China’s Quantum Push: What Beijing Is Assembling in Plain Sight

 Q U A N T U M T E C H N O L O G Y I N T E G R A T I O N S E R I E S ·

CHINA FRONTIER TECH: AI AND QUANTUM · REPORT 3 OF 3 · 

FRIENDS OF IONQ 

How America, our allies and NATO respond to China's significant advances in quantum technology may be the defining technology decision of the century

Special Note from The Friends of IonQ: We believe the material contained in this research report is perhaps the most important information we have ever shared. The data strongly supports a more expansive China quantum program that is expanding into new horizons.  Our position is that Congress must expand all quantum funding programs, that allied nations must expand their own quantum national efforts, and NATO must utilize their increased budget marked for infrastructure to utilize quantum technology.  Our position is that leading full quantum platforms - specifically IONQ - is best positioned to answer the challenges in a geopolitical quantum theater that has rapidly expanded. 

4 US-blacklisted bodies inside Beijing’s quantum-AI consortium, including its lead	<0.4% calculated share of a trapped-ion error-correction cycle used by the link to NVIDIA’s GPUs	0.32 vs 0.24 quantum against best classical, radar change detection on IonQ hardware
C R I T I C A L N O T I C E T O A L L S T A K E H O L D E R S The assessments most decision-makers rely on no longer reflect China’s quantum capabilities — and those outdated reports undermine American, allied and NATO security. [FACT] China’s quantum program is substantially further along than the major published assessments US and allied officials rely on describe — and further along than Beijing’s own headline statements show. This analysis documents both from China’s own procurement, standards and government records, the primary sources those assessments did not use, backed by a 342-link source register covering 247 cited sources, drawn from a review of hundreds more. [FACT] China’s quantum technology activity, capability, intent and execution have all moved substantially beyond what is described in the public assessments most often cited by congressional leaders, administration officials and allied stakeholders. Those assessments include the U.S.-China Economic and Security Review Commission staff report and Quantum First recommendation (November 2025); the CSIS assessments (October 2025 and January 2026); the SIPRI primer (July 2025); the Department of Defense China Military Power Reports (December 2024 and December 2025); the Defense Intelligence Agency’s 2025 threat assessment; the 2026 Annual Threat Assessment of the U.S. Intelligence Community; NATO’s Joint Air Power Competence Centre study (September 2026); the German Marshall Fund’s analysis (May 2026); the Commission’s 2025 Annual Report to Congress (November 2025); and RAND’s testimony to the Commission (February 2024), still widely cited. The change is documented in Chinese and US government sources. [S214, S215, S179, S216, S181, S218, S180, S219, S150, S151, S223, S222, S213, S200, S198, S206, S210, S217] [FACT] Used today, those assessments omit or understate developments recorded in Chinese and US government sources — the integration center, the satellite procurement, the standards rosters, the sensing trials. Funding, export-control, force-posture, standards and alliance decisions calibrated to them rest on a baseline China has already passed, and that undermines the security of the United States, its allies and NATO. [FACT] Recent independent and allied sources — ASPI (December 2025), MERICS (2026) and NATO’s Science & Technology Trends 2025–2045 — corroborate the direction of change. The statements this report examined are set out, assessment by assessment, in the then-and-now table. As of September 2026, Congress is reauthorizing the National Quantum Initiative, so the baseline it uses matters for that reauthorization. [S225, S224, S226, S227] Activity. [FACT] Beijing launched a full-stack quantum integration center benchmarked against

 

Why this matters to you

• Quantum computing’s first real payoff will come inside AI data centers, and whoever sets how the two connect sets the platform. That choice is being made in 2026 and 2027. [S41, S30, S49]
• China is racing toward that point on the ground and in orbit, run by institutions the United States already restricts, and it discloses less than it builds. [S24, S12, S14, S20, S21, S23]
• IonQ holds the most concrete American position: a US-fabricated QPU on NVIDIA’s AI rack, and the first proof that its acquired companies are producing results together. [S30, S31, S38]

What you will find only here

• The only assessment built on China’s own records — standards-drafting rosters, provincial procurement tenders, municipal notices and the Pentagon’s Mandarin 1260H list, sources the USCC, CSIS, SIPRI, DoD and ODNI assessments did not use (“The Baseline Has Moved”).

[S198], [S206], [S200], [S201], [S213], [S29]

• A statement-by-statement audit of the assessments Washington and NATO rely on, with a then-and-now table showing what each said and what the record now shows (“The Baseline Has Moved”). [S214], [S215], [S216], [S219], [S222]
• The integration layer China is building — a China Telecom-controlled center Beijing benchmarks, in its own words, against IBM Qiskit and NVIDIA NVQLink — and four US-blacklisted bodies in Beijing’s quantum-AI consortium, including its lead (Section 1). [S213], [S196], [S20], [S21], [S23]
• PLA universities named as drafters of China’s national quantum-computing and QKD standards (Section 1). [S198], [S206]
• China’s next quantum satellites, visible in procurement before any launch is announced, and a district policy that writes defense, drone swarms and reconnaissance into quantum demonstrations (Sections 2 and 9). [S200], [S201], [S18]

• Current to September 27, 2026 — published in the week of the Xi–Trump summit, the new U.S.–China Super Intelligence Dialogue and National Quantum Initiative reauthorization bills pending in Congress (About This Report). [S242], [S243], [S247]

Fully traceable — every claim carries an evidence tag and a source number that resolves in a 342-link source register covering 247 cited sources, drawn from a review of hundreds more (Appendix B and the Link Register).

Its limits stated openly — what open sources cannot show, listed as collection requirements, and the conditions that would change this view (Section 9 and the Honest Concessions).

Why trapped ions pair naturally with NVIDIA, and an American company already fielding both quantum-key and post-quantum security (Sections 12 and 15). [S42, S105]

How to Read This Report

Reading paths. 5 minutes: the Critical Notice and the bottom line. 20 minutes: the Executive Summary and “The Baseline Has Moved.” Full read: Parts One and Two, the Honest Concessions and the appendices. Two parts. Part One maps China and names no company; Part Two is the IonQ case. Tracing a source. Any [S#] resolves in Appendix B and in the 342-link Link Register.

Every analytical sentence carries one of the Series’ evidence tiers. Every INFER and ARG statement closes with its basis — the numbered sources in Appendix B ([S1] to [S247]) and, where relevant, the section of this report from which the position is drawn — so that a reader can trace each inference or argument to its origin.

ALL REFERENCES ARE AVAILABLE UPON REQUEST.

