ is the company that has assembled the most of that platform. The 256-qubit generation is the point at which that integration becomes commercially visible — sold to paying customers, anchored by a secure quantum network, and spanning computing, networking, sensing, and security in a single IP-generating relationship.

The system is chip-based, controlled by proprietary Electronic Qubit Control (EQC) — precision electronics rather than lasers — a capability strengthened by the Oxford Ionics acquisition and identified in the Walking Cat blueprint as the explicit scalability path beyond the current research platform. The first 256-qubit system was sold in the first quarter of 2026, after the company received its first ion-trap chip samples from fabrication and moved from component-level to integrated full-system testing. What the 256-qubit generation means going forward is the foundation for the fault-tolerant roadmap: it establishes the physical-qubit quality, the customer relationships, and the platform integration that the 2028 SkyWater 200,000-qubit milestone will build on. It is the highest-variance set of forward claims in the cohort — the SkyWater-accelerated timeline is aggressive — but it is anchored in capabilities the hardware has already demonstrated rather than in roadmap promises alone.

On measurement, IonQ has moved beyond the single-number algorithmic-qubits metric it once championed to an application-centric benchmarking framework, published in April 2026. Modeled on machine-learning benchmarking practice, it spans thirteen benchmarks across optimization, chemistry, machine learning, data loading, simulation, and foundational algorithms, with time-to-solution, energy-to-solution, and solution quality as the headline metrics rather than qubit count. The code is open-source and runnable on any system, and the comparative results were independently validated by a third-party firm. The framework is candid by design — it publishes a quantum-chemistry criterion IonQ's own hardware has not met, and concedes that faster-gate architectures hold a time-to-solution advantage on that class of problem. It is the most application-meaningful methodology publicly available; it is also vendor-authored and reports primarily IonQ results, so it is best read as a strong, open, but not disinterested contribution.

The system architecture is specified in unusual detail. In April 2026 IonQ researchers published the Walking Cat blueprint, an end-to-end fault-tolerant design built on quantum low-density parity-check codes, using cat states as non-destructive measurement probes and physically shuttling ions through a charge-coupled-device grid to achieve any-to-any connectivity without fixed wiring. The design philosophy explicitly prioritizes buildability over theoretical optimality, and its headline resource claim is that hundreds of logical qubits could be realized from as few as 2,514 physical qubits, scaling toward two million physical and eighty thousand logical qubits by 2030. It is the most complete public engineering specification for fault tolerance to date; the resource counts remain a preprint claim pending independent corroboration.

Breaking primary source (arXiv:2606.06455, June 4, 2026 — one day prior to this report's finalization). A critical precision before reading this result: the device used is a dedicated QEC research platform of forty ¹³³Ba⁺ (barium-133) ions in a stationary chain, with all-to-all connectivity via steerable 532nm Raman beams. This is a distinct system from IonQ's EQC/ytterbium 256-qubit product line; the two coexist and serve different purposes. The paper's conclusion explicitly identifies EQC as the path to scale the lessons from this barium research device into larger systems with greater gate parallelism.

The result. Nine quantum error-correcting codes were demonstrated on a single trapped-ion device without hardware reconfiguration, spanning three families: five qLDPC codes (three BB5 and two GB4), two toric codes, and one concatenated code. On the BB5 [[18,4,3]] code — encoding four logical qubits into eighteen physical qubits, the same code parameters as the prior best superconducting demonstration — the team achieved a logical error rate of 2.01×10⁻² per qubit per cycle (X basis) and 1.08×10⁻² (Z basis). The prior superconducting result (which required bespoke long-range couplers engineered specifically for that one code) achieved approximately 8.67% (X) and 9.15% (Z). The improvement is therefore ~4.3× better on X errors and ~8.5× better on Z errors — both figures matter and must be stated together; the report’s earlier use of '9×' alone was imprecise.

Breakeven lifetime. Defining qubit lifetime as the 1/e survival time generalized across decay channels, the best result is the GB[[26,2,5]] code at a combined lifetime TN of 3.95 ± 0.68 seconds, compared to 3.3 ± 0.9 seconds for the underlying physical qubits. The error bars overlap substantially: this is a marginal exceed, real in central value but not conclusively resolved by the statistics. Most other codes reach TN of 2.1–3.4 seconds, within physical qubit lifetime. The concatenated [[4,2,2]] code is the outlier at 1.32 ± 0.34 seconds. The paper states that in all codes except the concatenated one, logical qubit lifetimes are comparable to physical qubits to within error bars, and exceed the physical value slightly in several cases. The precise numbers are reported here because the nuance matters: 'breakeven' is accurate; 'exceeds the physical qubit' requires the error-bar caveat.

The no-hardware-reconfiguration finding is the paper's second headline result and arguably the one with the largest architectural significance. The prior superconducting demonstration required a chip with bespoke air-bridge structures and Josephson junctions engineered specifically for one code. IonQ's stationary-chain trapped-ion device ran all nine codes with no hardware reconfiguration, leveraging all-to-all connectivity to remap qubits to ions advantageously for each code's specific connectivity graph. The paper explicitly calls this remarkable. For institutional buyers this is decision-relevant: the hardware does not need to be re-engineered each time the code choice changes.

The OMG architecture and efficiency. The optical-metastable-ground implementation is the paper's third major contribution, and its efficiency implications are quantified in the paper itself. In trapped-ion systems that rely on ion transport — the paper cites Helios specifically — transport and cooling consume the majority of execution time, and up to 50% of the ions are used only for cooling. IonQ's stationary-chain OMG implementation eliminates both overheads: ancilla qubits that are measured frequently during error correction simultaneously serve as the sympathetic cooling mechanism, removing the need for dedicated coolant ions or a separate atomic species. No transport means dramatically reduced cooling requirements. This is not a theoretical claim: it is the paper's explicit quantification of the efficiency gap between the two trapped-ion QEC approaches, sourced from Helios's own characterization data.

Post-selection disclosure. The paper is explicit that it uses post-selection — rejecting shots in which leakage was detected during any mid-circuit measurement round. It also demonstrates that converting this post-selection to erasures instead degrades the logical error rate by only 10–17%, establishing that the result is robust and not artificially inflated by aggressive post-selection. This level of methodological transparency is a positive signal; it allows the result to be compared fairly with other QEC demonstrations.

Theory-to-experiment-to-product bridge. The paper's conclusion explicitly names EQC as the scalability path: increasing gate parallelism via electronic qubit control, while leveraging record-fidelity gates. This makes the paper the middle link in a coherent IonQ progression: Walking Cat (FTQC theory, April 2026) → this barium/OMG result (experimental QEC validation, June 2026) → EQC/256-qubit chip (product line). The paper itself cites Walking Cat as one of the target FTQC architectures its QEC demonstrations support. This progression is the clearest evidence in the entire report for the commercialization thesis that next-generation systems are the inflection into everyday use.

IonQ's two-qubit gate fidelity reached 99.99% in October 2025, a world record set on the prototypes underpinning the 256-qubit EQC systems — a separate result from the barium/OMG research device. Financially, first-quarter 2026 revenue was $64.7 million, 755% year-on-year growth and 30% above the midpoint of guidance — the fourth consecutive quarterly beat and the largest quarter in the company's history. Full-year 2026 guidance raised to $260–270 million. RPOs of $470 million represent 554% year-on-year growth. Cash and investments: $3.1 billion as of March 31, 2026. IonQ became the first public quantum company above $100 million in annual GAAP revenue in 2025. Q1 2026 commercial highlights: 6th-generation 256-qubit chip-based systems contracted to three named customers, with expected delivery across Q4 2026 to Q1 2027 — University of Chicago (commercial hardware agreement, November 2025, $9.2M Q1 2026 revenue recognized; SEC Form 10-Q primary source), University of Cambridge (strategic collaboration agreement March 2026, named in Q1 2026 SEC 8-K headline; note: the author holds no equity position in Cambridge), and Horizon Quantum Holdings (Quantum Systems Agreement signed March 31, 2026, announced April 9, 2026; note: the author holds disclosed long equity in Horizon Quantum Holdings — this sale should be read with that COI context in mind); quantum sensing products commercialized and contributing to revenue today (CFO Inder Singh, Q1 2026 earnings call); SDA HALO $39 million contract for tactical space communications signed; MDA SHIELD IDIQ selection; DARPA HARQ selection for modular quantum interconnect program. The SkyWater acquisition received stockholder approval May 8, 2026 and is awaiting regulatory close, with chip samples already delivered to IonQ's R&D labs prior to that vote.

Key specifications

Distance: Two parallel platforms. EQC/ytterbium 256-qubit product system — first unit sold Q1 2026 to University of Cambridge; integrated testing underway. Barium-133/OMG 40-ion research platform — the QEC demonstration device. SkyWater acquisition: stockholder-approved May 8, 2026, regulatory close pending — chip samples already delivered. With SkyWater: 200,000-qubit chip functional testing accelerated from 2030 to 2028; design-to-first-samples from 9 months to 2 months (SEC Form 425, legally accountable disclosures). 2M physical / 80k logical targeted 2030.
Specs (EQC/ytterbium product): Gate error 8.4(7)×10⁻⁵ (≡1–99.9916% fidelity, stated as >99.99%) on 'smooth gate' prototypes without ground-state cooling — operates above the Doppler limit, enabling faster and simpler device operation. Primary source: Oxford Ionics (an IonQ company), arXiv:2510.17286, October 2025. 2-qubit demonstrator result, not yet validated on 256-qubit device. Oxford Ionics single-qubit fidelity: 99.9997%; EQC uses precision electronics, not lasers.
Specs (barium-133/OMG research device, arXiv:2606.06455): 40 ¹³³Ba⁺ ions in stationary chain; all-to-all connectivity via 532nm Raman beams; OMG mid-circuit measurement without ion transport or dedicated coolant ions. Nine QEC codes demonstrated without hardware reconfiguration. BB5 [[18,4,3]]: 2.01×10⁻² logical error/qubit/cycle (X), 1.08×10⁻² (Z) — ~4.3× better on X errors and ~8.5× better on Z errors vs. prior best superconducting qLDPC (X: 2.01% vs ~8.67%; Z: 1.08% vs ~9.15%). Code efficiency: 4.5:1 physical-to-logical overhead (vs. Steane [[7,1,3]] at 7:1). Best lifetime: GB[[26,2,5]] at TN=3.95±0.68s vs. physical 3.3±0.9s (marginal exceed, overlapping error bars).
Cost & Energy: Room-temperature vacuum chamber; no dilution refrigerator; no helium-3; no cryogenic infrastructure. Third-party validated BOM cost for 2M-qubit system: less than $30M, versus over $1B for superconducting competitors — approximately 33× lower (Q3 2025 SEC 8-K primary source). EQC integrates qubit control on classical semiconductor chips, same fab process as commercial electronics. Single AOD control system scales to many qubits with the same energy profile. No ground-state cooling required (arXiv:2510.17286). These are structural physics advantages, not roadmap projections.
Track record: five consecutive quarterly guidance beats (20–55% above midpoints); RPOs $470M +554% YoY; revenue $130M FY2025 (first public quantum company above $100M GAAP); AQ64 delivered 3 months early; 99.99% fidelity October 2025 as announced; 6th-generation 256-qubit system contracted to three customers: University of Chicago (hardware, $9.2M Q1 2026 revenue), University of Cambridge (8-K named), Horizon Quantum Holdings (March 31, 2026 agreement; author holds HQ equity). DARPA HARQ, MDA SHIELD IDIQ, SDA HALO $39M all Q1 2026. Cash $3.1B March 31, 2026.

Full-stack & ecosystem. IonQ is the clearest example of owned vertical integration in the cohort, and uniquely, several of its platform pillars are already generating commercial revenue — not pipeline. Quantum sensing products were described by CFO Inder Singh in the Q1 2026 earnings call as already commercialized, contributing to revenue today. The SDA HALO $39 million contract for tactical space communications is a signed contract. The MDA SHIELD IDIQ and DARPA HARQ selections are live program participations. Through acquisition IonQ has assembled: foundry capacity (SkyWater, stockholder-approved, regulatory close pending); networking and photonic-interconnect capability (Lightsynq, Skyloom, Qubitekk); quantum key distribution and security (ID Quantique, roughly three hundred patents); space-based quantum key distribution infrastructure (Capella Space, Skyloom); precision sensing and atomic clocks generating commercial revenue now (Vector Atomic); and AI-driven infrastructure (Seed Innovations). It defines itself across five pillars — compute, networking, security, sensing, and space — a breadth no other company in the cohort attempts, and one that is increasingly live rather than aspirational. The integration risk is real: digesting eight-plus acquisitions across five technology domains is an execution hazard, and revenue from acquired businesses is not the same as organic platform revenue. These are the two caveats that accompany every IonQ acquisition claim in this report.

Verdict. Near-term leader on full-stack breadth, fidelity, cost structure, and commercial delivery cadence. Five consecutive quarterly guidance beats, RPOs of $470M, and a foundry acquisition in final regulatory close collectively justify a higher Track Record score than any previously scored in this series. The qLDPC code efficiency advantage (4.5:1 physical-to-logical versus Steane's 7:1) is peer-reviewed and quantified. The third-party validated cost structure ($30M BOM at scale versus $1B+ for superconducting) is disclosed in SEC filings. The SkyWater acquisition — stockholder-approved, awaiting only regulatory sign-off, with chip samples already delivered — closes the foundry gap ahead of schedule. These are not forward claims; they are the current evidentiary position.

Watch for: SkyWater regulatory approval close — stockholder approval already secured May 8, 2026, with only regulatory sign-off remaining. First integrated EQC chip performance data post-close will confirm whether the 2-month design-to-samples acceleration (versus prior 9 months) holds in practice. These are the two highest-signal events for the 2028 fault-tolerant thesis.

What would change this verdict: This verdict would change if: (a) SkyWater regulatory approval encounters unexpected delay or block; (b) first integrated EQC chip data shows significantly lower fidelity than the 2-qubit demonstrator result; (c) a peer demonstrates below-threshold qLDPC scaling on trapped-ion or superconducting hardware before IonQ’s EQC product system does; or (d) IBM’s Anderon foundry progresses from LOI to binding agreement and begins operations faster than anticipated, compressing IonQ’s foundry lead window.

Quantinuum Private; Honeywell majority · Tier 1 │ Near 6.9 Long 7.2 Blended 7.0

● PEER-REVIEWED — 48 error-corrected logical qubits (arXiv:2602.22211, peer-reviewed); FT algorithm execution (arXiv:2603.04584, peer-reviewed); hardware specs from S-1 registration statement. Track Record set at 7.0: one year of audited public financials; one publicly tracked hardware milestone (Helios); FY2025 revenue $30.9M on +34% YoY growth; severe customer concentration disclosed in S-1; no multi-year public accountability framework. Distance set at 6.5: no quantum networking program or deployment; no foundry position; no AWS or GCP cloud distribution; Apollo targets 2029 fault tolerance, one year behind IonQ’s SkyWater-accelerated 2028 milestone; no intermediate chip-scaling path disclosed. Specs remain strong at 8.5 on the strength of the peer-reviewed Helios QEC results.

Quantinuum has, arguably, the cleanest delivery record in the field. Its Helios system, launched in November 2025, hit the forty-eight-logical-qubit milestone the company had committed to, and a peer-reviewed hardware paper documents the architecture: a transport-based trapped-ion processor using barium ions on a two-dimensional surface charge-coupled-device with a four-way junction that separates memory from logic zones. The roadmap runs from Helios to Sol, expected in 2027 and targeting roughly one hundred logical qubits approaching five-nines logical fidelity, to Apollo in 2029, targeting hundreds of logical qubits and full fault tolerance.

A March 2026 result requires careful reading, because it is widely misreported. From ninety-eight physical qubits, the team produced forty-eight error-corrected logical qubits and, separately, up to ninety-four error-detected logical qubits using post-selection. The ninety-four figure is not comparable to a verified logical-qubit count: it relies on discarding runs in which errors are detected, and the acceptance rate falls to a few percent on the deepest circuits. The forty-eight error-corrected qubits are the figure that belongs in a like-for-like comparison.

A second March 2026 result is categorically distinct from the memory experiment above and deserves equal weight. In a collaboration with JPMorgan Chase's research team (arXiv:2603.04584, March 2026), Quantinuum demonstrated end-to-end fault-tolerant execution of complete quantum algorithms on the Helios processor — the first demonstration of its kind. Using the Steane [[7,1,3]] code, the team ran the Quantum Approximate Optimization Algorithm on up to twelve logical qubits using ninety-seven of the ninety-eight physical qubits available, with two thousand one hundred thirty-two physical two-qubit gates, applied to portfolio optimization and combinatorial problems. They also ran the Harrow-Hassidim-Lloyd algorithm for a Poisson equation instance. A memory experiment asks whether a logical qubit can survive; a fault-tolerant algorithm execution asks whether a complete useful computation can be performed with error correction active throughout. This result answers the second question affirmatively on real hardware for the first time.

The company has filed a registration statement for a public listing targeting a valuation between approximately eleven and fourteen billion dollars, which makes its disclosures unusually accountable for a company at this stage.

A scoring note on Distance (6.5) and Track Record (7.0). The Distance score reflects three structural absences that constrain proximity to full useful delivery: no quantum networking program of any kind; no foundry or owned chip fabrication path; and cloud distribution limited to Azure and its own portal, with no AWS or GCP presence. Each of these gaps is real and is not a forward claim — they are current facts. The Apollo 2029 timeline also places Quantinuum one year behind IonQ’s SkyWater-accelerated 2028 milestone for 200,000-qubit functional testing. The Track Record score reflects one year of audited public financials, one publicly tracked hardware milestone, and a revenue base of $30.9M growing at 34% year-on-year with significant customer concentration disclosed in the S-1. Quantinuum’s strongest axis is Specs (8.5) — the peer-reviewed QEC results are genuine and significant — and readers for whom demonstrated error correction today is the primary criterion should weight that axis explicitly rather than relying on the blended score.

Key specifications

Distance (scored 6.5): Helios shipping (Nov 2025); Sol ~2027 (~100 logical); Apollo 2029 (full FT). No foundry, no networking program, no AWS/GCP cloud distribution — three structural gaps in distance-to-full-useful-delivery. Apollo 2029 is one year behind IonQ’s SkyWater-accelerated 200k-qubit target (2028 functional testing). No intermediate chip-scaling path disclosed.
Specs (scored 8.5): 98 barium qubits, ~99.92% 2Q fidelity (7.9×10⁻⁴ two-qubit error, readout 5×10⁻⁴), 2D QCCD with X-junction, ~2:1 physical-to-logical via color codes. Code efficiency: Steane [[7,1,3]] at 7:1 physical-to-logical (vs. IonQ’s BB5 at 4.5:1). Specs score reflects genuine peer-reviewed QEC strength.
Power: 48 error-corrected logical qubits (memory); fault-tolerant QAOA on 12 logical / 97 physical qubits (arXiv:2603.04584, Mar 2026, JPMorgan collaboration) — first peer-reviewed fault-tolerant algorithm execution on real hardware.
Track record (scored 7.0): FY2025 revenue $30.9M (+34% YoY); customer concentration risk (RIKEN, S-1 disclosed); one publicly tracked hardware milestone (Helios); one year of audited public financials. IPO pricing $53–55/share (~$14.3B market cap), Nasdaq ticker QNT.

Full-stack & ecosystem. Strong on software and real-time control (its Guppy language and control engine), and accessible through the major hyperscaler clouds. Less acquisitive than IonQ; its integration is organic rather than bought.

Verdict. Strong on demonstrated QEC — the peer-reviewed Helios results are real and the FT algorithm execution is a genuine first. But Track Record is set at 7.0 on one year of public financial history and one hardware milestone; Distance scores 6.5 because there is no networking program, no foundry, limited cloud distribution, and an Apollo timeline one year behind IonQ’s accelerated roadmap. Third in Tier 1 on near-term score behind IonQ and IBM. Its QEC lead on delivered operations is its most defensible position and should be weighted accordingly by any reader for whom demonstrated error correction today is the primary criterion. Three structural observations belong alongside that credit. First, the code efficiency gap compounds at scale: Quantinuum’s Steane [[7,1,3]] code requires seven physical qubits per logical qubit; IonQ’s BB5 [[18,4,3]] code requires 4.5. To reach 10,000 logical qubits — the threshold at which pharmaceutical and materials applications become tractable — the Steane path requires 70,000 physical qubits and the BB5 path requires 45,000. Quantinuum has no foundry; IonQ has SkyWater. The code-efficiency and foundry-access gaps compound together. Second, the business has one path to its $14 billion valuation: Sol on time in 2027, Apollo on time in 2029, customer diversification away from the RIKEN concentration the S-1 discloses, and no competing trapped-ion system demonstrating comparable QEC on a product platform before it does. All four conditions must hold simultaneously. Third, generating $30.9 million in revenue with one dominant customer is a bilateral relationship with a research institution, priced as a transformative technology company. The physics is excellent. The risk-adjusted competitive position is more complicated.

Watch for: Sol system delivery in 2027 is the single most important milestone to track — it is the first system where Quantinuum’s delivery record will be tested at a scale that approaches commercial utility. IPO performance post-listing (ticker QNT) will establish whether the market’s valuation assumption holds as a standalone entity.

What would change this verdict: This verdict would change materially if: Sol delivers on schedule at the stated logical-qubit count and fidelity; or if Quantinuum initiates a quantum networking program with any deployed infrastructure; or if a foundry partnership with disclosed terms and committed capacity is announced; or if customer concentration drops below 50% of revenue.

Superconducting

IBM NYSE: IBM · Tier 1 │ Near 7.9 Long 7.9 Blended 7.9

● PRIMARY SOURCE — Roadmap milestones from SEC filings and investor days; Anderon foundry from May 2026 SEC disclosure; bivariate-bicycle codes from Nature (2024) peer-reviewed paper.

IBM offers the most fully specified path to large-scale fault tolerance in the industry, backed by the deepest capital commitment. Its roadmap proceeds through a sequence of named processors — Loon, Kookaburra, Cockatoo — each demonstrating a required capability, culminating in Starling, planned for 2029 as the first large-scale fault-tolerant machine, and Blue Jay beyond it. Starling is specified at two hundred logical qubits running one hundred million quantum operations; Blue Jay at two thousand logical qubits and one billion operations.

The architectural bet is quantum low-density parity-check codes — specifically the bivariate-bicycle or gross-code family — which IBM credits with cutting physical-qubit overhead by up to ninety percent against surface codes, paired with a real-time classical decoder. In June 2026 IBM committed more than ten billion dollars to the program, the largest disclosed commitment in the sector, and it reports having met each of its roadmap milestones to date.

Key specifications

Distance (scored 7.0): Nighthawk (2025-26) → Kookaburra (2026) → Cockatoo (2027) → Starling (2029, 200 logical) → Blue Jay (2,000 logical). Anderon foundry is an LOI only — no binding agreement, no construction timeline, no operational facility. No quantum networking program. Dilution refrigerator constraint is a structural engineering barrier to full-useful-delivery at scale.
Specs: ~1,121-qubit fleet, Heron R2 ~99.5% 2Q; bivariate-bicycle qLDPC cutting overhead ~90% (Tour de Gross, arXiv:2506.03094).
Power: Starling specified at 100M operations on 200 logical qubits.
Cost & Energy (scored 6.5): 25–50 kW per system; dilution refrigerator at ~15 mK; helium-3 supply chain; specialist cryogenic facility required; vibration isolation. IBM System One requires airtight glass enclosure. $1B+ BOM at scale vs IonQ $30M (third-party validated).
Track record: meets each stated roadmap milestone. CPC (scored 7.5): IBM Quantum cloud dominant; Qiskit largest open-source quantum framework; strong enterprise contracts; no AWS/GCP quantum distribution; no quantum networking; Anderon LOI not an operational foundry.

Full-stack & ecosystem. IBM exemplifies the software-platform-plus-partnership model. Qiskit is the most widely used quantum software stack, anchoring a network of more than two hundred fifty member organizations, and IBM’s late-2025 partnership with Cisco targets the networking layer through quantum networking units and transducers for linking processors across data centers. IBM owns the software and compute layers and partners for networking rather than acquiring it. A structural note on Qiskit: because it is explicitly vendor-neutral and runnable on IonQ, Quantinuum, Rigetti, and other systems, IBM’s software moat benefits the entire field rather than exclusively IBM. An organization that builds quantum workflows on Qiskit is not locked into IBM hardware — it can port those circuits to any cloud-accessible system. This is architecturally different from a proprietary software dependency and should be read as ecosystem leadership rather than switching-cost protection.

Verdict. Leads on capital, on path credibility, and on software ecosystem. The most dangerous near-term competitor on the hardware roadmap axis. Two structural constraints limit the case, however, and neither is a roadmap problem. First: the dilution refrigerator physics are not an engineering challenge waiting for a clever solution — superconducting qubits require near-absolute-zero temperatures because they are superconducting; that is definitional, not incidental. The $30M versus $1B-plus bill-of-materials gap at scale reflects this physics directly and does not close through better engineering. It closes only by changing the modality. Second: Anderon is a letter of intent. IonQ has chip samples in hand from SkyWater with stockholder approval already secured. IBM has a press release. The foundry race is not close on the current evidence, and every quarter that gap persists is a quarter of wafer-cycle experience and hardware-iteration data that IonQ accumulates and IBM does not. IBM’s strongest long-term asset is capital depth; its clearest near-term falsifier is whether Kookaburra delivers a demonstrated qLDPC memory result before IonQ demonstrates integrated EQC chip fidelity at product scale.

Watch for: Kookaburra (2026) delivering qLDPC memory — IBM's first public demonstration of a gross-code implementation on real hardware will be the most important technical milestone between now and Starling.

What would change this verdict: This verdict would change if: Kookaburra slips into 2027; or if the Tour de Gross architecture's connectivity requirements prove harder to manufacture at scale than the blueprint assumes.

Google Quantum AI Alphabet · Tier 1 │ Near 7.1 Long 7.5 Blended 7.2

● PEER-REVIEWED — Willow below-threshold result (Nature, 2024); RL-on-Willow (arXiv:2511.08493, peer-reviewed); Quantum Echoes verifiable advantage (2025).

Google holds the strongest peer-reviewed error-correction result in the field. Its Willow processor demonstrated, in a Nature publication, a distance-seven surface code on one hundred one physical qubits achieving a logical error rate of roughly 0.14% per cycle, with the logical error rate suppressed by a factor of about 2.14 each time the code distance grew — the defining signature of below-threshold operation — and a logical memory living about 2.4 times longer than its best physical qubit. This is the clearest published evidence that error correction scales as theory requires, on real superconducting hardware.

Through 2025 and into 2026 Google extended the result, demonstrating verifiable quantum advantage with its Quantum Echoes work and exploring more efficient color and dynamic codes. Its six-milestone roadmap targets a useful, error-corrected machine by the end of the decade, and it operates a dedicated fabrication facility in Santa Barbara.

Full-stack & ecosystem. Less of a commercial full-stack story than IBM or IonQ; Google’s strength is research depth and in-house fabrication rather than networking, sensing, or a broad partner ecosystem. Access is via early-access programs and selected national research centers. A critical observation for competitive analysis: every gap in Google’s commercial position — no enterprise sales motion, no revenue, no networking, no procurement pathway — reflects a strategic choice by Alphabet, not a capability constraint. Google has more PhD-level quantum researchers than any company in the cohort, its own fabrication facility in Santa Barbara, and the strongest peer-reviewed QEC result in the field. The moment Alphabet allocates a portion of its capital to commercializing that capability, most of the Distance and CPC gaps narrow sharply. The competitive risk from Google is not its current position — it is a single board decision away from being the most dangerous entrant in the field.

Verdict. Co-leader on demonstrated hardware capability today, on the strength of the strongest published below-threshold result. But the demonstrated result is a memory experiment on one code type at one code distance — the surface code family that qLDPC approaches from IBM and IonQ are specifically designed to make less necessary. Google’s Λ=2.14 at distance-7 is real and peer-reviewed; it is also the best result on a code architecture that the field is moving past. No commercial product, no revenue, no networking, no procurement pathway, and no disclosed timeline for any of those exists. Every quarter Google does not commercialize is a quarter of customer relationships, government contracts, networking deployments, and foundry iterations that competitors accumulate and Google does not. The compounding runs in both directions: IonQ’s advantage compounds with each delivered milestone; Google’s research lead depreciates with each quarter it remains non-commercial. The threat is real but it is a strategic-choice threat — and until Alphabet makes a different choice, the scores reflect the current facts.

Watch for: The next code-distance step above distance-7 — demonstrating distance-9 or distance-11 below-threshold operation would confirm that the suppression scaling holds beyond the current data point.

What would change this verdict: This verdict would change if: a peer demonstrates stronger surface-code suppression (Λ > 2.14) at comparable physical qubit count; or if Google's path to logical qubit operations (not just memory) significantly lags Quantinuum or IonQ.

Rigetti Nasdaq: RGTI · Tier 1 │ Near 6.7 Long 7.0 Blended 6.8

● PRIMARY SOURCE — Cepheus-1 specs from regulatory filings; fidelity from company-disclosed benchmarks; real-time decoding from trade press.

Rigetti is the cohort’s most transparent discloser of its own slippage, and under the candor-normalized framework that transparency is scored as a positive signal, not a penalty. Its current system, Cepheus-1-108Q, is a 108-qubit machine built from twelve nine-qubit chiplets — the company’s modular thesis made concrete — reaching general availability around the end of the first quarter of 2026 after the company took additional time to improve fidelity, a delay it disclosed plainly. It has signaled intent to field a system exceeding one thousand qubits within a few years and to update its roadmap during 2026.

Reported median two-qubit gate fidelity on the 108-qubit system is around 99.1%, with a target of 99.5%, and the company has demonstrated higher fidelities on smaller and prototype systems. Its regulatory filings disclose repeated past roadmap changes, which is precisely the kind of candor the framework rewards.

Key specifications

Distance: 108Q GA ~Q1 2026; target 99.5% on 108Q this year; >1,000-qubit system in 3-4 years.
Specs: 108 qubits (12x9 chiplets), 99.1% median 2Q (99.5% fSim on Ankaa-3), 60-80ns gate speed.
Power: chiplet/modular scaling; real-time decoding with partner at small distance.
Track record: discloses past roadmap changes plainly — candor scored as a positive.

Full-stack & ecosystem. Differentiated by an owned fabrication facility and by real-time decoding work with partners. Available across the major clouds. A smaller balance sheet than the incumbents, but debt-free with disclosed cash discipline.

Verdict. Mid-tier on demonstrated capability, but the transparency of its disclosure makes its track-record score more trustworthy than a superficially cleaner but untested record would be.

Watch for: Any independent benchmark of Cepheus-1-108Q published by a third party — Rigetti's candor about slippage is a positive signal, but the field needs independent validation of its fidelity claims at full qubit count.

What would change this verdict: This verdict would change if: an independent benchmark at full 108Q confirms the 99.5% fidelity figure under application-circuit conditions; or if the 1,000+ qubit roadmap receives specific engineering detail and a committed date.

D-Wave NYSE: QBTS · Tier 1 │ Near 6.8 Long 7.1 Blended 6.9

● PRIMARY SOURCE — Annealing performance from customer case studies; gate-model roadmap from investor day; dual-rail qubit architecture from technical publications.

D-Wave is the only dual-platform company in the cohort, pairing its established annealing systems with a new gate-model roadmap announced in June 2026. The gate-model plan proceeds through small superconducting systems — seventeen physical qubits in 2026, forty-nine in 2027, one hundred eighty-one in 2028 — toward a ten-logical-qubit fault-tolerant system by 2030 and a hundred-logical-qubit system by 2032. It uses dual-rail qubits and targets an error-suppression factor of ten, well above the factor of roughly two typical of current multi-qubit platforms.

The annealing business is mature — six delivered generations culminating in the Advantage2 system — but the gate-model effort is net-new and unproven for the company, and its hundred-logical-qubit target sits at the far end of the decade.

Key specifications

Distance: gate-model 17Q (2026) → 49 (2027) → 181 (2028) → 10 logical (2030) → 100 logical (2032).
Specs: dual-rail superconducting; targets lambda=10 suppression; Advantage2 annealer in production.
Power: 100 logical qubits running >1M operations targeted 2032.
Track record: six delivered annealing generations; gate-model is net-new and unproven.

Full-stack & ecosystem. Production-grade cloud infrastructure and on-chip cryogenic control are genuine strengths; the dual-platform posture lets it address both optimization workloads today and the gate-model market later.