Tier	Meaning	Worked example from this report
[FACT]	Documented in a primary or near-primary source.	A CETC research institute leads Beijing’s Quantum-AI Collaborative Innovation Consortium.
[CO-STATED]	A company’s or official’s own statement, not independently confirmed.	Qingxing Yigou says it has been an AMD AI system supplier since 2023.
[INFER]	A reasoned conclusion drawn from facts cited alongside it.	Quantum-inspired AI falls outside most quantum control categories.
[ARG]	The authors’ argument or recommendation.	Federal quantum-AI work should run on a US quantum-GPU reference architecture.
[UNDISC]	Information not publicly disclosed.	Board seats and information rights attached to the CETC fund’s investment.

 

Terms. A glossary is in Appendix C. “Quantum-inspired” (量子启发) means classical software that borrows mathematics from quantum physics and runs without a quantum computer. “Quantum-AI” covers both quantum-inspired methods and real quantum processors inside AI workflows. ¥100 million is written 1 亿元 in Chinese sources; figures are approximate as disclosed. “Department of Defense” is used throughout.

Contents

About This Report 6

Executive Summary 8

Three Critical Issues 13

Report at a Glance 14

The Baseline Has Moved 16

Context: Two Days in September 26

Post-Summit Note 28

Part One — China’s Quantum-AI Lane 30

1. Who Leads the Lane 30
2. What the District Policy Says 34
3. The Method Beneath the Controls 34
4. Quantum Processors Inside AI 35
5. The Distribution Layer 36
6. Research Scale, Capital and Control 39
7. US-Origin Touchpoints and Talent 43
8. The Military and Dual-Use Lane 45
9. Orbit and Counterspace 50
10. Part One Actions 58

Part Two — The American Answer: IonQ and NVIDIA 61

What This Means for IonQ 61

1. The Announcement and the Stack 62
2. The Timing Budget 63
3. Three Weeks of Quantum-AI Evidence 64
4. The Radar Result 65
5. The Operational Case and Allies 66
6. The Tests Ahead, and What IonQ Should Do Next 70
7. Part Two Actions 72
8. Honest Concessions 73
9. What Would Change This View 73
10. Conclusion — Closing the Series 74

Appendix A. Confirmed US Designations 75

Appendix B. Primary-Source Log 76

Appendix C. Glossary 85

Appendix D. Leaked China–Russia Forum Documents 86

Disclosure 87

About This Report

The real issue. [FACT] China’s quantum program is substantially further along than the major published assessments US and allied officials rely on describe — and further along than Beijing’s own headline statements show. This analysis documents both from China’s own procurement, standards and government records, the primary sources those assessments did not use, backed by a 342-link source register covering 247 cited sources, drawn from a review of hundreds more.

The final report in a three-part series. Report One: “An IonQ Solution for the AI Force — and the China Quantum Stakes Behind the Xi Summit” (QTIS Part IV, including The Conversion Machine). Report Two: “Land, Sea, Air and Space: IonQ’s Full-Stack Bid to Keep America’s Quantum Lead over China” (QTIS Part V).

At the White House, September 24–25. [FACT] On September 24 the President wrote that “super intelligence” would be a big topic and that he wanted to leave it “exactly where it is.” Both leaders raised AI; Xi stressed human control and offered more direct flights and exchanges for 100,000 American students. The trade truce runs to January 10. The visit ended on September 25; Xi also visited the National Archives and promised a pair of pandas to Atlanta. Basis: [S87].

[ARG] The summit’s steps lower tension without changing the competition this report describes.

Basis: [S3], [S4], [S5].

[FACT] Two days earlier, at the UN General Assembly, the President said that whoever wins superintelligence wins, and that the United States is leading China by a lot. Basis: [S59].

What the summit produced, September 25–27. [FACT] A White House fact sheet issued late on September 25 established the U.S.–China Super Intelligence (SI) Dialogue to exchange views on “risks and benefits related to SI,” with the next exchange by November 2026, and a bilateral communication channel for SI incidents; the two leaders agreed to use the term “super intelligence” in place of “artificial intelligence.” The same fact sheet put the U.S.–China Board of Trade and Board of Investment into operation, with recommendations for more favorable tariff treatment on $30 billion of goods in each direction. China’s Foreign Ministry readout on September 26 confirmed an AI-incident communication mechanism and a November AI dialogue, and the two sides agreed to sign a memorandum of understanding on military crisis communications and to meet again at APEC in Shenzhen in November and later at the G20 in Miami. On September 26 the President said the two sides would not integrate on super intelligence because the United States is “leading China by a lot.” Basis: [S242], [S243], [S244].

[FACT] US Trade Representative Jamieson Greer said chip export controls were not on the agenda of the preliminary talks. A Taipei-based research firm’s forecast, reported on September 27, puts the share of Nvidia, AMD and other foreign suppliers in China’s AI-chip market at 21% in 2026, down from 34% a year earlier. Senate Democrats used the visit to press for votes on chip export-control legislation. Basis: [S245], [S246].

[FACT] The National Quantum Initiative reauthorization is before Congress: the Senate Commerce Committee passed S.3597, the National Quantum Initiative Reauthorization Act of 2026, unanimously on April 14, 2026, and the House Science Committee reported H.R. 8462 on April 29, 2026. Both await floor action and a reconciled text. Basis: [S247].

Quantum and the summit. [FACT] None of the published summit outcomes — the White House fact sheet or China’s readout — mentions quantum technology. [INFER] That does not mean quantum was absent from the talks. AI and quantum have moved closely together: Beijing benchmarks its

quantum center against the platforms that link quantum processors to AI supercomputers, and quantum computing’s computational power is expected to accelerate AI and super-intelligence capability. Side discussions likely touched on quantum even though no outcome records them; this report has no evidence of what was said. [ARG] The new channel covers AI; the domain where China’s own records show the most movement has none. Basis: [S242], [S243], [S213]; Part One, Section 1.

This is the third and final report in the authors’ series on China’s frontier technology in AI and quantum. The first two were published on September 20 and 22, 2026.

• Report One (Part IV, September 20), An IonQ Solution for the AI Force, set out the infrastructure layer beneath the AI transition that the United States can contract now. Its China half, The Conversion Machine, showed how Beijing mobilizes money, laboratories, people and law.
• Report Two (Part V, September 22), Land, Sea, Air and Space, mapped what that machinery is building, layer by layer, and set IonQ’s full-stack assembly against it, from a trusted foundry to orbit. It proposed nine actions for the Administration.

 This report closes that gap, and it is the last report because the AI lane was the last gap.

How this report is built. Part One maps China’s quantum-AI lane and stands on its own: it makes no company recommendation and can be circulated separately. Part Two sets out the American answer, IonQ’s integration with NVIDIA, and the two actions that name it. The opening, conclusion and appendices serve both.