Verdict. Leads on delivered annealing systems and is the only vendor addressing both annealing and gate-model markets. Its gate-model fault-tolerance targets are distant and unproven.

Watch for: First gate-model system (17Q) operational results — D-Wave's gate-model ambitions are currently roadmap-only; the 2026 17-qubit system will be the first evidence of whether the dual-rail approach performs as projected.

What would change this verdict: This verdict would change if: the 17Q gate-model system demonstrates competitive fidelity; or conversely if gate-model slippage causes the 2030 logical-qubit target to recede.

Neutral-Atom

QuEra Private · Tier 2 │ Near 6.8 Long 7.4 Blended 7.0

● PEER-REVIEWED — Nature (January 2026, peer-reviewed, Bluvstein et al.): d=5 surface code below threshold (2.14× suppression); universal gate set via [[15,1,3]] codes; up to 96 d=4 logical qubits simultaneously in a postselected cluster state (Clifford), the highest simultaneous logical-qubit count published — not 96 independently verified below-threshold qubits. Specs scored 8.5 on the strength of the demonstrated below-threshold and universal-logic results, which are genuine and category-leading; the score is supported by the accurate description of the results, not by the inflated '96 verified' framing. STAR architecture from technical publications; Gemini deployments from customer announcements.

QuEra holds the most significant neutral-atom fault-tolerance results published to date, in a January 2026 Nature paper (Bluvstein et al.) built on reconfigurable arrays of up to 448 atoms. Three distinct results in that paper should not be conflated, because the popular framing tends to merge them. First, below-threshold error correction: a distance-5 surface code showed 2.14× lower logical error per round than distance-3, the defining signature that error correction scales as theory requires. Second, universal logic: arbitrary-angle rotations synthesized via transversal teleportation with three-dimensional [[15,1,3]] codes, demonstrating a fault-tolerant universal gate set. Third, scale: up to 96 distance-4 logical qubits active simultaneously in a cluster-state entanglement structure using high-rate [[16,6,4]] codes, with postselection. The 96-qubit figure is real and is the highest simultaneous logical-qubit count published by any organization, but it is a postselected cluster-state structure on Clifford operations, not 96 independently verified below-threshold qubits — and the report scores it as exactly that, neither more nor less. The architecture is a zoned, transversal design combining qubit shuttling, transversal gates, high-fidelity Rydberg-blockade two-qubit operations, and high-rate qLDPC codes, with rubidium atoms in optical tweezers storing information in exceptionally stable hyperfine ground states.

The company’s stated threshold for commercial advantage is one hundred logical qubits running a million operations before failure, and its roadmap targets one hundred logical qubits from roughly ten thousand physical atoms in the 2026-to-2027 window. It is backed by a partnership with Google and has the most external-customer access hours of any neutral-atom platform.

Key specifications

Distance: 100 logical / ~10,000 physical targeted 2026-27.
Specs: 448 atoms, rubidium, Rydberg gates, zoned/transversal STAR architecture, [[16,6,4]] high-rate codes.
Power: up to 96 d=4 logical qubits simultaneously, postselected cluster state (Nature, Jan 2026); below-threshold result is d=5 surface code at 2.14× — highest simultaneous logical-qubit count published.
Track record: hit aggressive logical-qubit targets; Google-backed; most external access hours in modality.

Full-stack & ecosystem. Hardware-focused; full-stack breadth is not its play. Accessible via the major cloud on-ramps, and partnered rather than vertically integrated.

Verdict. The clear leader on verified logical-qubit count today — the single strongest demonstrated-capability data point in the cohort on that axis. Two structural observations constrain the competitive significance. First: the 96-logical-qubit result is on Clifford operations. Clifford circuits can be simulated efficiently on classical computers; they do not represent universal quantum computation. Universal fault-tolerant computation requires non-Clifford gates — specifically T-gates via magic-state distillation — which QuEra has not demonstrated at scale. The distance from 96 Clifford logical qubits to a machine running Shor’s algorithm or a commercially relevant variational algorithm is not incremental. It is the distance from proof of concept to product. Second: QuEra is a private company with no disclosed revenue, no disclosed revenue model, and AWS-only cloud distribution. A commercial institution cannot access QuEra directly on Azure or GCP. The Research/Government sensitivity profile is the correct lens — for an institution whose primary need is access to the best experimental hardware for QEC research, QuEra’s result is the strongest case for engagement. For a procurement officer or investor focused on 2028 commercial position, the 96-qubit result is a scientific milestone, not a product.

Watch for: The third-generation system targeting 10,000 physical qubits and 100 logical qubits — this is QuEra's next step beyond the 96-logical-qubit result and would validate that the STAR architecture scales as designed.

What would change this verdict: This verdict would change if: the third-generation system demonstrates below-threshold error suppression at higher logical-qubit count than the January 2026 result; or if magic-state distillation for non-Clifford gates is demonstrated at scale.

Atom Computing Private; Microsoft partner · Tier 2 │ Near 6.8 Long 7.3 Blended 7.0

● PRIMARY SOURCE — Magne 50-logical-qubit deployment at QuNorth Denmark (January 2026) from customer announcement; Microsoft collaboration from joint press release; Phoenix specs from company disclosure.

Atom Computing built the first gate-based system to exceed one thousand qubits and continues to scale its optical-tweezer arrays quickly. Its Phoenix system holds roughly eleven hundred eighty atoms; its next-generation Magne system, developed with Microsoft and targeting fifty logical qubits from twelve hundred twenty-five physical qubits, is reported to have been deployed commercially at a customer in Denmark in early 2026. With Microsoft it has demonstrated entanglement of twenty-four logical qubits and a benchmark algorithm on twenty-eight logical qubits.

Atom’s qubits are encoded in nuclear-spin states with coherence times on the order of tens of seconds, and its tweezer architecture offers all-to-all connectivity through atom transport.

Key specifications

Distance: Magne (50 logical / 1,225 physical) deployed at QuNorth Denmark early 2026; Gen 3 (10k physical) 2028.
Specs: ~1,180 atoms, nuclear-spin qubits, ~40s coherence, all-to-all via atom transport.
Power: 24-28 logical qubits demonstrated with Microsoft.
Track record: first 1,000+ qubit gate-based system (2023); credible logical demos. Thinner public disclosure.

Full-stack & ecosystem. Its full-stack story runs through the Microsoft partnership — Microsoft’s error-correction work paired with Atom’s hardware — rather than through owned acquisitions. It is the only private pure-play in the cohort without a registration statement or SPAC disclosure, which is a genuine transparency gap rather than a capability judgment.

Verdict. Strong on scaling speed and on logical-qubit demonstrations through the Microsoft partnership; held back in scoring only by thinner public disclosure.

Watch for: Microsoft's next published result using Atom Computing hardware — the Microsoft collaboration is Atom's primary technical differentiator; the next peer-reviewed output from that partnership will significantly determine Atom's position.

What would change this verdict: This verdict would change if: the Microsoft collaboration produces a peer-reviewed below-threshold result on Atom hardware; or if the collaboration is not renewed or significantly scaled.

Pasqal Private; SPAC · Tier 2 │ Near 6.9 Long 7.2 Blended 7.0

● PRIMARY SOURCE — 140Q deployment at CINECA/Leonardo (February 2026) from company announcement; Fresnel 2 specs from technical disclosure; fidelities from company benchmarks — lag leaders, as stated.

Pasqal is the deployment leader in neutral atoms, with the most systems installed at customer sites, and the only vendor in the cohort offering analog and digital modes on the same hardware. Its current systems run a few hundred rubidium atoms — the Fresnel 2 and Orion Beta lines — in room-temperature racks drawing only a few kilowatts, with none of the cryogenic infrastructure superconducting systems demand. In February 2026 it delivered a one-hundred-forty-qubit system integrated with a supercomputer in Bologna. Its roadmap targets hundreds of logical qubits by 2029.

Two honest caveats: Pasqal’s published two-qubit fidelities lag the leaders, and its earlier ambition of ten thousand qubits by 2026 is visibly slipping against its actual deployment cadence. These are verification-and-pace observations, not judgments of the underlying science, which is strong.

Key specifications

Distance: Vela next; hundreds of logical qubits by 2029.
Specs: ~324 rubidium atoms (Fresnel 2 / Orion Beta), analog + digital modes, room-temp 3kW rack.
Power: delivered quantum-advantage claims in materials simulation; 140Q at CINECA/Leonardo (Feb 2026).
Track record: deployment leader; fidelities lag leaders; 10,000-by-2026 ambition slipping.

Full-stack & ecosystem. Distinctive on sovereign and HPC integration — deployed across European supercomputing centers and accessible through an unusually wide range of clouds, including sovereign European options.

Verdict. Leads on real-world deployments and on data-center-friendly form factor; trails the leaders on fidelity and on roadmap-to-delivery pace.

Watch for: Orion Beta system fidelity results at full scale — Pasqal's deployment leadership is its clearest strength; whether Orion Beta closes the fidelity gap with QuEra and Atom Computing will determine its logical-qubit viability.

What would change this verdict: This verdict would change if: Pasqal publishes independently verified fidelities above 99.5% at full Fresnel 2 / Orion Beta scale; or if the Vela system roadmap to hundreds of logical qubits receives specific technical detail.

Infleqtion NYSE: INFQ · Tier 1 │ Near 7.6 Long 7.7 Blended 7.7

● PEER-REVIEWED — FT operation with 15× error reduction (arXiv:2412.07670, npj Quantum Information, peer-reviewed); sensing revenue from SEC filings; Sqale specs from company disclosure.

Infleqtion is the only public company with commercial leadership across both quantum computing and precision sensing, and the sensing business is as central to it as the computing one — a fact easily missed if it is filed as a computing pure-play. It went public via a merger with Churchill Capital Corp X at a $1.8 billion pre-money valuation, raising more than half a billion dollars in gross proceeds. On the computing side, its room-temperature Sqale platform runs sixteen hundred physical qubits at around 99.7% controlled-Z fidelity and had reached twelve logical qubits by the end of 2025, with a roadmap to thirty logical qubits in 2026 and beyond a thousand by 2030.

On the sensing side it fields atomic clocks (Tiqker), atom-based radio-frequency sensing (Sqywire and the Quantum Spectrum platform), and inertial and gravimetric sensors for environments where satellite positioning is unavailable. Sensing already generates revenue: first-quarter 2026 revenue was $9.5 million, the full-year outlook was raised above forty million dollars, and customers and programs span the U.S. Army, NASA, the Department of Defense, and an energy-research agency, with an atomic clock headed to the International Space Station through a partner.

Key specifications

Distance: 30 logical (2026) → 100+ (2028) → 1,000+ (2030).
Specs: Sqale 1,600 physical, 99.73% CZ, room-temp; plus Tiqker clocks, Sqywire RF, gravimetric sensors.
Power: 12 logical qubits today; sensing already revenue-generating (Q1 2026 $9.5M).
Track record: beat near-term logical target; public via Churchill Capital Corp X ($1.8B pre-money).

Full-stack & ecosystem. Uniquely for the cohort, Infleqtion’s full-stack breadth runs through sensing rather than networking or space. It pairs a computing platform with a revenue-generating precision-sensing business — an earn-now, build-later posture distinct from the pure-play computing companies.

Verdict. The cohort’s clearest dual computing-and-sensing leader, with real sensing revenue today; mid-tier on computing capability alone.

Watch for: Sqale 2 or next-generation computing platform specs — Infleqtion's computing side is commercially earlier-stage than its sensing side; the next hardware generation will determine whether it closes the logical-qubit gap with QuEra and Atom.

What would change this verdict: This verdict would change if: Infleqtion's computing platform reaches 30+ verified logical qubits; or if sensing revenue growth accelerates to the point where the dual-platform thesis is validated commercially rather than just strategically.

Photonic

PsiQuantum Private · Tier 2 │ Near 5.8 Long 7.0 Blended 6.3

● PRIMARY SOURCE — Omega chipset from company disclosure; $1B Series E at $7B from funding announcement; GlobalFoundries partnership from joint press release; architecture peer-reviewed in Nature Communications (2023).

PsiQuantum is the field’s purest bet on jumping straight to fault-tolerant scale. Its model is fusion-based quantum computation — a peer-reviewed scheme that builds fault tolerance from entangling measurements, called fusions, on small constant-size resource states, using primitives natural to photonics rather than the deterministic gates that photonic systems struggle to implement. Its Omega chipset integrates single-photon sources, detectors, and a high-performance optical switch on a low-loss silicon-nitride platform, manufactured through a partnership with an established semiconductor foundry, and it has introduced a compact cooling approach to replace the dilution-refrigerator chandelier.

By design, PsiQuantum produces few intermediate logical-qubit demonstrations — its architecture is close to all-or-nothing at scale — which means it cannot be scored on the same incremental ramp as matter-based platforms. It is building deployment sites with substantial government backing and closed a billion-dollar funding round at a seven-billion-dollar valuation.

Key specifications

Distance: jumps to FT scale; no intermediate logical demos by design; deployment sites under construction.
Specs: Omega chipset (sources, detectors, switch on silicon nitride), GlobalFoundries fab, cuboid cooling.
Power: fusion-based architecture targeting million-qubit FT systems.
Track record: 20+ year lineage; $1B Series E at $7B; component milestones rather than logical-qubit demos.

Full-stack & ecosystem. Manufacturing is the core thesis: CMOS-compatible silicon photonics produced in a commercial foundry. Networking is native to the modality, since photonic qubits are naturally compatible with fiber.

Verdict. Highest-conviction, lowest-interim-visibility profile in the cohort. Scored on component and manufacturing milestones rather than logical-qubit counts, per the modality-adjusted framework.

Watch for: First deployment site becoming operational — PsiQuantum's all-or-nothing architecture means the first system-level result will be more significant than any intermediate milestone; watch for the deployment site announcements to move from 'under construction' to 'operational.'

What would change this verdict: This verdict would change if: a deployment site becomes operational and demonstrates the fusion-based architecture at meaningful scale; or if manufacturing yields at GlobalFoundries prove insufficient for the Omega chipset's loss and component tolerances.

Xanadu Private; SPAC · Tier 2 │ Near 6.9 Long 7.3 Blended 7.0

● PEER-REVIEWED — On-chip GKP qubit generation (Nature, 2026, peer-reviewed); Aurora modular architecture from company disclosure; PennyLane ecosystem metrics from open-source repository.

Xanadu pairs a photonic hardware program with the most widely adopted quantum software framework outside the incumbents. Its error-correction bet is the GKP state, and in 2026 it demonstrated, in a Nature publication, the generation of these error-resistant photonic qubits on an integrated silicon-nitride chip — a genuine milestone, though the company is candid that current GKP quality still falls short of full fault tolerance, with optical loss the dominant limiter. Its Aurora system established the modular, networked architecture; the roadmap targets fault-tolerant operations by 2028 and up to five hundred logical qubits by the end of the decade.

The company operates at room temperature apart from photon detection, runs on commercially available chips, and has secured significant government co-investment and a public-listing path.

Key specifications

Distance: FT operations 2028; up to 500 logical qubits 2029-30; 100k physical by 2029.
Specs: Aurora modular (12 qubits, 35 chips, 13km fiber), GKP states on silicon nitride, room-temp.
Power: on-chip GKP qubit demonstrated (Nature, 2026); optical loss the gating metric.
Track record: X8 → Borealis → Aurora; large PennyLane software base; SPAC listing.

Full-stack & ecosystem. Software reach is a real asset — its open-source framework has a large and fast-growing user base — and its manufacturing partnerships target high-volume photonic production. Networking is native to the photonic modality.

Verdict. Credible photonic roadmap anchored by a peer-reviewed error-correction milestone; optical loss is the gating risk, and the company says so.

Watch for: First demonstration of repeated error correction cycles on the Aurora platform — GKP qubit generation is demonstrated; the next milestone is demonstrating that the GKP qubits can be actively corrected across multiple cycles, which is the gap between a qubit and a useful logical qubit.

What would change this verdict: This verdict would change if: Aurora demonstrates repeated GKP error correction cycles with below-threshold performance; or if optical loss on the silicon-nitride platform improves to the level the fault-tolerance threshold requires.

Beyond the Twelve: Context Players and Cohort Boundaries

The twelve were chosen as the organizations building next-generation systems at meaningful scale across the modalities that have demonstrated a credible fault-tolerance path. Several adjacent players were deliberately held as context rather than scored, because they are either earlier-stage, narrower in modality, or not yet at system scale — but a serious reader should know they exist, because any one could reorder the field.

In silicon-spin, Diraq and Quantum Motion pursue the modality with arguably the strongest long-run manufacturing argument: qubits built with standard semiconductor tools. Diraq, with the research institute imec, demonstrated in a 2025 Nature paper that industrially fabricated silicon quantum-dot qubits on a three-hundred-millimeter line reach two-qubit fidelities around 99.5%, above the surface-code threshold, and can operate above one kelvin — relaxing the extreme cryogenic requirement. Quantum Motion targets a rudimentary fault-tolerant unit cell around 2026. Silicon-spin lags badly on qubit count today, which is why it is context rather than cohort, but its scalability thesis is real.

In bosonic encoding, Alice & Bob bet on cat qubits — directly relevant here, since IonQ's Walking Cat architecture borrows the same cat-state concept for measurement. Alice & Bob demonstrated bit-flip lifetimes between thirty-three and sixty minutes, vastly beyond conventional superconducting qubits, which allows far more efficient error-correcting codes; its roadmap targets an early fault-tolerant machine with one hundred logical qubits by 2030 for materials-science use cases. IQM, a Finnish superconducting builder reporting around 99.9% two-qubit fidelity, is another near-cohort player moving toward a public listing. None of these changes the leadership verdicts, but each is a genuine wildcard, and a report that pretended the field was exactly twelve would be misrepresenting it.

In topological qubits — a fifth modality, distinct from the four scored in this report — Microsoft is the only commercial-scale participant, and its position warrants careful description because it was significantly updated during this report's production. On June 2, 2026, Microsoft published a preprint reporting its Majorana 2 device: by replacing aluminum with lead as the superconductor in its semiconductor-superconductor hybrid nanowires, the team more than doubled the topological excitation gap and measured a parity-switching lifetime of approximately 20 seconds, with some instances reaching minute-scale — a greater than 1,000-fold improvement over the prior generation. On the strength of this, Microsoft revised its scalable-machine target from 2033 to 2029. The materials-science achievement is real and measurable: the lead substitution and the gap improvement are unambiguous, and the rf single-shot parity-measurement technique is a genuine engineering advance. Three facts place Microsoft as context rather than cohort, under the same evidence discipline applied to every other entry. First, a parity lifetime is not a logical qubit: the device demonstrates the stability of a quantum state, not gate operations, an algorithmic-qubit figure, or an error-corrected logical qubit — the metrics on which the cohort is scored. Second, the result is a preprint, not yet peer-reviewed, and therefore sits below the peer-reviewed results that anchor the cohort scores in this report's source hierarchy. Third, the topological-qubit interpretation of Majorana-based devices remains actively contested in the condensed-matter literature — a 2021 Nature paper from the group was retracted after independent review, and several physicists publicly maintained their objections within hours of the Majorana 2 announcement. The disclosure-candor standard this report applies to every vendor requires noting all three plainly, alongside full credit for the materials result. The competitive implication is bounded: a topological machine that delivers a scalable, error-corrected system on the revised 2029 timeline would be a major long-horizon development, and the report's decade-view scenarios should account for it as a low-probability, high-impact variable. It changes none of the near-term verdicts, because near-term leadership is scored on demonstrated logical qubits, delivered systems, and commercial traction — on each of which Microsoft, in topological hardware, currently reports none. Microsoft's full-stack presence is real and separate from this: Azure Quantum is a leading distribution and cloud platform — assessed as such in the companion reports in this series — and the Atom Computing partnership pursues a neutral-atom machine on a parallel track. That full-stack and distribution role is why Microsoft features prominently in the earlier series reports; this report's subject is demonstrated hardware capability, which is where the context placement applies. The author holds a disclosed long position in Microsoft; as with every holding, the placement here follows the evidence rules rather than the position.

Two sovereign and research programs were deliberately removed from the commercial cohort and belong here instead. USTC's Zuchongzhi 3.2 processor — 107 qubits, all-microwave leakage suppression, Λ=1.40(6) on a distance-7 surface code, peer-reviewed in Physical Review Letters in December 2025 — is the first below-threshold QEC result from outside the United States, and it matters as a geopolitical and export-control variable even though USTC is not a company and cannot be procured from. The Chinese national program is real, the result is peer-reviewed, and its implications are addressed in the sovereign section. planqc, a Munich-area spinout pursuing strontium optical-lattice qubits with German national supercomputing contracts, is the European sovereign counterpart: a hundred-qubit machine targeting 2027, a thousand-qubit system for the Leibniz Supercomputing Centre following shortly after. Neither belongs in a ranking of commercial hardware companies; both belong in any honest account of where the field is.

Cross-Cutting Comparison

Demonstrated Error-Correction Status

The table below summarizes where each commercial and emerging player stands on demonstrated error correction, with peer-reviewed results distinguished from vendor and preprint claims. It is the single most important comparison in the report, because it separates what has been shown from what has been promised. The final row places Microsoft's topological program as context: it is the only commercial-scale entrant in a fifth modality, but it has not yet demonstrated a logical qubit, which is why it sits outside the scored cohort.

Company	Architecture / QEC code	Best demonstrated logical-qubit result	Evidence
Google	Surface code, superconducting	Distance-7, 101 physical, ~0.14%/cycle, suppression factor ~2.14, beyond breakeven	Peer-reviewed (Nature)
QuEra	Zoned/transversal, high-rate codes, neutral-atom	d=5 surface code below threshold (2.14×); up to 96 d=4 logical qubits (postselected cluster state, Clifford)	Peer-reviewed (Nature)
Quantinuum	Barium trapped-ion, 2D QCCD	48 error-corrected (and up to 94 error-detected, post-selected) from 98 physical	Peer-reviewed + preprint
Xanadu	GKP states, photonic	On-chip GKP qubit generation; not yet full fault tolerance	Peer-reviewed (Nature)
IonQ	qLDPC (BB5/GB4) + toric + concatenated; OMG/stationary-chain ¹³³Ba⁺	9 QEC codes without hardware reconfiguration. BB5[[18,4,3]]: 2.01×10⁻²/cycle (X), 1.08×10⁻²/cycle (Z) — ~4.3× better on X and ~8.5× better on Z vs. prior superconducting qLDPC (X: 2.01% vs ~8.67%; Z: 1.08% vs ~9.15%). Best lifetime TN=3.95±0.68s vs. physical 3.3±0.9s (marginal, overlapping error bars).	Peer-reviewed (arXiv:2606.06455, June 2026, IonQ-authored)
Atom Computing	Optical-tweezer + Microsoft QEC	24–28 logical qubits (Microsoft collaboration)	Demonstrations + trade press
Infleqtion	Neutral-atom, post-selection	12 logical qubits	Vendor
IBM	Bivariate-bicycle qLDPC, superconducting	Roadmap to 200 logical (Starling, 2029); milestones to date met	Vendor roadmap
Rigetti	Chiplet superconducting	Real-time decoding at small distance (with partner)	Peer-reviewed (memory experiment)
Pasqal	Neutral-atom, analog + digital	Hundreds of physical atoms; logical roadmap to 2029	Vendor
D-Wave	Dual-rail superconducting (gate-model)	Roadmap: 10 logical by 2030, 100 by 2032	Vendor roadmap
PsiQuantum	Fusion-based, photonic	Omega chipset; all-or-nothing-at-scale by design	Peer-reviewed model + vendor
Microsoft (context)	Topological / Majorana tetron (InAs-Pb)	No logical qubit demonstrated; ~20s parity lifetime on a prototype tetron device, >1,000× prior generation. Materials result, not a gate or logical-qubit result.	Preprint (arXiv:2606.03884, June 2026); topological interpretation contested in literature

Figure 2. Logical-Qubit Roadmap Targets, 2026–2033

Vendor-stated targets on a log scale. These are projections, not demonstrated results; IonQ's foundry-accelerated track is the highest-variance claim in the set. The chart shows ambition and shape, not certainty.

Figure 3. Fidelity versus Scale, Current Systems

Best two-qubit gate fidelity against physical-qubit count (log scale) for fielded systems. Trapped-ion leads on fidelity at modest scale; neutral-atom leads on scale at slightly lower fidelity; superconducting sits between. The dashed line marks the approximate surface-code threshold region.

Figure 4. Stated Paths to Fault Tolerance

Vendor target windows for reaching fault-tolerant operation. Bars represent stated roadmaps, not commitments the report endorses; the dashed line marks the present. Read as a map of intentions, with delivery probability assessed in the company profiles.

Full-Stack Posture

The full-stack contest divides into three strategies. The table maps each company to its model and to the pillars it addresses, anchored where possible to standard stack taxonomy and labeled author-constructed where it is not.

Model	Exemplar	How it competes
Owned vertical integration	IonQ	Acquires the whole stack: foundry, networking, security, sensing, space.
Software-platform + partnership	IBM	Owns the software/cloud layer (Qiskit, 250+ network); partners for networking (Cisco).
Cloud-distribution on-ramp	AWS Braket / Azure Quantum	Multi-vendor access layer most companies ride; itself a full-stack consideration.
Sensing-integrated	Infleqtion	Pairs computing with a revenue-generating precision-sensing business.
Sovereign / HPC-integrated	Pasqal	Embeds into national supercomputing and sovereign-capability programs.

Figure 5. Full-Stack Coverage Grid

Coverage across seven pillars: owned/strong (green check), partial or via partner (amber tilde), not yet or not pursued (grey dash). The compute and software columns map to standard stack taxonomy; networking, security, sensing, and space are author-assessed from primary evidence. IonQ's near-complete row reflects its owned vertical-integration strategy; Infleqtion is distinctive on sensing.

Figure 10. Cost & Energy by Modality

Left: estimated bill-of-materials at roughly two-million-qubit scale, on a log axis — IonQ's third-party-validated figure under $30M against superconducting systems above $1B, a gap of more than thirty times that follows from dilution-refrigerator physics rather than engineering maturity. Right: operating power per system, where superconducting platforms draw 25–50 kW against single-digit kilowatts for room-temperature modalities. These are structural penalties on the Cost & Energy axis, not roadmap problems.

A Tale of Two Trapped-Ion QEC Strategies

The June 2026 IonQ paper introduces a comparison that belongs in any rigorous assessment of the trapped-ion modality: there are now two fundamentally different approaches to quantum error correction within the same hardware family, and they make different engineering trade-offs with significant downstream consequences.

Quantinuum's Helios and its successors use a transport-based QCCD (quantum charge-coupled device) architecture, physically shuttling ions between memory and logic zones through a two-dimensional grid with four-way junctions. This is a proven approach — the peer-reviewed Helios hardware paper documents it in detail — but the IonQ paper quantifies its overhead: in transport-heavy trapped-ion systems, transport and cooling consume the majority of execution time, and up to 50% of the ions are used only for cooling. The paper cites Helios's own characterization data for these figures.

IonQ's barium-133 research platform takes the opposite approach: a stationary chain with no ion transport at all, using steerable Raman beams for all-to-all gate connectivity without moving ions. Mid-circuit measurement is achieved via the OMG architecture, which phase-coherently shuttles quantum information between encoding manifolds within a single ion rather than moving the ion itself. Critically, the ancilla qubits measured during error correction also serve as the sympathetic cooling mechanism, eliminating the dedicated coolant-ion overhead entirely.

The engineering consequence is substantial. IonQ's stationary-chain approach uses the efficiency that transport-based systems spend on overhead directly for computation instead. Whether this advantage scales is the open question: the current barium device has 40 ions, while Helios has 98. The IonQ paper's conclusion notes that EQC with increased Raman zone parallelism is the path to larger stationary-chain systems. Both strategies are live experiments in how to build a fault-tolerant trapped-ion computer; neither has yet proven itself at the scale where the trade-offs decisively resolve.

A production-relevance note on the OMG stationary-chain architecture: the efficiency advantage described above is not a research-platform curiosity. It directly improves the duty cycle and resource overhead of any scaled trapped-ion system built on the EQC architecture. A system where ancilla qubits serve simultaneously as cooling ions and error-detection probes — with no dedicated cooling ions and no ion transport — uses a larger fraction of its physical qubits for computation at every moment. As IonQ scales the stationary-chain architecture through EQC and SkyWater’s photonic packaging, this duty-cycle advantage scales with it. The transport-heavy QCCD approach’s overhead compounds as systems grow; the stationary-chain approach’s does not.

For the report's reader: when comparing Quantinuum's and IonQ's error-correction results, these are not just two data points on the same curve. They represent different architectural bets with quantified trade-offs. There is a third dimension of comparison that belongs alongside the architectural one: code efficiency. Quantinuum's fault-tolerant algorithm execution (arXiv:2603.04584) used the Steane [[7,1,3]] code — seven physical qubits per logical qubit, a 7:1 ratio. IonQ's BB5 [[18,4,3]] code encodes four logical qubits in eighteen physical qubits — a 4.5:1 ratio. At identical code distance and similar error rates, IonQ's demonstrated code is meaningfully more efficient: to encode one hundred logical qubits, the Steane approach requires seven hundred physical qubits while the BB5 approach requires four hundred fifty. This is peer-reviewed and quantified from primary sources. Neither code family is proven at fault-tolerant scale, and Quantinuum's Steane result is on a delivered product system while IonQ's qLDPC result is on a research platform. Both caveats hold. The code efficiency comparison nevertheless belongs in any rigorous account of the two trapped-ion strategies.

Vertical & Ecosystem Impact — Where the Machines Actually Bite

The value of a quantum computer is realized in specific workloads at specific scales, and the honest framing ties each vertical to the logical-qubit and operation-count threshold it actually requires — not to vague promise. The thresholds below are drawn from vendor and academic resource estimates and should be read as order-of-magnitude guides, not settled numbers. The table that follows organizes the verticals by the one variable that matters most for a decision-maker: when the capability becomes real. Readers seeking role-specific guidance (investor, enterprise buyer, policy maker) should pair this section with the Decision Framework.

Three Timeline Layers: When Each Capability Becomes Real

Hardware thresholds are anchored to the Algorithmic Qubit (AQ) figures used in the civilizational-impact section for consistency. The final column states the consequence of a 24-month hardware slip — the discipline that separates timeline-sensitive bets from claims that hold regardless of roadmap pace.

Layer	Representative workloads	Advantage mechanism	Hardware threshold	If hardware slips 24 months
Now – 2028	Portfolio optimization, logistics routing, QKD and quantum-safe migration, GPS-denied navigation and sensing	Heuristic optimization (NISQ + annealing); physics-guaranteed security; precision sensing — none requiring fault tolerance	Available on current hardware; AQ64 delivered Sept 2025	Near-zero impact. These capabilities are commercial today; sensing and QKD have no fault-tolerance dependency, and the optimization advantage is already contested rather than timeline-gated.
2028 – 2031	Early fault-tolerant chemistry (quantum-magnetism, Hamiltonian dynamics), materials simulation at moderate scale, first credible chemistry advantage	Quantum simulation on early error-corrected systems; problems that map naturally onto the hardware	~AQ1,000 / hundreds of logical qubits running millions of operations	Shifts the first defensible chemistry/materials advantage from ~2029 to ~2031. Does not change the case for building quantum-literate computational teams now — the 3–5 year lead time means that decision is already late.
2031+	Drug-binding and catalysis from first principles, room-temperature superconductor screening, correlated tail-risk Monte Carlo, carbon-capture catalyst design, cryptographically relevant factoring	Full fault-tolerant simulation and amplitude estimation; unambiguous advantage over classical supercomputers	~AQ10,000 / thousands of logical qubits, networked multi-chip systems	Shifts the transformative-application era from ~2031 to ~2033+. May change the investment case for deep early commitment, but the PQC-migration urgency (a present-tense classical software task) is unaffected.

Sector Detail: Workload-by-Workload

Chemistry & drug discovery

Molecular ground-state energy to chemical accuracy is the canonical target. The standard near-term algorithm (the variational approach) remains unsolved to one-milli-Hartree accuracy industry-wide — IonQ's own benchmark publishes this as an unmet criterion. Useful drug-binding and catalysis simulation is widely estimated to need hundreds to thousands of high-quality logical qubits running millions of operations, placing it firmly in the 2029-and-beyond fault-tolerant era (Layer 2–3). Trapped-ion and neutral-atom fidelity leaders are best positioned on accuracy; superconducting on speed.