Executive Summary

[ARG] The disclosure-gap thesis in two sentences. China fields more than it declares, spends more than it announces and discloses less than it builds. What can already be observed is enough to act on; what cannot is a collection requirement. Basis: Sections 8 and 9; The Baseline Has Moved.

The argument is set out on page 1, under “Two Days in September.” This summary gives the findings behind it, the actions that follow, and what to watch.

Key numbers

Number	Type	What it measures	Source
4	Observed	US Entity-Listed bodies inside Beijing’s Quantum-AI Collaborative Innovation Consortium, including its lead	[S20], [S21], [S23]
3	Observed	Regions where the consortium lead, CETC’s CAEIT, runs quantum centers: Beijing, Suzhou, Shenyang	[S23], [S89], [S91]
100 / 31	Plan / target	Cities and provinces targeted by the “Light Up 100 Cities” quantum network, with satellite-ground links	[S99]
16	Observed	City quantum networks China Telecom had built by 2024	[S100]
¥15 billion	Observed	Shanghai Future Industry Fund, 15-year state patient capital, reviewed by Premier Li on September 22	[S63], [S64], [S107]
¥3.204 billion	Reported	China’s first-quarter 2026 quantum financing, more than all of 2025	[S9]
32–63%	Estimate	Pentagon estimate of Chinese military spending above the announced budget	[S76], [S86]
~200	Observed (state TV)	Drones directed by one soldier in a National University of Defense Technology swarm test	[S75]
2.198 exaflop/s	Observed	LineShine, the world’s fastest supercomputer, built on domestic CPUs	[S15]
< 0.4%	Calculation	Share of a trapped-ion error-correction cycle used by the NVQLink round trip to NVIDIA GPUs	[S42], [S36]
0.32 vs 0.24	Company research	Quantum against best classical radar change detection, run on IonQ hardware	[S38]
2027	Schedule	Installation of IonQ’s Superion 256 at NVIDIA’s quantum research center	[S30]
2030 / 2031	Deadline	US federal deadlines for post-quantum key establishment and signatures	[S110]
0	Declared	Publicly identified operational dedicated quantum-key satellites in orbit (Jinan-1 re-entered October 2025; Micius January 2026)	[S132], [S2]
1	Reported	Medium-high-orbit quantum satellite its team said in January 2026 it would prepare “all-out” to launch; no launch date announced	[S138], [S145]
1	Observed	Cold-atom gyroscope operating on China’s space station since 2022	[S153]
10+	Official claim	Types of experimental equipment a PLA Information Support Force center has led work on, with quantum among its priorities	[S186]

What China is building

1. The baseline has moved. [FACT] The public assessments most often cited in Washington and across NATO — among them USCC (November 2025), CSIS (January 2026) and SIPRI (July 2025) — describe Chinese integration as limited, quantum sensing as lab-bound, Jinan-1 as in service and new satellites as due in 2025. Each statement has since been overtaken by documented Chinese and US government records. [INFER] Used today, they understate China’s advances and positioned growth, which undermines American, allied and NATO security. Basis: [S214], [S215], [S216], [S213], [S132], [S211]; The Baseline Has Moved.
2. China is connecting the pieces into a system. [FACT] Beyond quantum processors, China is building the integration layer: a China Telecom-controlled quantum-electrical fusion center in Beijing E-Town aims at a full-stack ecosystem benchmarked, in the municipal government’s words, against IBM Qiskit and NVIDIA NVQLink; national standards for testing quantum computers and for quantum key distribution security are co-drafted by PLA universities; national tasks call for real-time error correction; and satellite hardware for the next quantum constellation is in procurement. [INFER] The competition is increasingly architecture versus architecture, not processor versus processor. Basis: [S213], [S196], [S197], [S198], [S199], [S200]; Sections 1 and 9.
3. The PLA is trialling more than ten experimental systems, with quantum a priority. [FACT] A center under the Information Support Force has led research on more than ten types of experimental equipment across cloud-native computing, AI and quantum technology, and trialled several results on the mission front line. [INFER] The likely set spans secure command links, jam-resistant navigation, submarine detection and intelligence processing. Basis: [S186], [S180], [S181]; Section 8.
4. The next orbital quantum layer is in procurement. [FACT] China has no publicly identified operational dedicated quantum-key satellite in orbit (Jinan-1 re-entered in October 2025, Micius in January 2026), yet its satellite team said in January it would prepare “all-out” for a new launch; quantum sensors already operate on its space station; and a geostationary quantum satellite is announced for about 2027. [ARG] What is in orbit beyond the public record is a collection requirement. Basis: [S138], [S153], [S145], [S200]; Section 9.
5. Orbit is already contested. [FACT] The State Department found that Chinese firms’ satellite imagery enabled Iranian strikes on US forces. The rest of the non-quantum counterspace record is set out in one box in Section 9. Basis: [S217]; Section 9.
6. China fields more than it declares. [FACT] Drone swarms under a single operator, two sixth-generation fighter prototypes known only by analysts’ designations, military humanoid research, a planned quantum constellation under development, and military spending the Pentagon puts at 32 to 63% above the announced budget. [ARG] Chinese disclosure is an incomplete indicator of total capability; the gap is a collection requirement. Basis: [S73], [S75], [S76], [S77], [S61], [S86]; Sections 8 and 9.
7. The premier is driving the money. [FACT] Li Qiang wrote quantum and a state risk-sharing mechanism for future industries into national policy in March, chaired a State Council meeting on them in June, and on September 22 reviewed the state-funded Shanghai fund that backs a quantum start-up in this report. Basis: [S70], [S71], [S63], [S64]; Section 1.
8. A documented method beneath the controls. [FACT] RiverONE generates model parameters with simulated quantum computation, fixes them as classical tensors and runs on classical GPUs; its 1.9-billion-parameter model reaches at least 95% of NVIDIA’s Ising

Calibration 1 with under 10% of its parameters by the abstract’s measure (the paper body states under 5%). [INFER] It is a method plus a small firm (about ¥100 million raised), not yet a sector — but it needs no controlled item. Basis: [S24], [S6], [S28].