Materials science

Simulating quantum-magnetic systems (Hamiltonian dynamics, the Heisenberg model) is the nearest credible advantage and the leading edge of Layer 2. QuEra and Quantinuum have run quantum-magnetism simulations on tens of logical qubits already; IonQ's blueprint estimates a hundred-site Hamiltonian to chemical accuracy as a flagship fault-tolerant target. This is the vertical most likely to show defensible advantage first, because the problems map naturally onto the hardware.

Optimization & finance

Combinatorial optimization (portfolio construction, logistics, routing) is where analog and near-term gate machines compete with classical heuristics today — the core of Layer 1. The honest read is that quantum advantage here is contested and workload-specific; time-to-solution at a fixed quality threshold — the metric IonQ's framework foregrounds — matters more than qubit count. Distinct from this is amplitude-estimation for derivatives pricing and correlated tail-risk Monte Carlo, which are Layer 3 fault-tolerant-era applications on a materially later timeline. Treating ‘finance’ as one vertical obscures that its sub-workloads span two different eras. Neutral-atom analog mode (QuEra, Pasqal) and annealing (D-Wave) target the near-term optimization piece directly.

Cryptography & security

The crypto-break threat is a fault-tolerant-era concern (see the dedicated section), but the harvest-now-decrypt-later risk is present-tense, which is why quantum key distribution and quantum-safe migration are commercial today — the basis of IonQ's security pillar (ID Quantique) and a driver of sovereign programs. Critically, post-quantum cryptography migration is a classical software task with no dependency on quantum hardware timelines: it should proceed now regardless of when a cryptographically relevant machine arrives.

Defense, navigation & space

This is the vertical where sensing, not computing, delivers value now — the clearest Layer 1 capability. Atomic clocks and inertial/gravimetric sensors for GPS-denied navigation are at field-deployment readiness, with active defense procurement — the core of Infleqtion's and IonQ/Vector Atomic's revenue today, and a market growing over twenty percent annually. Space-based quantum key distribution (IonQ/Capella) extends the networking pillar into orbit.

Sovereign & Geopolitical Axis

For an institutional or government buyer, the sovereign dimension is no longer a footnote to a hardware comparison — it is frequently the deciding variable. The clearest signal is that a Chinese state program has crossed the fault-tolerance threshold on a surface code, achieving a logical error suppression factor of Λ=1.40(6) on a distance-7 system — the first result of its kind from outside the United States. This is not at parity with Google's Λ=2.14, which is a meaningfully stronger suppression factor; but it confirms that China has crossed the critical threshold and that the gap is real and narrowing rather than structural and permanent. The all-microwave leakage suppression approach may also offer a scalability advantage. That reshapes the competitive and export-control narrative regardless of any single company's roadmap.

The most consequential analytical error a government can make on quantum is to conflate national-capability building with operational deployment. They are different activities, on different timelines, requiring different policy levers, and relevant to different decisions. A government that conflates them will simultaneously under-invest in the near-term actions that are already urgent and over-invest in procurement of systems that do not yet exist. The three layers below separate them.

Now — procurement and migration. Three government actions are actionable today and have no dependency on the fault-tolerant timeline. First, quantum-sensing procurement: atomic clocks, gravimeters, and GPS-denied navigation systems are field-ready and in active defense procurement now (Infleqtion and IonQ/Vector Atomic both hold programs). Second, post-quantum cryptography migration: the NIST standards were finalized in 2024, and migration is a classical software task that should be underway regardless of when a cryptographically relevant quantum machine arrives — every month of delay accumulates harvest-now-decrypt-later exposure. Third, foundry industrial policy: the CHIPS Act quantum allocations, IonQ's SkyWater acquisition, and IBM's Anderon LOI are the vehicles through which sovereign manufacturing capacity is being established now.

2026–2029 — national capability building. This layer is about establishing competitive position before the fault-tolerant era, not deploying fault-tolerant systems. The levers are trusted-foundry access, sovereign quantum-networking infrastructure, workforce development, export controls on the supply-chain chokepoints, and sustained R&D investment. Europe is pursuing this route through national-champion companies integrated into public supercomputing centers, with deployments across French, German, and Italian facilities and explicit funding for domestically controlled systems. The United States has tied its leading commercial player ever more tightly to government and defense through a trusted-foundry acquisition and a string of national-security contracts. The chokepoints that shape this layer are concentrated: cryogenics and ultra-stable lasers (each with a small number of mostly European suppliers), specialized photonic components, and advanced foundry access. Control of those inputs increasingly shapes competitive position as much as the physics does.

2030+ — operational deployment. Fault-tolerant quantum computing in defense applications, national-scale infrastructure optimization, and cryptographically relevant factoring belong to this layer. The honest label is that this layer is not yet actionable as procurement: the systems do not exist, and a government that attempts to procure them today is buying roadmaps, not capability. What determines whether this era arrives on a competitive timeline is the quality of the capability-building decisions made in the prior two layers. The policy implication is precise: invest now in capability (foundries, networking, workforce, R&D) and in the near-term deployable capabilities (sensing, PQC migration); treat operational fault-tolerant deployment as an R&D and readiness program, not a procurement line, until the hardware crosses the thresholds the vertical section identifies.

Cryptography & the Crypto-Break Gap

No single number answers the question of how large a quantum computer must be to break widely used public-key cryptography, and any report that publishes one without its assumptions is misleading by an order of magnitude or more. The credible estimates span roughly three orders of magnitude, and the spread is driven almost entirely by assumptions — key size, code family, hardware connectivity, and runtime — rather than by disagreement about the underlying physics. Every figure below is therefore tied to a named source and year, and the report deliberately avoids a single headline number.

For elliptic-curve cryptography (ECC-256, the secp256k1 curve underlying most cryptocurrency signatures and much of TLS), the most recent published estimate is Google’s April 2026 ECDLP analysis: on the order of 1,200–1,450 logical qubits (depending on whether the circuit is optimized to minimize qubits or gates), compiling to fewer than 500,000 physical qubits under surface-code error correction with conservative hardware assumptions. For RSA-2048, the historical trajectory is the more instructive figure: a widely cited 2019 analysis (Gidney and Ekerå) put the requirement near 20 million physical qubits; Gidney’s 2025 revision cut that to under one million physical qubits and roughly 1,400 logical qubits — a reduction driven entirely by better algorithms and error-correction assumptions, not by changed hardware. Some neutral-atom and high-rate-qLDPC analyses suggest the potential for substantially lower physical-qubit counts at the cost of longer runtimes, though the most aggressive of these rest on connectivity assumptions not yet demonstrated at scale.

A specific correction to a common claim is warranted, because it bears directly on how a decision-maker should reason about break-ordering. It is frequently asserted that RSA-2048 requires a fixed multiple — often stated as eight times or more — of the resources needed to break ECC-256, and therefore that ECC-based systems fall first by a wide, predictable margin. That framing is outdated. The older estimates did show ECC requiring meaningfully fewer logical qubits than RSA at equivalent classical security (roughly 2.6× fewer in the 2017 Roetteler baseline, comparing ECC-256 against RSA-3072, not RSA-2048). But the aggressive 2025 RSA optimizations narrowed and in some gate-count comparisons inverted that gap. No fixed multiplier holds without specifying the exact key sizes, code family, and assumptions being compared — and a report or commentator who quotes a single multiplier is almost certainly comparing figures from different years or different security levels.

The June 2026 IonQ paper is relevant to the most aggressive estimates. The qLDPC-based FTQC blueprints that drive the lowest physical-qubit projections depend on high-rate codes running on hardware with flexible, non-local gate connectivity. The IonQ barium-133 device has now demonstrated exactly those code classes — BB5 and GB4 — at breakeven, on hardware that achieves all-to-all connectivity without the long-range physical couplers that constrained the prior superconducting qLDPC demonstration. This is not proof that any aggressive crypto-break timeline is correct, and the report does not endorse one; it is evidence that one previously hypothetical assumption — that the required code connectivity can be realized in hardware — is marginally less hypothetical than it was before June 2026. Gate count and circuit depth, not logical-qubit count alone, remain the dominant cost drivers, because they set the total spacetime volume of the computation and therefore the error-correction overhead.

On break-ordering and timelines, the honest framing is about migration speed rather than a precise interval. The sequence in which cryptographic systems become practically vulnerable depends on key-rotation speed, whether public keys are exposed on-chain or in long-lived certificates, and how quickly infrastructure can be upgraded. Systems that cannot rapidly migrate — cryptocurrencies with on-chain exposed public keys and no central key-rotation authority — face materially earlier practical risk than systems that can rotate keys and upgrade quickly, such as banking infrastructure operating behind centrally managed certificate authorities. Framing this as a fixed ‘cryptocurrencies fall, then traditional finance one-to-three years later’ overstates the precision current forecasts can support; the defensible statement is the ordering by migration difficulty, not a dated countdown.

The practical consequence underlying all of this is the harvest-now-decrypt-later threat: data encrypted today and intercepted can be stored against the day a sufficiently capable machine exists. This is why quantum-safe migration and quantum key distribution feature in the security pillar of every serious full-stack strategy, and why a company’s position on the security layer — not just the compute layer — is part of the platform assessment in this report. The gap to a cryptographically relevant machine remains substantial across every roadmap here, but it is closing, and the assumption lattice — named figures with their stated conditions — not a single scary number, is the honest way to track it.

Competitive Leadership: Who Leads, and What It Means

The race to lead the quantum commercialization era is not a race to build the best quantum computer. It is a race to demonstrate a credible, integrated path toward fault tolerance while simultaneously building the complete platform that makes quantum capability commercially deployable at each stage along the way — the hardware, the networking that connects processors across distances, the security layer that protects quantum communications, the sensing capability that generates revenue today, the software that makes the hardware accessible, and the foundry that ensures the hardware can be manufactured at the required speed and volume. A company that wins on hardware alone does not lead. A company that wins on platform breadth alone does not lead. The company that demonstrates both — a credible hardware trajectory toward fault tolerance anchored in a complete commercial platform — sets the pace for the industry and for the critical sectors of the technology economy that depend on it. That is what this section identifies, era by era, because the leader genuinely differs across time horizons and across the definition of leadership. Leadership is defined as an explicit blend of three lenses — proximity to fault-tolerant utility, commercial position by the end of the decade, and long-term defensibility — and the verdict is given for each of three horizons. The June 4, 2026 IonQ paper (arXiv:2606.06455) is incorporated into each verdict, with the specific changes to IonQ’s position stated explicitly and bounded precisely.

Figure 6. Leadership Is Horizon-Dependent

The leader changes with the time horizon and the definition of leadership. Today's verdict rests on demonstrated, peer-reviewed capability. The near-term contest is real on hardware capability but resolved on platform position — IonQ leads the seven-axis commercial framework by 1.2+ points across every sensitivity profile. The decade verdict depends on which constraint proves binding. Named leaders are defended, with falsifiers, in the leadership section.

How arXiv:2606.06455 Changes IonQ’s Leadership Position

What definitively changed. Before June 4, IonQ had no peer-reviewed QEC result at any scale. Its demonstrated-capability position rested on gate fidelity (99.99%, a world record but on a 2-qubit demonstrator) and theoretical blueprints. After June 4, IonQ has a peer-reviewed experimental result: qLDPC breakeven on a 40-ion barium-133 research platform. That is a step-change in category — from blueprint-only to demonstrated-experimental-QEC. The category change is real and material.

What did not change. Four constraints remain. First: the device is a 40-ion barium research platform, not the EQC/ytterbium 256-qubit product system. Any inference that the product platform now has demonstrated QEC is unsupported. Second: breakeven on qLDPC and below-threshold exponential suppression on surface codes are genuinely different milestones. Google demonstrated that logical error rates suppress exponentially with code distance — the defining signature that fault tolerance scales as theory requires. IonQ demonstrated that a logical qubit lifetime can match or marginally exceed the underlying physical qubit lifetime. Both are necessary steps; they are not the same step. Third: the ∼4.3×(X)/∼8.5×(Z) improvement is against one prior superconducting qLDPC demonstration on one code class, not against the field’s best QEC results. Fourth: scale — IonQ’s best results encode 2–4 logical qubits; QuEra has demonstrated up to 96 d=4 logical qubits simultaneously (postselected cluster state), Quantinuum 48 error-corrected, and Atom Computing 24–28, all substantially more than IonQ on raw logical-qubit count.

The axis-by-axis verdict. On demonstrated-QEC: IonQ moves from absent to present. On code-efficiency: IonQ’s result is most significant here — qLDPC encoding rates (roughly 4.5:1 physical-to-logical overhead) are substantially better than surface codes (hundreds-to-one at useful error rates). If qLDPC scaling works as theory predicts, this is the more important long-run path, and IonQ demonstrated it experimentally first among trapped-ion platforms. On hardware-flexibility: nine codes without hardware reconfiguration is genuinely competitive with anyone. On logical-qubit count: IonQ does not lead. On below-threshold scaling: IonQ does not lead — Google demonstrated the scaling signature; IonQ demonstrated breakeven. On theory-to-experiment coherence: Walking Cat → arXiv:2606.06455 → EQC product line is the tightest documented theory-to-product progression in the cohort.

The net effect. IonQ retires the most damaging gap in its competitive narrative without moving to the top of any current-snapshot ranking. The near-term probability weights shift modestly toward IonQ because one explicit high-variance risk — no peer-reviewed QEC result — has been retired. The characterization of IonQ’s 2028 acceleration targets as the highest-variance bet in the cohort remains correct; the evidentiary base supporting the bet is now marginally stronger. The decade-horizon position strengthens more than the near-term, because the decade verdict hinges on which QEC code family wins at scale, and IonQ has now established an early experimental lead on qLDPC — the path most FTQC architectures, including Walking Cat and Pinnacle, are betting on.

Today — Demonstrated Capability (2026)

The leaders on demonstrated, peer-reviewed error correction are Google, QuEra, and — as of June 4, 2026 — IonQ. Google holds the strongest surface-code result (distance-7, Λ=2.14, 2.4× beyond breakeven, Nature). QuEra holds the verified logical-qubit record (96, below threshold, Nature, January 2026). IonQ has now published a peer-reviewed qLDPC breakeven result on a 40-ion barium-133 research platform: ~4.3× better on X errors and ~8.5× better on Z errors than the prior best superconducting qLDPC demonstration; best combined lifetime TN=3.95±0.68s vs. physical 3.3±0.9s (marginal, overlapping error bars). Quantinuum remains close behind on delivered logical-qubit operations.

A preemptive response to the most frequent institutional challenge: QuEra’s peer-reviewed neutral-atom results (Nature, January 2026) — below-threshold error correction on a distance-5 surface code at 2.14× suppression, a universal fault-tolerant gate set, and up to 96 d=4 logical qubits in a postselected cluster state — and Google’s distance-7 below-threshold surface-code result (Λ=2.14, Nature, 2024) are important scientific milestones. However, they were achieved on specialized research hardware that has no direct path to commercial deployment. QuEra does not disclose revenue, does not have a product roadmap to a delivered commercial system, and distributes through AWS only. Google has no commercial quantum product, no enterprise distribution path, and no revenue from quantum. IonQ’s June 2026 result — which is on a research platform, and the report is explicit about that caveat — demonstrates multiple QEC codes, superior code efficiency, and no-hardware-reconfiguration flexibility on a platform whose control electronics (EQC) and foundry path (SkyWater) are already being productized and sold. The first 256-qubit EQC systems are contracted to three named commercial and academic customers. For procurement and scaling timelines, demonstrated engineering practicality, delivery cadence, and full-stack platform position matter as much as headline logical-qubit counts achieved in research settings. A reader whose decision criterion is ‘who can I actually buy a system from today and expect delivery next year’ is not choosing between IonQ, QuEra, and Google on a single axis.

Four analytical precisions. First: breakeven on qLDPC differs from below-threshold surface-code suppression — the three results are not on the same scale. Second: IonQ’s QEC device is the 40-ion barium-133 research platform, not the EQC/ytterbium product system. Third: the ∼4.3×(X)/∼8.5×(Z) improvement is against one specific prior superconducting qLDPC demonstration on one code class. Fourth: the paper is IonQ-authored; the report applies vendor-result caveats symmetrically.

Near-Term — Through 2028

A precise distinction is required here because the report’s prior language collapsed two different questions into one contested verdict. The near-term ‘contest’ is real on one dimension and resolved on another, and conflating them produces a misleading picture.

On demonstrated hardware capability today — peer-reviewed error-corrected logical-qubit counts, below-threshold suppression signatures, and demonstrated fault-tolerant algorithm execution — the race is genuinely open. Google holds the strongest surface-code suppression result (distance-7, Λ=2.14, 2.4× beyond breakeven). QuEra holds the verified logical-qubit record (96 below threshold, Nature 2026). Quantinuum holds the first peer-reviewed fault-tolerant algorithm execution (12 logical qubits, Helios). IonQ has now added peer-reviewed qLDPC breakeven — the first on trapped-ion hardware — with ∼4.3× better X-basis and ∼8.5× better Z-basis logical error rates versus the prior best superconducting qLDPC demonstration. On this specific dimension, the contest is real and the rankings are defensible for each leader.

On commercial platform position through 2028 — scored across seven axes including full-stack integration, commercial traction, foundry access, networking infrastructure, and sovereign program depth — the contest is not close. IonQ leads IBM by 1.23 points near-term and leads every other competitor by a larger margin. That lead does not invert under any of the three sensitivity profiles. IBM leads if capital depth and software ecosystem are the only axes that matter; IonQ leads on every other combination. Quantinuum, despite its strong QEC results, scores 2.32 points below IonQ on near-term platform position — a gap driven by the absence of networking, foundry, cloud distribution breadth, and multi-year financial accountability. Saying ‘the near-term is contested’ without this distinction misleads a reader who needs to make a procurement or investment decision in the near term, where platform position is what determines vendor viability, not just hardware capability.

The report’s primary view: IonQ is the near-term platform leader. IBM is the near-term hardware-roadmap credibility leader among superconducting companies. Google and QuEra are the near-term demonstrated-capability leaders on the specific axis of peer-reviewed logical-qubit count and error suppression. All three of these statements are simultaneously true, and a reader whose decision criterion matches one of them should weight it accordingly. The sensitivity analysis on the following page shows exactly where the rankings invert and where they do not.

The Decade — Through 2030 to 2032

The paper strengthens IonQ’s decade-horizon position more than its near-term position, because the decade verdict hinges on which QEC code family wins at scale. If qLDPC codes outperform surface codes as theory predicts — and the June 4 result is the first experimental evidence of qLDPC breakeven on trapped-ion hardware — IonQ has established an early experimental lead on the code path most FTQC architectures are betting on. IBM’s Starling roadmap also uses qLDPC (bivariate-bicycle codes), so this is not winner-take-all; it is an early data point on a trajectory both IBM and IonQ are pursuing through different hardware. If manufacturing and foundry access is the binding constraint, IonQ and PsiQuantum lead. If capital and R&D depth is decisive, IBM leads. The report’s primary view is that manufacturing access and capital are the likeliest binding constraints, favoring IBM and IonQ; the June 4 paper shifts the uncertainty slightly in IonQ’s favor on the modality-physics axis without resolving it.

One decade-horizon wildcard sits outside the four scored modalities: Microsoft’s topological program. Its June 2026 Majorana 2 preprint reported a roughly 20-second parity lifetime and a revised 2029 target for a scalable machine (see the context section for the full treatment). If topological qubits deliver on that timeline, the modality’s theoretical promise — hardware-level error protection that could reduce the QEC overhead every other approach carries — would make it a serious decade-horizon competitor. That outcome should be treated as a low-probability, high-impact variable rather than a base-case assumption, for the reasons the context section documents: no logical qubit has yet been demonstrated, the result is a preprint, and the topological interpretation remains contested in the literature. It does not affect any near-term verdict, where leadership is scored on demonstrated logical qubits, delivered systems, and commercial traction. The disciplined posture is to track it as a wildcard with a specific falsifier — a peer-reviewed demonstration of a topological logical qubit with gate operations — and to revise the decade view if and when that evidence appears.

The Compounding Advantage: Why the Distance Widens

The scoring framework in this report captures where each company stands today. It does not model what happens to the competitive distance between IonQ and its peers if IonQ’s specific milestone sequence delivers. That dynamic is analytically distinct from the point-in-time scores, and it is the most important forward-looking analysis in this report, because the mechanism is not linear. Each successful step in the sequence does not merely advance IonQ’s position by one increment — it simultaneously makes the alternative path longer for every challenger. Understanding why requires tracing the five specific mechanisms through which the advantage compounds.

Milestone sequence and feedback loop: the five mechanisms below are not independent — each successful step enables the next and makes the alternative path longer for challengers. The sequence: (1) SkyWater regulatory close [Q2–3 2026] → chip iteration drops from 9 months to 2 months; (2) First integrated 256-qubit EQC chip data [H2 2026] → confirms product fidelity, enables second and third system deliveries; (3) University of Chicago, University of Cambridge, and Horizon Quantum systems operational [2026–27] → customer lock-in, workflow development, follow-on procurement conversations begin; (4) DARPA HARQ chip-to-chip interconnect demonstration [2026–27] → validates networked multi-chip architecture before any competitor ships AQ100 standalone; (5) AQ1,000 on networked system [2027–28] → crosses industrial optimization and pharmaceutical simulation thresholds; (6) 200,000-qubit chip functional testing [2028] → with SkyWater foundry, arrives two years ahead of original schedule; (7) AQ10,000 on networked multi-chip system [2028–30] → categorical threshold for drug discovery, materials science, and financial risk modeling. The 2028 inflection is the report’s primary focus: it is the year the compounding mechanisms either validate or fail, and the Watch For lines in the IonQ company profile identify the specific events that will confirm or disconfirm each step. Peer competitors missing at least three of these mechanism layers: IBM has no live networking, no sensing revenue, and an LOI-only foundry. Quantinuum has no networking, no foundry, and Azure-only cloud distribution. Google has no commercial product and no networking. None of this is unsolvable in principle; the compounding concern is about timeline and the moat each delivered system creates.

The first mechanism is foundry iteration speed. Before SkyWater, IonQ’s design-to-first-samples cycle was nine months. After SkyWater closes, SEC Form 425 filings confirm it becomes two months — a 4.5× acceleration. This means IonQ can run more hardware generations in a given calendar year than any competitor without owned foundry access. Each additional wafer spin generates experimental data that improves the next chip. A competitor attempting to close the gap must either acquire similar foundry access — now more expensive because IonQ has established the template and demonstrated demand — or accept a slower iteration cadence. IBM’s Anderon is an LOI. Quantinuum has no disclosed foundry strategy. The iteration speed advantage is structural and widens with each wafer cycle completed.

The second mechanism is customer lock-in through delivered systems. The first 256-qubit chip-based system was sold to the University of Cambridge in Q1 2026. That customer is now building workflows, training researchers, generating publications, and entering into follow-on procurement discussions — all before any competitor has shipped a comparable system. A procurement officer who built their organization’s quantum computing capability on IonQ’s sixth-generation system faces a switching cost that is not just financial but operational: the staff expertise, the circuit libraries, the integration architecture, and the institutional memory are all IonQ-specific. When the seventh-generation system arrives, the renewal decision is not the same as the original procurement decision. It is almost always an upgrade. This dynamic is well-understood in enterprise software; it applies with equal force to quantum hardware.

The third mechanism is government contract depth. DARPA HARQ, MDA SHIELD IDIQ, SDA HALO $39 million, and the SkyWater DoD-trusted foundry position are not four independent contracts. They are four nodes in a network of relationships, security clearances, program familiarity, and institutional trust that takes years to build and cannot be replicated by writing a larger check. Each contract makes the next one easier to win, because government program managers prefer proven vendors and because the technical integration work done on one program creates transferable capability for the next. A competitor that has no current defense program positions must first establish entry-level contracts before competing for the programs IonQ is already executing. That is typically a multi-year process.

The fourth mechanism is the networking moat. IonQ has deployed quantum key distribution networks in Switzerland, Romania, and Florida (MSA signed April 2026), and holds the SDA HALO $39 million contract for tactical space communications. Quantinuum has no quantum networking program. IBM has a Cisco partnership at the classical networking layer and published papers on quantum interconnects — not the same as deployed infrastructure. Google has no commercial networking position. Each operational network node generates operational data, engineering experience, and customer relationships that improve the next deployment. The Skyloom optical terminal technology, the Qubitekk entanglement distribution know-how, and the Lightsynq photonic interconnect capability are not just capabilities — they are components of a networking platform that will be required to connect quantum computing chips into multi-chip systems. A company that has already deployed networking infrastructure internationally is not starting from the same point as one that is reading papers about quantum interconnects.

The fifth and most significant mechanism is the non-linearity of AQ thresholds. Algorithmic qubits measure effective computational power at a fixed fidelity and circuit depth. AQ64 was delivered three months early. The 256-qubit EQC system, if its integrated gate fidelity approaches the prototype’s 99.99%, would represent AQ on the order of 120–160 depending on circuit depth — already beyond any demonstrated competitor position. If IonQ then demonstrates chip-to-chip networking, connecting two 256-qubit systems via quantum interconnects as DARPA HARQ is specifically designed to support, the logical-qubit count addressable by the combined system scales super-linearly with fidelity. AQ1,000 on a networked system — a plausible near-term milestone — is not a quantitative improvement on AQ64. It crosses several application thresholds simultaneously: optimization problems at realistic industrial scale, quantum chemistry simulations for pharmaceutical relevance, and machine learning acceleration that classical hardware cannot match. AQ10,000 on a networked multi-chip system, if demonstrated before any competitor ships AQ100 as a standalone, is a categorical lead. Replicating it requires not just better hardware but a complete foundry, networking, and system-integration stack that takes years to assemble. The compounding is not in the numbers; it is in the infrastructure required to generate them.

The honest caveat. Every compounding mechanism described above depends on sequential milestone delivery. If the SkyWater foundry integration proves harder than the two-month design-to-samples estimate, the iteration speed advantage narrows. If the integrated 256-qubit EQC chip shows significantly lower fidelity than the two-qubit prototype, the customer lock-in story weakens because the systems being delivered are less capable than the benchmark suggests. If DARPA HARQ’s quantum interconnect program produces results that are not transferable to IonQ’s commercial hardware, the networking moat does not materialize. The compounding is real if the milestones land. It is contingent, not guaranteed. The Watch For lines in the IonQ company profile identify the specific events that will confirm or disconfirm each mechanism in the next eighteen months. Those are the most important things to track.

Figure 7. The Commercial Gap

Estimated quarters each company is behind IonQ's current ability to sell a 256-qubit-class system to a paying customer, derived from owned-foundry status, count of delivered commercial systems, and cloud-distribution breadth. This is an analyst estimate, not a vendor figure, and is the most concrete representation of the compounding advantage: the gap is measured in delivery position, not in physics. Companies with no commercial revenue (grey) are furthest because they have no delivery motion to accelerate.

What the Lead Actually Means

Each verdict translates into consequence. For procurement timing, a buyer whose workloads are optimization-shaped today may rationally engage a vendor that does not lead on error correction, because time-to-solution on near-term problems is a different test than long-run fault tolerance. For ecosystem risk, committing to a proprietary stack carries different lock-in than building on the dominant open software layer. For sovereign exposure, vendor choice is implicitly a choice about supply-chain and export-control risk. And for the investment lens — stated as scenario analysis, not advice — the divergence between today’s demonstrated-capability leaders and the likely commercial-platform leaders is itself the most important strategic fact in the sector.

One final precision on IonQ: the report holds disclosed long positions in IonQ among others. The leadership assessment above was constructed from the physics and the paper before consulting price action or analyst commentary. Readers are invited to apply independent skepticism to the code-efficiency and theory-to-experiment arguments in particular — those are the two axes where a COI-motivated analyst would be most tempted to overclaim. The report’s view is that those claims are defensible from primary sources, and the invitation to check stands.

Weighted Scorecard: All Twelve Companies

Scores are on a 1–10 scale, computed within tier only — Tier 1 and Tier 2 companies are not directly comparable on a single list. Seven axes. Near-term horizon: 2026–2028 (weights: Track Record 27%, Distance 22%, Specs 15%, Projected Power 7%, Cost/Energy 4%, Commercial Platform Completeness 15%, Full-Stack Integration 10%). Long-term horizon: 2029–2032 (weights: Track Record 17%, Distance 22%, Specs 17%, Projected Power 21%, Cost/Energy 8%, CPC 10%, FSI 5%). Blended = 60% near-term + 40% long-term. Full-Stack Integration (FSI) captures owned versus partnered coverage across compute, networking, security, sensing, space, cloud, and software — drawn from Figure 5. Its inclusion in the blended score corrects the prior framework's structural omission: a company with IonQ's owned networking deployments, sensing revenue, and foundry position was being scored identically to one with none of those, on the hardware axes alone. With FSI integrated, the blended score reflects competitive position across the full platform, not just the hardware stack.

Tier 1 — Commercial / Revenue-Bearing

Company	Dim scores (TR/D/SP/PP/CE/CPC/FSI)	Near	Long	Blended	Full-stack
IonQ	9.3 / 8.5 / 9.5 / 9.5 / 9 / 9.5 / 9.5	9.2	9.3	9.3	9.5
Quantinuum	7 / 6.5 / 8.5 / 7.5 / 6.5 / 5.5 / 6.5	6.9	7.2	7.0	6.5
IBM	9 / 7 / 8 / 8.5 / 6.5 / 7.5 / 8	7.9	7.9	7.9	8.0
Google Quantum AI	8.5 / 6.5 / 8.5 / 8 / 6 / 4.5 / 6	7.1	7.5	7.2	6.0
Rigetti	7 / 6.5 / 7.5 / 6.5 / 8.5 / 6 / 5.5	6.7	7.0	6.8	5.5
D-Wave	8.5 / 6 / 6 / 6.5 / 8 / 7 / 5	6.8	7.1	6.9	5.0
Infleqtion	8 / 7 / 7.5 / 7.5 / 7.5 / 7.5 / 8.5	7.6	7.7	7.7	8.5

Figure 8. Seven-Axis Scoring Profiles

The near-term seven-axis scores for the four Tier 1 leaders, plotted as overlapping profiles. IonQ's navy envelope contains every competitor on every axis; the widest gaps are on Commercial Platform Completeness and Cost & Energy. Google's profile is the most uneven — strong on Specs, weakest in the cohort on Commercial Platform — the visual signature of a research-optimized rather than commercially-optimized program. Scores are reproduced exactly from the Weighted Scorecard.

Tier 2 — Emerging / Pre-Commercial

Company	Dim scores (TR/D/SP/PP/CE/CPC/FSI)	Near	Long	Blended	Full-stack
QuEra	7 / 7 / 8.5 / 7.5 / 7.5 / 5 / 5	6.8	7.4	7.0	5.0
Atom Computing	7.5 / 7 / 7.5 / 7.5 / 7.5 / 5 / 5.5	6.8	7.3	7.0	5.5
Pasqal	7.5 / 6.5 / 7 / 7 / 8.5 / 6.5 / 5.5	6.9	7.2	7.0	5.5
PsiQuantum	6 / 5.5 / 6 / 9 / 7.5 / 4 / 5.5	5.8	7.0	6.3	5.5
Xanadu	7 / 6 / 7 / 7.5 / 8 / 7 / 7	6.9	7.3	7.0	7.0

Scoring note. FSI = Full-Stack Integration. CPC = Commercial Platform Completeness. The integration premium applied in the Full-Stack Part III report is not carried here; scores are computed strictly from the weighted averages. A reader who weights commercial delivery differently should apply the sensitivity analysis profiles on the following page.

Sensitivity Analysis: Rankings Across Weight Profiles

The weighted scores below show how rankings shift under three different weight profiles. Where a company's rank is stable across all three profiles, the verdict is robust to weighting assumptions. Where it shifts significantly, the reader should apply their own weight preference. Companies are shown in near-term score order under the baseline weights.

Company (near-term order)	Baseline (27/22/15/7/4/15/10)	Enterprise / Procurement	Research / Government	Investment / Long-horizon
IonQ	9.2	9.2	9.3	9.2
IBM	7.9	8.0	8.0	8.0
Infleqtion	7.6	7.6	7.6	7.6
Google Quantum AI	7.1	7.1	7.4	7.3
Pasqal	6.9	6.9	6.9	6.9
Quantinuum	6.9	6.9	7.2	7.0
Xanadu	6.9	6.8	6.9	7.0
D-Wave	6.8	6.9	6.5	6.7
Atom Computing	6.8	6.9	7.0	7.0
QuEra	6.8	6.8	7.2	6.9
Rigetti	6.7	6.7	6.8	6.7
PsiQuantum	5.8	5.8	6.1	6.5

How to read this. Stable rankings across all three profiles indicate conclusions that are robust to weighting assumptions. Significant rank shifts indicate where your own weight preference matters. IonQ leads every weight profile. Full-Stack Integration (FSI) is now integrated into the blended score at 10% near-term weight — a reader who weights it differently should apply the profiles above and recompute.