1. The fusion is national policy. [FACT] The State Council’s “AI+” Opinions direct AI–quantum collaborative innovation, and a quantum-intelligence alliance was launched at China’s flagship AI conference in July. Basis: [S16], [S2].
2. An Entity-Listed defense-electronics institute leads it in Beijing. [FACT] CAEIT leads the consortium; the Beijing Academy of Quantum Information Sciences, the Beijing Academy of Artificial Intelligence and the CAS Institute of Computing Technology — all members — are also Entity-Listed. Basis: [S20], [S21], [S23].
3. Quantum processors are already in the AI loop. [CO-STATED] Origin Quantum reports fine-tuning a billion-parameter model on its superconducting computer, a claim repeated by state media but not independently replicated; QBoson offers Boltzmann-machine training on Alibaba Cloud. Basis: [S12], [S11].
4. The distribution layer is national. [FACT] Tianyan — one 24-qubit, two 176-qubit and one 504-qubit machine — sits on SCNet beside China’s supercomputing centers, and China’s classical supercomputing now tops the TOP500 list without US accelerators. Basis: [S25], [S14], [S15].
5. American tools help validate the build. [CO-STATED] Google, IBM and Xanadu frameworks were tested on an Entity-Listed firm’s GPUs, NVIDIA’s open model is the benchmark, and the firm reports supplying AMD since 2023. Basis: [S6], [S24], [S28].

The American answer

1. A repeatable allied playbook. [INFER] IonQ places its processors inside allies’ national AI supercomputing through their lead institutes and local partners, as in Korea’s Hangang program, and fields both quantum-key and post-quantum security. Basis: [S93], [S94], [S105], [S106]; Section 15.
2. A US-fabricated QPU wired to America’s leading AI rack. [FACT] Superion 256, fabricated at IonQ’s SkyWater subsidiary, will link to an NVIDIA GB200 NVL72 over NVQLink at NVAQC, with installation in 2027. Basis: [S30], [S31], [S43].
3. Physics favors the pairing. [INFER] NVQLink’s round trip is under 4 microseconds; trapped-ion error-correction cycles run 1 to 5 milliseconds. By calculation from published specifications, the link uses under 0.4% of each cycle, and IonQ’s decoder fits on one CPU, leaving the whole GPU rack free for AI work. Basis: [S42], [S36], [S44]; Section 12.
4. Customers and allies before installation. [FACT] Superion 256 has been sold to the University of Cambridge. [CO-STATED] IonQ reports sales to SDT for a Korean hybrid quantum-classical data center and to Florida International University. [FACT] Korea’s science ministry has announced the plan to link IonQ’s Tempo to the national Hangang supercomputer. Basis: [S40], [S35], [S37], [S93], [S94].
5. Quantum-AI has reached ISR. [FACT] On Capella radar imagery of Marine Corps Air Station Miramar, the quantum model scored a filtered F1 of 0.32 on IonQ hardware against 0.24 and

0.16 for classical baselines, and 0.41 on an ideal simulator. It is a feasibility result. Basis: [S38].

1. IonQ builds both layers of quantum-safe security. [FACT] Through ID Quantique it fields quantum key distribution and post-quantum cryptography in one architecture — the dual approach China is building and US policy has not adopted. Basis: [S105], [S106], [S2]; Section 15.

20. It is not the only contender. [FACT] Quantinuum’s Helios has demonstrated real-time decoding over NVQLink. [ARG] IonQ is an early anchor of the architecture, not the architecture itself. Basis: [S41]; Section 16.

What this paper proposes

[ARG] Eight actions specific to the quantum-AI lane (Q-1 to Q-8), each with an owner and a 90-day deliverable. They add to Report One’s five charter asks and Report Two’s nine actions and repeat none of them. Basis: [S1], [S2]; Sections 10 and 17.

1. Q-1 Anchor federal hybrid quantum-GPU work on a reference architecture: a US-fabricated QPU beside AI supercomputing over an open interconnect (DOE Office of Science).
2. Q-2 Name quantum-inspired AI as a tracked category and screen every consortium member against the Entity List and Section 1260H list (Commerce BIS and DoD).
3. Q-3 Diligence and customer vigilance US chip, cloud and software vendors’ supply relationships with Chinese quantum-AI firms (Commerce BIS outreach).
4. Q-4 Keep the yardstick American with open quantum-AI benchmark sets (NIST with DOE).
5. Q-5 Standardize with allies on the quantum-GPU interface with NVQLink adopters in Japan, Korea, Taiwan, Singapore, Australia and Europe (State with DOE).

1. Q-6 Update the baseline used by Congress: a Commission staff update or hearing drawing on the documentary sources identified here, and a Quantum First goal extended beyond computing to quantum networks, space and sensing (USCC).
2. Q-7 Watch the standards Chinese national quantum standards and their PLA drafters, and Chinese submissions to ISO/IEC and the ITU (NIST and State, with NATO).
3. Q-8 Secure the anomaly data allied gravity and magnetic anomaly datasets for navigation (Department of Defense with NATO allies).

What to watch in the next twelve months

• NVAQC installation and the first Superion 256 deliveries, including FIU and SDT (2027).
• The TuringQ and Origin Quantum prospectuses, which will show what China’s champions certify about their AI claims and shareholders.
• Any independent replication of RiverONE or of Chinese quantum fine-tuning results.
• The next Section 1260H update, and whether CAEIT or other consortium members are named.
• Whether an AI incident-notification channel is agreed after the summit.
• The APEC leaders’ meeting in Shenzhen in November, which both leaders pledged to support and where AI is on the agenda. [S51], [S62]
• Whether the chip-control bills survive in the defense authorization bill after the elections. [S57]
• Any launch of the three promised low-orbit quantum satellites or the medium-high-orbit satellite, and whether it is declared as quantum. [S138], [S135]
• Europe’s Eagle-1, now late 2027 or early 2028, against China’s “Dawn” geostationary quantum satellite, announced for about 2027. [S152], [S145]
• Whether US officials name the Chinese entity behind the Jordan imagery, and any further sanctions. [S161], [S163]

Why now. [ARG] The integration layer is being chosen in 2026 and 2027, in machine rooms rather than at summits. Whichever architecture laboratories adopt first is the one they will keep. Basis: adoption record [S41]; deployment schedules [S30], [S37].

Source basis and confidence

Each key judgment is shown with its sources, whether those sources originate independently or repeat one underlying source, and the authors’ confidence. A single strong primary source is not a weak source; independence is noted so readers can weigh it.