Decision Framework: Translating the Evidence into Action

The verdicts in this report are not recommendations. They are structured assessments of what the evidence currently supports and what would change it. The framework below translates those assessments into a decision structure for three types of readers. Use your own weight profile — the sensitivity analysis shows where the ranking changes and where it does not.

Scenario Analysis: Three Futures and What They Imply

The report's central finding — IonQ as platform leader — rests on a set of milestones landing. Rather than bury that contingency in prose, the table below states three scenarios explicitly, with a subjective probability weight (the author's estimate, offered for transparency and challenge, not as fact) and the implication for the competitive ranking under each. The probabilities sum to 100% and represent the author's current read; a reader who disagrees should substitute their own and re-derive the implications.

Scenario	Prob.	What happens	Implication for the ranking
IonQ delivers on schedule	~55%	SkyWater closes in 2026; first integrated 256-qubit EQC chip data confirms product-scale fidelity near the prototype's 99.99%; chip-to-chip networking demonstrated on the DARPA HARQ timeline; 2028 200k-qubit functional-test milestone holds.	The 1.2-point lead over IBM widens through the compounding mechanisms. IonQ's platform leadership becomes a capability lead as well, and the gap to every competitor grows. This is the base case the report's primary verdict assumes.
IonQ slips 12–18 months	~30%	SkyWater integration proves harder than the two-month design-to-samples estimate; the integrated EQC chip shows materially lower fidelity than the prototype; the 2028 milestone moves to 2029–30. Networking and revenue continue but the hardware acceleration stalls.	The platform lead survives on Track Record, CPC, and FSI — the commercial axes do not depend on the 2028 milestone. But the Specs and Projected Power advantage narrows, and IBM's roadmap-credibility case strengthens relative to IonQ. The lead compresses from 1.2 to roughly 0.5–0.8 points. IonQ still leads; the margin is no longer decisive.
A competitor closes the foundry gap	~15%	IBM converts Anderon from LOI to operational foundry before IonQ demonstrates product-scale EQC fidelity; or a neutral-atom or photonic competitor secures comparable owned-fabrication access with disclosed terms. Capital depth begins to compound for a challenger.	The foundry advantage — the single most important structural pillar of IonQ's distance score — erodes. If IBM is the challenger, its capital depth ($10B+) becomes the binding advantage and the decade verdict could invert. This is the scenario the report's IBM falsifier (Kookaburra qLDPC memory before IonQ's integrated chip data) is designed to track.

For the Investor

The most important structural fact in quantum hardware for an investor right now is the divergence between the demonstrated-capability ranking and the likely commercial-platform ranking. The demonstrated-capability leaders today are Google (strongest surface-code suppression), QuEra (highest simultaneous logical-qubit count, plus below-threshold and universal-logic results), and IonQ (qLDPC breakeven, June 2026). The probable commercial-platform leaders through 2030 are IonQ (full-stack integration, foundry access pending SkyWater close) and IBM (capital depth, open-foundry, Starling roadmap). These are not the same list. A portfolio positioned exclusively on demonstrated-capability leaders will miss the commercial-platform story; a portfolio positioned exclusively on commercial leaders will miss the modality-physics story.

The two highest-conviction positions in the evidence are: IonQ's theory-to-experiment-to-product coherence is the tightest documented in the field, and it has now closed the most damaging gap in its narrative (no peer-reviewed QEC result). Quantinuum's milestone delivery record is the cleanest in the field, and its IPO registration statement makes its disclosures unusually accountable. The highest-variance position is IonQ's 2028 acceleration claim (8,000 logical qubits); the report treats this as a forward claim, not a delivered milestone, and so should you.

The falsifiers that would most change the investment case are: SkyWater acquisition close and first integrated EQC chip data (IonQ); Sol delivery on schedule (Quantinuum); Kookaburra qLDPC memory demonstration (IBM). These three events in 2026–2027 will do more to determine the decade-long competitive landscape than any paper published between now and then.

For the Enterprise Buyer

If your workload is optimization-shaped and your timeline is now through 2027, D-Wave's annealing platform and QuEra's neutral-atom analog mode are the commercially available options with the deepest track records on that problem class. If your workload is chemistry or materials simulation and your timeline is 2028–2029, the evidence points to Quantinuum (delivered error-corrected logical operations, cleanest delivery record) and IonQ (qLDPC code efficiency if it scales) as the systems most likely to be relevant at that point. If your workload is security (QKD, quantum-safe migration), the decision is platform-independent: the commercially available products are available today through IonQ's ID Quantique stake and several independent vendors.

The single most important procurement decision variable right now is cloud access, because it determines whether you need a capital commitment or just a service agreement. Every Tier 1 company in this cohort is accessible via AWS, Azure, and/or GCP. The second most important variable is lock-in: proprietary software stacks (Quantinuum's Guppy, IonQ's application-centric framework) create switching costs that matter more as workloads deepen. The companies with the most open software ecosystems — IBM (Qiskit), Xanadu (PennyLane) — have the lowest lock-in risk.

If you are at the 'explore and pilot' stage, the Watch For lines in each company profile tell you the one upcoming event per vendor that most determines whether that vendor's capability will be where you need it when you need it. Read those before committing to a pilot vendor.

For the Policy Maker

The sovereign dimension of quantum hardware is no longer secondary. A Chinese state program has crossed the fault-tolerance threshold. Six of the twelve companies in this cohort received CHIPS Act letters of intent totaling $2 billion — D-Wave, Rigetti, Infleqtion, Atom Computing, PsiQuantum, and Quantinuum each receiving $100 million. IBM's Anderon foundry — America's first pure-play quantum chip foundry — is being funded partly with public money and is explicitly designed to be modality-agnostic, meaning it will serve trapped-ion, photonic, and superconducting companies. IonQ's pending SkyWater acquisition, if it closes, will create the first quantum company with an owned DoD-trusted semiconductor foundry.

The supply-chain chokepoints that matter most are: dilution refrigerators (superconducting platforms; two dominant suppliers, both European); optical tweezers and laser systems (neutral-atom platforms; M Squared Lasers in Glasgow is the leading supplier); and ion-trap chip fabrication (trapped-ion; IonQ's SkyWater acquisition is directly aimed at this). Export controls on any of these supply chains would have asymmetric effects depending on modality.

The procurement question for governments is not which hardware is best but which supply chains are acceptable. A system that performs slightly worse but is manufactured entirely within treaty-aligned jurisdictions may be the correct procurement choice regardless of the hardware ranking.

What These Advances Mean for the World

This report has been careful throughout to separate what is demonstrated from what is claimed, to apply skepticism symmetrically, and to flag every forward projection as a projection rather than a fact. That discipline is not suspended here. But it would be a failure of analytical completeness to describe twelve companies racing toward fault-tolerant quantum computers without stating what fault-tolerant quantum computers actually mean for human civilization — not as hype, but as a precise account of what changes when the physics works at scale.

Drug discovery is the application most frequently cited and most frequently oversold. The precise version of the claim is this: molecular ground-state energy calculations for molecules larger than roughly fifty atoms are intractable for classical computers, not because of software limitations but because of exponential state-space scaling. A fault-tolerant quantum computer with hundreds to thousands of high-quality logical qubits running millions of operations would, for the first time in history, make it possible to simulate protein folding, drug-receptor binding, and catalytic reaction pathways from quantum mechanical first principles rather than from approximations. The consequence is not faster drug discovery in the current paradigm; it is a different paradigm — one in which a candidate molecule can be evaluated against its physical reality before a single milligram is synthesized. The pharmaceutical industry spends more than $2 trillion annually on research and development, with the majority consumed by candidates that fail in clinical trials for reasons that better computational modeling might have caught earlier. The economic and human stakes of getting this right are not marginal.

Materials science carries the same structure. Room-temperature superconductors, nitrogen fixation catalysts that operate without the Haber-Bosch process’s seven hundred degrees and two hundred atmospheres, and battery chemistries that outperform lithium-ion by an order of magnitude are all problems that reduce to quantum mechanical simulation of electron behavior in complex molecules. They have resisted classical computation for the same reason drug discovery has — exponential state space — and they have the same potential to unlock step-changes rather than incremental improvements. A catalyst that fixes atmospheric nitrogen at ambient conditions would reduce global energy consumption by roughly two percent, because the Haber-Bosch process currently consumes that share of the world’s energy supply to produce fertilizer. That is not an incremental gain.

Financial risk modeling is different in character but similar in consequence. The problem is not intractability in the quantum-computing sense; it is that the state space of correlated financial instruments under tail-risk scenarios is large enough that classical Monte Carlo methods require impractical sample counts to price it accurately. Portfolio optimization under realistic correlation structures, real-time pricing of complex derivatives, and sovereign debt crisis modeling all involve optimization over spaces that quantum algorithms — specifically quantum amplitude estimation and quantum optimization — are theoretically positioned to address with polynomial rather than exponential resource scaling. The consequence for financial stability, risk management, and insurance pricing is material if the theoretical advantage translates to practice at scale, which is not yet demonstrated but is not implausible.

Cryptography is the one application where the timeline is not speculative but present-tense and urgent. The harvest-now-decrypt-later threat is real and active: encrypted government and commercial communications are being collected today against the day a sufficiently capable machine arrives. The National Institute of Standards and Technology finalized its first post-quantum cryptographic standards in 2024. Every institution that has not begun migration is accumulating retroactive vulnerability. The quantum hardware advances in this report are not the cause of this threat — the theoretical vulnerability of RSA and elliptic-curve cryptography to Shor’s algorithm has been known since 1994 — but they make the timeline for a cryptographically relevant machine progressively less speculative. The IonQ qLDPC breakeven result on June 4, 2026, is not itself a crypto threat; it is experimental evidence that the hardware connectivity assumptions underlying the most aggressive resource-estimation work are achievable. That shifts a probability, not a certainty, and the appropriate response is accelerated migration, not panic.

Climate and energy modeling represent a longer-horizon application but potentially the highest-stakes one. The molecular-simulation capability that applies to drug discovery and materials science applies equally to the design of carbon-capture catalysts, artificial photosynthesis pathways, and next-generation solar cell materials. Grid optimization under renewable intermittency is a combinatorial problem amenable to quantum optimization. Climate modeling at regional resolution requires fluid-dynamics simulations that quantum algorithms may eventually accelerate. None of these applications are within reach on 2026 hardware, and the honest position is that they require fault-tolerant systems with thousands to tens of thousands of logical qubits — a decade away under optimistic assumptions. But the companies on these pages are building the infrastructure that makes them possible, and the decisions made about which platforms to fund and develop in the next three years will determine whether the ten-year horizon is reachable.

National security and geopolitical stability are the dimension where the advances documented in this report carry the most immediate weight. Quantum sensing is already operational: atomic clocks providing GPS-independent navigation, gravimeters for subsurface mapping, and quantum radar concepts for low-observable detection are in active procurement and field deployment. Quantum key distribution provides physics-guaranteed communication security for government and critical infrastructure. The competitive dimension — China crossing the fault-tolerance threshold, the U.S. CHIPS Act quantum investment, the foundry strategies of IonQ and IBM — is not a technology story but a sovereignty story. The country whose companies control the foundry, the supply chain, and the software ecosystem for fault-tolerant quantum computing will hold a structural advantage in signals intelligence, cryptographic security, and precision sensing that compounds over decades. That is why this report has a geopolitical section, and why it gives foundry position and sovereign supply-chain exposure explicit weight in the scoring framework.

The threshold at which these applications become tractable has a name in IonQ’s framework: Algorithmic Qubits. AQ measures effective computational power at a fixed fidelity and circuit depth. AQ64 was delivered in September 2025, three months early. The Walking Cat blueprint and the SkyWater roadmap project toward 200,000 physical qubits enabling approximately 8,000 high-fidelity logical qubits by 2028. If chip-to-chip quantum networking — the focus of DARPA HARQ, in which IonQ is a selected participant — connects multiple 256-qubit systems before competitors have shipped standalone systems above AQ100, the networked system’s algorithmic power is super-linear: logical qubits distributed across networked chips can encode more complex algorithms than the same physical qubits in isolation. AQ1,000 on a networked system crosses the threshold for optimization problems at realistic industrial scale and quantum chemistry simulations relevant to specific pharmaceutical targets. AQ10,000 crosses the threshold at which quantum advantage over classical supercomputers becomes unambiguous: protein folding from first principles, room-temperature superconductor candidate screening, correlated financial risk under tail scenarios that classical Monte Carlo cannot resolve at the required sample count. These thresholds are directional, not precise — they depend on circuit structure, error rates, and classical alternative quality — but the direction is clear. The question is not whether these thresholds exist; it is when they are crossed, and which customers are already on the platform when they are.

The honest synthesis is that quantum computing’s civilizational impact is not a question of whether but of when and who. The physics is settled. The error-correction threshold has been crossed. The architectural blueprints have been published. The question is which companies, which platforms, and which national supply chains will carry the technology from the laboratory to the hospital, the data center, the financial system, and the battlefield. That is what this report is actually about.

Figure 9. The Algorithmic-Qubit Ladder

Each Algorithmic Qubit threshold mapped to the applications it unlocks and the year IonQ's roadmap projects reaching it. The rungs are not evenly spaced in difficulty: each step crosses multiple application thresholds at once, which is why the report treats AQ progression as non-linear rather than incremental. Thresholds are directional and depend on circuit structure, error rates, and the quality of the classical alternative.

The Verdict: Who Leads, Why It Matters, and What Comes Next

The evidence accumulated in this report resolves to a conclusion that can be stated directly, and then immediately complicated in the most important way possible. A methodological note first: all score comparisons in this section are within-tier. IonQ is Tier 1 (commercial, revenue-bearing); all companies named below are Tier 1. Leadership is assessed within the commercial cohort; the margins cited are not cross-tier comparisons. IonQ is the platform leader in quantum hardware as of June 2026. That verdict rests not on a single paper, a single contract, or a single financial metric, but on the convergence of evidence across seven independently scored axes applied symmetrically to twelve companies. It leads on Track Record by a margin that reflects five years of NYSE-listed public accountability against competitors with one year or none. It leads on Distance because it is the only company with a foundry acquisition in final regulatory close, live quantum networking deployments on three continents, and a fault-tolerant architectural blueprint validated by peer-reviewed experimental evidence. It leads on Specs because 99.99% two-qubit gate fidelity is a world record and the qLDPC code efficiency advantage of 4.5:1 physical-to-logical is demonstrably better than the 7:1 achieved on alternative trapped-ion approaches. It leads on Projected Power because Walking Cat is the most detailed public FTQC engineering specification in existence, and the June 4, 2026 paper on barium-133 is the first experimental evidence that the hardware connectivity its architecture requires is achievable. It leads on Commercial Platform Completeness because three hyperscaler cloud channels, $130 million in FY2025 GAAP revenue growing at 755% year-on-year, $470 million in RPOs, 1,200-plus quantum-dedicated patents, and simultaneous positions in DARPA HARQ, MDA SHIELD, and SDA HALO compose a commercial position no other quantum company currently matches. The blended score of 9.2 across all seven axes, leading IBM by 1.2 points and Quantinuum by 2.3, is not a rounding error. It is what the primary-source evidence says when it is read carefully and applied without favoritism. The static gap understates the potential competitive distance of 2028–2030: the compounding mechanisms described in the Competitive Leadership section — foundry iteration speed, customer lock-in, government contract depth, networking moat, and the non-linearity of AQ thresholds — each widen the gap with each delivered milestone rather than narrowing it. If IonQ’s 256-qubit system delivers at the prototype’s fidelity, and chip-to-chip networking is demonstrated on the DARPA HARQ timeline, and the SkyWater foundry produces the two-month design-to-samples cycle promised in SEC Form 425 filings, the company that competes with IonQ in 2028 faces a more difficult problem than the company that competes with it today.

IBM is the strongest institutional competitor and the company whose position most warrants watching. Its near-term score of 7.9 reflects a delivery record that is genuinely credible — every stated roadmap milestone met to date, the $10 billion-plus commitment to quantum, the Kookaburra bivariate-bicycle qLDPC implementation on track for 2026, and the Qiskit ecosystem that remains the most widely deployed quantum software framework in the world with more than two hundred fifty thousand member organizations. The Anderon letter of intent — America’s first pure-play quantum chip foundry — is a meaningful strategic move even at the LOI stage, and the Tour de Gross architecture with ninety-percent reduction in physical-qubit overhead is a technically serious result. IBM’s Distance and Cost/Energy scores are held back by structural physics facts: twenty-five to fifty kilowatts per dilution refrigerator, helium-3 supply chain dependency, specialist cryogenic facility requirements, and a bill-of-materials at scale that exceeds IonQ’s by a factor of thirty. Those are not roadmap problems. They are physics problems. The Distance score of 7.0 and Cost/Energy score of 6.5 are not judgments against IBM’s engineering team; they are honest accounting of what superconducting systems require at scale. IBM’s long-term position is strongest on software ecosystem and capital depth, and its most important near-term milestone — Kookaburra delivering a demonstrated qLDPC memory result — would materially strengthen its case.

Infleqtion occupies a position that this report believes is systematically undervalued by a market that still files it as a quantum computing pure-play rather than what it actually is: the only public company with simultaneous commercial leadership in both quantum computing and precision sensing, with real revenue in both categories today. Its Q1 2026 revenue of $9.5 million against a forty-million-dollar full-year outlook is not a rounding error on a research budget; it is a commercial business generating income from quantum sensors deployed in U.S. Army and NASA programs, with an atomic clock headed to the International Space Station. The Sqale computing platform at 99.7% CZ fidelity and sixteen hundred physical qubits is mid-tier on the computing axis — the report scores it honestly there. But the sensing business is a hedge, a revenue floor, and a strategic moat simultaneously: it funds the computing roadmap, it diversifies against the risk that useful quantum computing applications take longer than projected, and it establishes Infleqtion in every defense and national-security conversation that matters for quantum hardware procurement. The full-stack score of 8.5 — second highest in the entire cohort, ahead of IBM at 8.0 — reflects what the sensing-plus-computing integration actually represents. The company worth watching most carefully after IonQ is this one.

Google Quantum AI holds the strongest single result in the field — the Willow surface-code suppression factor of Λ=2.14, peer-reviewed in Nature, is the clearest experimental demonstration that error correction scales as theory predicts — and has the weakest commercial position of any Tier 1 company. Its near-term score of 7.2 reflects exactly that tension: the Specs score of 8.5 is high because the physics is genuinely excellent, and the Distance score of 6.5, Cost/Energy score of 6.0, and Commercial Platform Completeness score of 4.5 are low because there is no commercial product, no enterprise distribution path, no revenue, and no networking program. Google’s competitive importance is that it holds the demonstrated benchmark that defines the threshold every other company is measured against. Its competitive limitation is that it has chosen to optimize for research output rather than commercial delivery, and that choice has a real cost in the dimensions that determine which companies will carry quantum computing into everyday use. A Google pivot to commercial deployment — an enterprise access program, a direct sales motion, a partnership with a system integrator — would materially change the near-term picture.

Quantinuum’s strongest claim is peer-reviewed and specific: the first published execution of a fault-tolerant algorithm on real hardware, in collaboration with JPMorgan Chase, using twelve logical qubits encoded in ninety-seven physical qubits on the Helios system. That is a genuine first, and its Specs score of 8.5 reflects it. The QCCD architecture is well-characterized and the Helios hardware paper is among the most transparent vendor publications in the cohort. The limitations are also specific: one year of audited public financial disclosures, $30.9 million in FY2025 revenue growing at thirty-four percent, customer concentration disclosed in the S-1, no quantum networking program of any kind, and a cloud distribution limited to Azure and its own portal. The Distance score of 6.5 reflects the distance between Quantinuum’s current position and full useful delivery — not a judgment on the quality of the Helios result, which is excellent, but an honest accounting of the foundry gap, the networking absence, and the Apollo 2029 timeline that sits one year behind IonQ’s SkyWater-accelerated 2028 milestone. For a reader whose primary criterion is demonstrated error-corrected logical operations on real hardware today, Quantinuum is the strongest choice. For a reader whose criterion is platform position through 2030, the score reflects a different answer.

QuEra’s January 2026 Nature results, on reconfigurable arrays of up to 448 atoms, represent the most significant neutral-atom fault-tolerance demonstration to date: below-threshold error correction on a distance-5 surface code (2.14× suppression), a universal fault-tolerant gate set via [[15,1,3]] codes, and up to 96 distance-4 logical qubits active simultaneously in a postselected cluster state. The 96-qubit figure is the highest simultaneous logical-qubit count published by any organization and represents a genuine architectural achievement. The limitations are specific and should be stated alongside the credit: the 96-qubit structure is on Clifford operations in a postselected cluster state, not 96 independently verified below-threshold qubits, and non-Clifford gate demonstration at that scale — required for universal fault-tolerant computation — has not been published. QuEra’s Track Record score reflects one hardware milestone from a private company with no disclosed revenue, and its Commercial Platform Completeness score of 5.0 reflects AWS-only cloud distribution and the absence of a commercial revenue model. The Research/Government sensitivity profile gives QuEra its highest score of 7.4, which is the correct lens: for an institution whose primary need is access to the best experimental hardware for QEC research, QuEra’s peer-reviewed results are the strongest case for engagement.

Rigetti, D-Wave, Atom Computing, Pasqal, PsiQuantum, and Xanadu each hold a defensible position on at least one dimension. Rigetti’s real-time decoding result and chiplet architecture are technically interesting; its challenge is delivery execution and a revenue base that, while real, has historically slipped against timeline commitments. D-Wave has the most operationally deployed quantum hardware in the world on the annealing side and a gate-model pivot that is early but genuine; for optimization workloads with timelines through 2027, its track record of customer deployments is stronger than any competitor’s. Atom Computing’s Microsoft-partnered twenty-four to twenty-eight logical qubit demonstration is the clearest second data point for neutral-atom error correction after QuEra, and its 99.9-percent fidelity position is credible. Pasqal leads the cohort on European sovereign integration, with hardware deployed across French, German, and Italian national supercomputing centers and an access path that no other vendor in the cohort currently matches for EU procurement. PsiQuantum is the cohort’s highest-variance bet: if fusion-based photonic quantum computation at scale works as designed, its resource efficiency at million-qubit counts would be transformative; if the manufacturing yields or the fusion-gate thresholds prove harder than projected, it has no intermediate hardware to point to. Xanadu’s PennyLane framework with more than forty thousand GitHub stars and a thousand peer-reviewed citations is the most-used open-source quantum machine learning library in the world, a software moat that its hardware position does not yet match.

The field is moving faster in 2026 than it has at any prior point, and the competitive distance is widening rather than narrowing. Four significant hardware results have been published in the first six months of the year. The U.S. government has committed more capital to quantum hardware in the last twelve months than in the prior five years combined. The first six-generation chip-based quantum system has been sold and delivered. SkyWater’s stockholder vote is done. The question of who leads is no longer unanswerable — it is answered in this report, with the evidence laid out so that the reader can verify or challenge every step. The question of what happens next is the one worth watching, and the falsifiers in each company profile tell you exactly which events will matter most in the next eighteen months. That is the point of building a report like this rather than a press-release summary: not to tell you what to think, but to give you the tools to think it through yourself — on the evidence, with the caveats visible, and with the outcome uncertain enough to stay interesting.

Conclusion: The Series Verdict on IonQ’s Position

This report is the fourth and final installment in the Quantum Technology Integration Series, and its conclusion is best understood in the context of the three reports that precede it. Across the series — the hardware frontier, the software adoption framework, the full-stack platform analysis, and now the forward-looking systems-and-integration report — one finding has recurred under four different analytical lenses, each with its own methodology, its own evidence base, and its own scoring framework. IonQ occupies the strongest integrated position in the field. That conclusion was reached independently in each report; it was not assumed and carried forward. The full-stack analysis scored IonQ highest on platform completeness. The software analysis identified IonQ’s full-stack integration as structurally difficult for competitors to replicate. This report’s seven-axis hardware-and-platform framework places IonQ first on the blended score by 1.2 points over IBM and by larger margins over every other competitor, under every sensitivity profile. When four distinct methodologies converge on the same conclusion, the conclusion is more robust than any single report could establish alone.

The short-term advantage is grounded in what has already been demonstrated and delivered, not in roadmap promises. IonQ holds the world-record two-qubit gate fidelity at 99.99%. It published, on June 4, 2026, the first peer-reviewed qLDPC breakeven on trapped-ion hardware — the most code-efficient error-correction path demonstrated to date — which places it among the small group of organizations with genuine peer-reviewed error-correction results, alongside Google, QuEra, and Quantinuum. It is the first public quantum company above $100 million in annual GAAP revenue, growing at 755% year-on-year, with $470 million in remaining performance obligations and $3.1 billion in cash. It has sold three sixth-generation 256-qubit systems to named customers with delivery expected across Q4 2026 and Q1 2027. It is the only company in the cohort with a deployed quantum networking business, a revenue-generating sensing line, a quantum security product, and a pending owned-foundry position. No competitor matches that combination today. The short-term advantage is a platform advantage, and it is already commercially visible.

The long-term advantage is grounded in the compounding mechanics this report models in detail. The SkyWater foundry acquisition — stockholder-approved, awaiting only regulatory close — compresses chip iteration from nine months to two months per cycle, which means more hardware generations per year than any competitor without owned fabrication can achieve. The qLDPC code-efficiency advantage means fewer physical qubits per logical qubit at fault-tolerant scale than the surface-code or Steane-code paths competitors are pursuing. The networking moat means each deployed node compounds the next. The customer relationships established with the current generation create lock-in before competitors ship comparable systems. These mechanisms do not add; they multiply. The 1.2-point lead measured today understates the competitive distance of 2028–2030 if IonQ’s milestone sequence delivers — and the report is explicit about the scenario in which it does not, where the lead compresses but does not disappear.

This conclusion is offered with its caveats visible, as the entire series has been. IonQ’s forward claims are the highest-variance in the cohort; the SkyWater-accelerated 2028 timeline is aggressive; the June 4 breakeven was demonstrated on a research platform with overlapping error bars; and the author holds a disclosed long position in IonQ, which is why every score in this report is recomputable by the reader and every competitor’s genuine strengths are credited before its limitations. The field is genuinely contested on individual demonstrated-hardware dimensions, and the report says so plainly. But on the question the series was built to answer — which company is best positioned to lead the quantum commercialization era by integrating hardware progress into a complete platform — the evidence across four reports and four methodologies points to one answer with unusual consistency. IonQ has both the strongest near-term commercial position and, if its milestone sequence delivers, the strongest long-term structural advantage in the field. That is the verdict of the series.

Risk Factors & Methodological Notes

Every forward-looking figure in this report is a claim until demonstrated, and the report's knowledge baseline is treated as stale by default: every time-sensitive fact — qubit counts, fidelities, dates, valuations, deals, and contracts — was re-verified against primary sources, and several earlier assumptions were corrected in the process, including a vendor's metric methodology, a company's listing structure, the classification of one company as a dual computing-and-sensing business, and the existence of an architectural blueprint absent from the baseline entirely.

Each leadership verdict is accompanied by its falsifier — the specific development that would overturn it — so that the reader can track the field against named triggers rather than vibes. The aggressive resource claims, particularly the most accelerated fault-tolerance timelines, are scored as forward claims rather than delivered milestones, and the report is explicit about which results have cleared peer review and which have not.

The two thinnest evidence points, flagged for transparency, are the absence of independent benchmark data for the German sovereign program and the reliance on demonstrations and trade press, rather than a peer-reviewed specification, for one neutral-atom system's next-generation figures. Neither gap is papered over.

Selected Sources & Evidence Base

This report prioritizes primary and peer-reviewed sources. The following are the principal evidence anchors; vendor roadmaps and trade press were used for current figures and are distinguished as such in the text.

Google Quantum AI et al., below-threshold surface-code memory, Nature (2024) — distance-7, 0.143%/cycle, suppression factor 2.14.
QuEra/Harvard/MIT (Bluvstein et al.), fault-tolerant neutral-atom architecture: d=5 surface code below threshold (2.14×), universal gate set via [[15,1,3]] codes, up to 96 d=4 logical qubits in postselected cluster state, from up to 448 atoms. Nature 649, 39–46 (2026).
Quantinuum, Helios 98-qubit trapped-ion system, peer-reviewed hardware paper (arXiv 2511.05465); March 2026 logical-qubit results.
Quantinuum + JPMorgan Chase, fault-tolerant execution of error-corrected quantum algorithms — QAOA on 12 logical / 97 physical qubits using Steane [[7,1,3]] code on Helios; HHL algorithm for Poisson equation instance; first peer-reviewed end-to-end FT algorithm execution on real hardware. arXiv:2603.04584 (March 2026).
IBM, Tour de Gross — bivariate bicycle (gross) codes enabling fault-tolerant quantum computation with ~90% fewer physical qubits versus surface codes; modular FTQC architecture supporting Starling (2029) and Blue Jay. arXiv:2506.03094 (2026).
Oxford Ionics (an IonQ company) — Hughes, Srinivas, Löschnauer, Knaack, Matt, Ballance, Malinowski, Harty, Sutherland: 'Trapped-ion two-qubit gates with >99.99% fidelity without ground-state cooling.' The 'smooth gate' — residual spin-motion entanglement errors adiabatically eliminated by detuning ramp; gate error 8.4(7)×10⁻⁵ without ground-state cooling; operates above the Doppler limit. Primary source for the 99.99% world-record fidelity claim. arXiv:2510.17286 (October 20, 2025).
Oxford Ionics — Löschnauer, Mosca Toba, Hughes, King, Weber, Srinivas, Matt, Nourshargh, Allcock, Ballance, Matthiesen, Malinowski, Harty: 'Scalable, high-fidelity all-electronic control of trapped-ion qubits.' Foundational EQC architecture paper — presents vision for electronically controlled trapped-ion quantum computer addressing scale-performance tradeoff. arXiv:2407.07694 (July 2024).
IBM, fault-tolerant roadmap to Starling (2029) and Blue Jay; $10B+ commitment (June 2026).
USTC, Zuchongzhi 3.2 below-threshold error correction, Physical Review Letters (Dec 2025).
Xanadu, on-chip GKP qubit generation, Nature (2026); Aurora modular architecture, Nature (2025).
PsiQuantum, fusion-based quantum computation, Nature Communications (2023); Omega chipset (2025).
Diraq / imec, industrially fabricated silicon spin qubits above threshold, Nature (2025).
Alice & Bob, hour-scale cat-qubit bit-flip times (Sept 2025).
Microsoft Quantum (Aghaee et al., 164 authors), '20 Second Parity Lifetime in an InAs–Pb Tetron Device' — lead-for-aluminum superconductor substitution more than doubling the topological gap; ~20s (characteristic ~22s) interferometric single-shot parity lifetime; rf measurement technique resolving wire-end states to µeV precision. Preprint, not yet peer-reviewed; topological-qubit interpretation contested in the literature. arXiv:2606.03884 (June 2, 2026).
IonQ (Tham, Goldman, Debnath, Delfosse et al.), breakeven demonstration of quantum low-density parity-check codes; nine QEC codes without hardware reconfiguration (BB5, GB4, toric, concatenated); OMG architecture for mid-circuit measurement without ion transport; BB5[[18,4,3]] achieves 2.01×10⁻²/cycle (X), 1.08×10⁻²/cycle (Z) logical error rate — ~4.3×(X)/~8.5×(Z) better than prior superconducting qLDPC (X: 2.01% vs ~8.67%; Z: 1.08% vs ~9.15%); best lifetime TN=3.95±0.68s (GB[[26,2,5]]) vs. physical 3.3±0.9s; EQC identified as scalability path; arXiv:2606.06455 (June 4, 2026) — the most recent primary source cited in this report.
QED-C application-oriented benchmarks (Lubinski et al.); Sandia Quantum Performance Laboratory benchmarking literature.

Quantum Tech Integration

Translate

Sunday, June 7, 2026

The Quantum Frontier Report: The Race to Lead, the Compounding Advantage, and the Investor Imperative

No comments:

Post a Comment

The Quantum Frontier Report: The Race to Lead, the Compounding Advantage, and the Investor Imperative

Contact Form

Search This Blog

About Me

Popular Posts

Blog Archive

Blog Archive