Key judgment	Sources	Independence	Confidence
China is building the integration layer: a China Telecom-controlled full-stack center in Beijing E-Town, benchmarked against IBM Qiskit and NVIDIA NVQLink	S213, S196, S197	Primary municipal government notice, plus two separate outlets (Beijing News; Xinhua Liaowang)	High
The builders are on US lists (CAEIT and three consortium members Entity-Listed; China Telecom, China Mobile, CETC on 1260H)	S20, S21, S29	Independent primary US government records, including the Mandarin 1260H list	High
Defense, drone swarms and reconnaissance are written into Beijing E-Town’s quantum policy	S18, S19	Single originating document, posted on two official sites	High
PLA institutions co-draft national quantum standards	S198, S206, S210	Two published standards records (quantum-computer testing; QKD security), plus the space-QKD project record	High that they are named drafters; not evidence of control. Space QKD: the original plan roster does not name them; whether they were added at review is a collection item
The premier and the state are steering capital into quantum	S70, S63, S64, S107, S195	Multiple official sources (Government Work Report, Xinhua, fund disclosures, SASAC)	High
A 31-province state quantum network on infrastructure the state can requisition	S99, S133, S137, S101	Multiple independent sources (operator, peer-reviewed paper, company results, statute)	High for scale and law; military use is inference (moderate)
The next quantum satellites are in procurement	S200, S201, S132	Two separate procurement records, plus re-entry reporting	High that the tenders were posted (one terminated, one open); June payload detail from an archive copy only (moderate-high), described as quantum communications, not QKD; undeclared payloads are a collection requirement (low)
The PLA is trialling more than ten experimental systems, quantum a priority	S186	Single originating official report (Science and Technology Daily)	High that it was stated; low on which systems are quantum
Quantum sensing has moved to field testing	S167, S193, S81, S153	Multiple indicators; the magnetometer paper is single-origin, relayed by two outlets	Moderate-high for field progress; operational military capability not established (low).

Key judgment	Sources	Independence	Confidence
			Marine gravimetry standard registered (S211)
China is working the real-time error-correction problem	S199, S202, S207	Multiple independent sources (ministry task list, peer-reviewed result, industry white paper)	High that the problem is being worked; moderate on maturity
An organized mechanism recruits overseas quantum talent	S208, S209	Multiple primary sources (national fund guide; consortium member’s recruitment)	High, at mechanism level only
Chinese firms’ imagery enabled Iranian strikes on US forces (a State Department finding)	S217, S163	Official US finding (State Department, recorded by the Commission)	High for the official finding
China and Russia cooperate on photonic and quantum technology	S205, S116	Official CAS and TASS texts	High for the memorandum
Chinese state media describes anti-satellite equipment while the foreign ministry opposes space weapons	S171, S174, S172, S173	Separate official statements and a published paper; HQ-29 role from footage only	High that the statements exist; low for HQ-29’s anti-satellite role
IonQ’s processor will sit on NVIDIA’s AI system; early quantum radar result	S30, S41, S38	Company-originating announcements; a single preprint	High that it is announced (delivery 2027); moderate for the radar result

Three Critical Issues

 Three issues decide the quantum-AI lane. Each is set out below with the evidence and the decision it demands.

 

Report at a Glance

Item	Scope
Question	How is China fusing quantum technology with AI before fault-tolerant hardware exists, who governs and pays for it, and what is the American answer?
Place in series	Report Three of Three (QTIS Part VI). It assesses the AI lane that Reports One and Two explicitly left open. Part One (China) stands alone and makes no company recommendation; Part Two (the American answer) names IonQ.

Item	Scope
Companies covered	Qingxing Yigou (清醒异构; Thewake Systems); QBoson (玻色量子); TuringQ (图灵量子); Origin Quantum (本源量子), AI work only; China Telecom Quantum Group; Taiyi Liangsheng (太一量生); Fuxi Quantum (伏曦量子; not NORINCO’s Fuxi robot); IonQ; NVIDIA.
Institutions covered	CETC’s Electronic Science Research Institute; Beijing E-Town; NUDT; the National Supercomputing Center in Shenzhen; NVIDIA’s Accelerated Quantum Research Center; KISTI.
Sources	Chinese-language government, financial and technology sources; US federal and law-firm material; IonQ, NVIDIA and TOP500 releases; technical preprints; Reports One and Two.
Not covered	A full assessment of the humanoid-robotics and drone-swarm lanes (Section 8 covers them only as context); classified material; independent replication of Chinese technical claims.

The Baseline Has Moved

Updating the assessments Washington and NATO are using

[FACT] This analysis compared its evidence against the public assessments most often cited by congressional leaders, administration officials and allied stakeholders: the U.S.-China Economic and Security Review Commission staff report Vying for Quantum Supremacy and the Commission’s 2025 Quantum First recommendation (November 2025); the CSIS assessments Understanding China’s Quest for Quantum Advancement (January 2026) and Quantum Sensing and the Future of Warfare (October 2025); the SIPRI primer Military and Security Dimensions of Quantum Technologies (July 2025); the Department of Defense China Military Power Reports for 2024 and 2025; the Defense Intelligence Agency’s 2025 threat assessment; the 2026 Annual Threat Assessment of the U.S. Intelligence Community; NATO’s Joint Air Power Competence Centre study (September 2026); the German Marshall Fund’s analysis (May 2026); and the Commission’s China–Iran timeline (September 2026). Basis: [S214], [S215], [S179], [S216], [S181], [S218], [S180], [S219], [S150], [S151], [S217].

[FACT] China’s quantum technology activity, capability, intent and execution have all moved substantially beyond what those assessments describe. The change is documented in Chinese and US government sources set out in the table below. Basis: [S213], [S200], [S201], [S198], [S206], [S210], [S211], [S217].

[FACT] Used today, those assessments omit or understate developments recorded in Chinese and US government sources — the integration center, the satellite procurement, the standards rosters, the sensing trials. Funding, export-control, force-posture, standards and alliance decisions calibrated to them rest on a baseline China has already passed, and that undermines the security of the United States, its allies and NATO. Each assessment was sound on the evidence of its time; the problem is that the record has moved and the documents in circulation have not.

Statements this report examined, by assessment

[FACT] This table counts only the statements this report examined in each assessment — a selected set, not an audit of any document. The Department of Defense’s December 2024 Military Power Report is not counted, because its December 2025 successor is the live baseline. RAND’s February 2024 testimony is shown separately below the total: most of its lines were overtaken by time, not wrong when written. Basis: [S214], [S223], [S215], [S216], [S222], [S219], [S224].

Assessment	Statement s checked	Overtaken, incomplete or overstated	Still hold	Direction
USCC staff report (Nov 2025)	6	4	2	Understates China
USCC Annual Report (Nov 2025)	2	1	1	Omits networks, space and sensing
CSIS (Jan 2026)	6	2	4	Understates China
SIPRI (Jul 2025)	8	4	4	Mixed: overstated orbit timing; understated the rest
ODNI Annual Threat Assessment (Mar 2026)	2	1	1	Omits networks, space and sensing
MERICS tracker (2026)	3	1	2	Overstates one item

Assessment	Statement s checked	Overtaken, incomplete or overstated	Still hold	Direction
Total, current assessments	27	13	14	Predominantly understates
Shown separately: RAND testimony (Feb 2024)	11	9	2	Overtaken by time, not wrong then

The documents reviewed

Document	Date	Primary audience	Method	China sourcing
USCC staff report, Vying for Quantum Supremacy, and 2025 Quantum First recommendation	Nov 18, 2025	Congress	US laboratory fact-finding (June 2025)	Secondary
CSIS, Understanding China’s Quest for Quantum Advancement	Jan 29, 2026	Executive branch (NIST-requested review)	Desk research	Secondary and state media
CSIS, Quantum Sensing and the Future of Warfare	Oct 9, 2025	Policy and defense community	Desk research	Secondary
SIPRI, Military and Security Dimensions of Quantum Technologies	Jul 2025	European governments and NATO	Desk research	Secondary
Department of Defense, China Military Power Reports 2024 and 2025	Dec 2024; Dec 23, 2025	Congress	Official assessment	Classified and open sources
Defense Intelligence Agency, 2025 threat assessment	May 2025	Congress	Official assessment	Classified and open sources
ODNI, Annual Threat Assessment 2026	Mar 18, 2026	Congress	Intelligence Community assessment	Classified and open sources
NATO Joint Air Power Competence Centre, quantum for air and space power	Sep 2026	NATO	Desk research	Secondary
German Marshall Fund, Quantum Entangles the Heavens	May 2026	Transatlantic policy community	Desk research	Secondary
USCC, Enabling Iran timeline	Sep 1, 2026	Congress	Official compilation	US government and press
USCC, 2025 Annual Report to Congress (Quantum First by 2030)	Nov 18, 2025	Congress	Commission report	Secondary
RAND, testimony to the Commission: The Chinese Industrial Base and Military Deployment of Quantum Technology (still widely cited)	Feb 1, 2024	Congress	Expert testimony	English-language publications

Document	Date	Primary audience	Method	China sourcing
Corroborating: ASPI Critical Technology Tracker update	Dec 1, 2025	Allied governments	Research-output data	Publication data
Corroborating: MERICS China Tech Observatory, quantum	2026	European governments	Tracker	Chinese and Western sources
Corroborating: NATO Science & Technology Organization, Science & Technology Trends 2025– 2045	Apr and Jul 2025	NATO	Foresight	Open sources

[INFER] None of these documents used the Chinese primary records on which this report rests: standards-agency drafting rosters, provincial and national procurement listings, municipal government notices and the Pentagon’s Mandarin-language 1260H list. That is why the update below is citable rather than merely more recent. Basis: [S213], [S198], [S206], [S210], [S200], [S201], [S29].

Then and now

 

Figure. The record moved after the assessments were written. Source: Assessment publication dates and developments as cited in this section; compiled by the authors.

Domain	What the assessment said	What the record now shows	Confiden ce	Direction	What it means
Integratio n	USCC (Nov 2025, p. 3): China’s centralized model may struggle with cross-domain integration. CSIS (Jan 2026, “Quantum High-Performance Computing	Mar 31, 2026: Beijing’s economy bureau announces a full-stack center benchmarked against IBM Qiskit and NVIDIA NVQLink; national real-time error-correction tasks; May 2026 white paper using NVQLink as its case study.	High	Understate d	The layer counted as a US structural advantage is now a named Chinese target. [S214], [S215], [S213], [S199], [S207]

Domain	What the assessment said	What the record now shows	Confiden ce	Direction	What it means
	Integration”): integration appears limited.
Quantum satellites	USCC (pp. 9–10): CN-QCN connected to Jinan-1 as of August 2025. SIPRI (p. 17): two to three low-orbit quantum satellites to launch in 2025.	Jinan-1 re-entered in October 2025; none launched in 2025. Medium-high-orbit satellite component tendered (Dec 2025); Henan microsatellite tendered (Sep 2026); space-QKD standard under review.	High	Mixed	A gap year in orbit, with the rebuild visible in procurement. [S214], [S216], [S132], [S200], [S201], [S210]
Allied space QKD	SIPRI (p. 16): Europe’s Eagle-1 to launch late 2025 or early 2026.	Eagle-1 has slipped to late 2027 or early 2028.	High	Understate d allied gap	The allied gap in space is about two years wider than NATO planning assumes. [S216], [S152]
Network scale	CSIS (“Quantum Communications”): 5.5 million QKD service users (June 2025).	7.2 million quantum-communication users reported by China Telecom (mid-2026); China Mobile’s 100-city, 31-province network; China Telecom, China Mobile and CETC on the 1260H list.	High	Understate d	A mass encryption layer run by carriers the Pentagon lists as military companies. [S215], [S137], [S99], [S29]
PLA role	SIPRI (p. 17, citing a secondary source) and USCC (p. 3, in general terms): the PLA is a user and beneficiary.	NUDT and the Information Engineering University named drafters of GB/T 47468-2026 (quantum-computer testing) and GB/T 47679.2-2026 (QKD security); the space-QKD project’s original roster did not name them.	High that they are named; an indicator, not control	Understate d	A general assertion becomes documentary evidence. [S216], [S214], [S198], [S206], [S210]
Standards	SIPRI (Executive summary): quantum standards may become a strategic battleground. CSIS (“Quantum Communications”): China positioned to lead QKD standard-setting.	Four 2026 Chinese national standards and projects in quantum computing, QKD, space QKD and marine gravimetry.	High	Understate d	The battleground is live at the national level. [S216], [S215], [S198], [S206], [S210], [S211]
Cryptogra phy	CSIS (“Quantum Communications”): China’s own post-quantum algorithm call could bifurcate global infrastructure.	China Telecom fields QKD with post-quantum cryptography; its 2030 plan calls for a new cryptographic system fusing both.	High for the plans; moderate for bifurcatio n	Understate d	A Chinese alternative security stack is forming. [S215], [S136], [S194]
Quantum sensing	USCC (pp. 12–13): largely in the	Marine absolute-gravimeter standard (Aug	Moderate -high for	Understate d	Moving from laboratory to field and

Domain	What the assessment said	What the record now shows	Confiden ce	Direction	What it means
	laboratory and fragmented. SIPRI (p. 26): prototypes and drone demonstrators.	2026); field-tested magnetometer; gravimeter on a procurement catalog; imaging radar in trial; an Information Support Force quantum receiver prototype field-tested at tens of kilometers (Dec 2025); cold-atom gyroscope in orbit.	progress; operation al capability not establish ed		standards. [S214], [S216], [S211], [S167], [S193], [S220], [S153]
Navigatio n data	SIPRI (pp. 18–19): gravity and magnetic anomaly data are becoming strategic resources.	China’s marine gravimeter standard is tied explicitly to its national positioning, navigation and timing system.	High	Understate d	Anomaly mapping is now a standardized Chinese national effort. [S216], [S212]
Commerci al space intelligenc e	Not addressed in the USCC report (Nov 2025).	State Department (May 8, 2026): Chinese firms’ imagery enabled Iranian strikes on US forces; some firms hold PLA certifications.	High for the official finding	Omitted	Chinese orbital intelligence has already been used against US forces. [S217]
Military applicatio n	DoD China Military Power Report (Dec 2024): the PLA may use QKD for command and control and sensing for submarine detection.	DoD (Dec 2025, as reported by Nikkei Asia): Beijing treats quantum as vital to future industry and national security. A PLA Information Support Force center is trialling more than ten experimental systems, quantum a priority.	High that stated	Understate d	From intention to trials. [S181], [S218], [S186]
Intelligenc e Communit y framing	ODNI (Mar 2026, pp. 12–13): quantum treated as a computing and encryption race; no country has yet built a cryptographically relevant quantum computer.	China’s documented lead lies in quantum networks, satellites, sensing and standards — none of which the public quantum section addresses.	High	Omitted	The most widely read public intelligence assessment omits the domains where China is furthest ahead. [S219]
Capital	CSIS (“Recent Developments”): a national venture guidance fund of about $138 billion, quantum a focus.	Premier Li reviewed Shanghai’s ¥15 billion fund; SASAC calls quantum a source of assassin’s-mace technologies; China Telecom’s 2030 plan.	High	Understate d	Capital now arrives with explicit strategic direction. [S215], [S63], [S195], [S194]
Talent	Not treated.	A national overseas-scientist fund (RMB 1–3 million over three years); an Entity-Listed quantum academy recruits through it.	High, at mechanis m level	Omitted	An organized channel for overseas expertise. [S208], [S209]
China–	SIPRI (p. 16): a	CAS–Rosatom	High for	Understate	An institutional

Domain	What the assessment said	What the record now shows	Confiden ce	Direction	What it means
Russia	2023 China–Russia quantum link test.	memorandum (May 2026) providing for personnel exchanges in quantum technologies.	the memoran dum	d	channel, now a collection requirement. [S216], [S205]
Opacity	USCC (p. 3): breakthroughs often unverified; advanced efforts likely concealed.	Consistent with this report’s findings.	Agreeme nt	Agrees	The unseen portion is listed as collection requirements. [S214]
Quantum First scope	USCC Annual Report (Nov 2025): Quantum First targets computational advantage in cryptography, drug discovery and materials science by 2030.	China’s documented lead lies in quantum networks, space and sensing, outside that target set.	High	Omitted	The flagship congressional goal is aimed where the US leads, not where China is advancing. [S223]
PLA weapons	MERICS (2026): the PLA is testing over ten quantum warfare weapons.	The Chinese source reports more than ten experimental systems across cloud-native computing, AI and quantum, trialled at the front; it does not say all are quantum.	High	Overstated	Circulating claims err in both directions; primary records settle them. [S224], [S186]

Exhibit: RAND’s 2024 testimony, then and now

[FACT] RAND’s February 2024 testimony to the Commission remains a standard reference on China’s military use of quantum technology. It was careful and noted how little public information existed; the comparison below is between the public record then and the public record now. Most lines were overtaken by time, not wrong when written; that is why the scorecard counts RAND separately. Basis: [S222].

Page	What the testimony said	What the record now shows	Status
3	Chinese researchers are fast followers, rarely at the forefront, except in quantum communications.	A full-stack center benchmarked against Qiskit and NVQLink; PLA-drafted national standards; national real-time error-correction tasks; a marine gravimeter standard.	Overtaken in several domains [S213], [S206], [S199], [S211]
3	Much of China’s open sensing research focuses on quantum radar.	Documented work now spans gravimetry, magnetometry, quantum receivers and cold-atom inertial sensing.	Overtaken [S211], [S167], [S220], [S153]
3–4	Quantum radar is impractical.	Still the assessment of this report.	Holds [S216]
4	China has the only two known QKD satellites.	Both have re-entered; replacements are in procurement.	Overtaken [S132], [S200], [S201]
4	The NSA assesses QKD as unsuitable for US national security systems.	Unchanged.	Holds [S214]

Page	What the testimony said	What the record now shows	Status
4	No public information on whether the PLA uses the national QKD network.	NUDT and the Information Engineering University are named drafters of the national QKD-security standard (GB/T 47679.2-2026).	Overtaken [S198], [S206]
4–5	No comparable public PLA operational test of quantum technology; no Chinese demonstrations of cutting-edge quantum sensors for navigation.	An Information Support Force quantum receiver field test (Dec 2025); more than ten experimental systems trialled at the front; a navigation-linked marine gravimeter standard; a cold-atom gyroscope in orbit.	Overtaken [S220], [S186], [S212], [S153]
4	Operational utility for quantum sensing expected around 2024–2029 (Defense Science Board).	That window is now current.	Overtaken by time
5	China’s military universities are not at the forefront; the Information Engineering University is not a leading publisher.	That university is now a named drafter of two national quantum standards.	Overtaken [S198], [S206]
6	Large Chinese technology companies appear to be pulling back from quantum.	Q1 2026 quantum investment exceeded all of 2025, with large-scale capital returning.	Overtaken [S224]
6	No formal institution-level quantum partnerships with foreign institutions are known.	CAS–Rosatom memorandum with personnel exchanges in quantum (May 2026); CETC helping establish Pakistan’s National Centre for Quantum Computing.	Overtaken [S205], [S224]

What recent sources add

[FACT] ASPI’s December 2025 tracker update finds China leading high-impact research in 66 of 74 critical technologies and now leading research on small satellites, previously a US-led area; its updated China Defence Universities Tracker maps how that research aligns with China’s civil–military fusion system. It also finds the United States still employs the largest share of top-tier technology talent. Basis: [S225].

[FACT] MERICS records Chinese quantum investment of CNY 3.2 billion in the first quarter of 2026, more than in all of 2025; CETC helping establish Pakistan’s National Centre for Quantum Computing; and commercial deployment of a Zuchongzhi 3.0-based superconducting computer by China Telecom Quantum Group and QuantumCTek. Basis: [S224].

[FACT] NATO’s Science & Technology Trends 2025–2045 identifies the race for AI and quantum superiority as one of six key trends over the next twenty years. Basis: [S226].

[INFER] Allied and independent sources confirm the direction of change. NATO frames the quantum race over twenty years; the documented Chinese record shows integration, standards and satellite procurement under way now, which compresses the Alliance’s planning horizon. Basis: [S225], [S224], [S226].

Domain by domain: what it means

Integration. [INFER] The most consequential change. Both the Commission and CSIS treated integration as a likely Chinese weakness. Beijing has since named two American commercial platforms, IBM’s Qiskit and NVIDIA’s NVQLink, as its benchmarks and put the software ecosystem

first — the same layer the Commission’s proposed Quantum Software Engineering Institute is meant to secure. The competition is architecture versus architecture, and the United States should plan as if China intends to contest the interface, not only the processor. Basis: [S214], [S215], [S213].

Orbit. [INFER] The satellite picture circulating in Washington and Brussels is a year out of date. There is no publicly identified Chinese quantum-key satellite in orbit, but the replacements are in procurement, a space-QKD standard is under review, and a geostationary satellite is announced for about 2027. Allied planning should track the rebuild, not the gap. Basis: [S132], [S200], [S201], [S210], [S145].

Standards and the PLA. [INFER] SIPRI anticipated that standards would become contested; the Chinese rosters show PLA universities inside that process. These are indicators of civil-military participation, not evidence of control, and they are exactly the signals that allied standards bodies should be tracking. Basis: [S216], [S198], [S206], [S210].

Sensing and navigation. [INFER] The Commission and SIPRI were right that operational quantum sensing is not established, and SIPRI rightly cautions against the recurring claim that quantum sensors will soon make the oceans transparent. What has changed is maturity: field tests, procurement catalogs and national standards for marine gravimetry. Submarine-survivability and navigation assessments should track this trajectory, not the old laboratory baseline. Basis: [S214], [S216], [S211], [S167].

Commercial space intelligence. [INFER] The Commission’s November 2025 report predates the Iran war. Its own September 2026 timeline records a US government finding that Chinese commercial imagery enabled strikes on American forces. Any assessment of China’s orbital program that omits this is incomplete. Basis: [S217].

Capital, talent and cryptography. [INFER] CSIS documented the capital; the record now shows the direction attached to it, the recruitment channel beside it, and a Chinese security stack forming around QKD and domestic post-quantum algorithms. Basis: [S215], [S195], [S208], [S136].

 Statements now outdated, and what to use instead

“China’s quantum network is connected to the Jinan-1 satellite” (USCC, Nov 2025, pp. 9–10). Jinan-1 re-entered in October 2025; no operational quantum-key satellite is publicly identified in orbit. [S132]

“Two to three low-orbit quantum satellites will launch in 2025” (SIPRI, Jul 2025, p. 17). None launched; the next satellites are in procurement. [S200], [S201]

“Eagle-1 will launch in late 2025 or early 2026” (SIPRI, p. 16). Now late 2027 or early 2028. [S152]

“China’s quantum–supercomputing integration appears limited” (CSIS, Jan 2026, “Quantum High-Performance Computing Integration”). Beijing launched a full-stack center benchmarked against Qiskit and NVQLink on March 31, 2026. [S213]

“China’s centralized model may struggle with cross-domain integration” (USCC, p. 3). China is now building that layer explicitly. [S213], [S199], [S207]

“China’s quantum sensing remains in the laboratory” (USCC, pp. 12–13). Field tests and national standards now exist; operational capability is still not established. [S211], [S167]

“5.5 million QKD users” (CSIS, “Quantum Communications”). China Telecom reports 7.2 million quantum-communication users by mid-2026. [S137]

“Chinese researchers are fast followers, rarely at the forefront, except in quantum communications” (RAND, 2024, p. 3). Overtaken in integration, standards, error correction and marine gravimetry. [S222], [S213], [S206]

“No public information on whether the PLA uses the national QKD network” (RAND, 2024, p. 4). PLA universities now draft national QKD standards. [S222], [S206]

“Quantum First” targets computing only (USCC Annual Report, Nov 2025). China’s lead lies in networks, space and sensing. [S223]

Overstated: “the PLA is testing over ten quantum warfare weapons” (MERICS, 2026). The Chinese source reports more than ten experimental systems, not all quantum. [S224], [S186]

“The PLA is among the network’s users” as a general assertion. Now documented: PLA universities are named drafters of two national quantum standards, GB/T 47468-2026 and GB/T 47679.2-2026. [S198], [S206]

 

 

Context: Two Days in September

On September 23, IonQ agreed to put its newest quantum computer inside NVIDIA’s quantum research center, wired to America’s most powerful AI hardware. On September 24, as Xi arrived at the White House, a Shanghai start-up announced new funding for AI built with simulated quantum computers and run on Chinese chips, using a workflow that does not obviously fall within the hardware-focused quantum export-control categories. In Beijing, the body leading the city’s quantum-AI consortium has been on a US export blacklist since 2021. [S30, S6, S24, S20]

The two leaders called it a good day. Neither agreed to slow down. [S3, S4]

 What we can watch Beijing assembling. [FACT] A China Telecom-controlled center in Beijing E-Town is building a quantum-electrical full-stack ecosystem that Beijing’s economy bureau says is benchmarked against IBM Qiskit and NVIDIA NVQLink; PLA universities co-drafted China’s national standards for testing quantum computers and for quantum key distribution security; the carriers building its quantum networks are on the Pentagon’s list of Chinese military companies; and hardware for its next quantum satellites is in procurement. On September 22, 2026, the eve of Xi’s arrival in Washington, Premier Li reviewed the state fund that backs a quantum start-up in this report. [INFER] China is connecting quantum, AI, classical compute, communications, sensing, space, talent and state capital into one system. [ARG] Beijing discloses only part of what it builds, and the rest is a collection requirement. But the part we can already see is enough to act on. [S213, S196, S198, S206, S29, S200, S201, S63]

S E R I O U S C O N C E R N

What Beijing fields, and what it admits

In orbit. [FACT] China has no publicly identified operational dedicated quantum-key satellite in orbit: Jinan-1 re-entered in October 2025 and Micius in January 2026. In January its satellite team told Xinhua it would prepare “all-out” in the new year for the launch of a new medium-high-orbit quantum satellite; no launch has been announced, and public timelines point to about 2027. Separately, and without declared quantum payloads, its Qianfan broadband constellation passed 250 satellites in September 2026; on September 23, the day Xi arrived in Washington, it launched a 26th batch of Guowang satellite-internet satellites, reported on the broadcaster’s military channel; and its computing satellites identify targets on board. [S138, S132, S74, S2]

On the battlefield. [FACT] State television showed one soldier directing some 200 drones in a National University of Defense Technology swarm test. The Pentagon assesses the WZ-9 drone as semi-operational and reports two sixth-generation fighter prototypes in flight test since December 2024, known publicly only by analysts’ tentative designations. NORINCO markets a military humanoid robot for sentry and reconnaissance duty, and a Reuters review of more than 100 procurement records finds China’s defense establishment accelerating humanoid research toward eventual wartime use. [S75, S76, S77, S61]

Quantum in the field. [FACT] China’s leading quantum scientist says quantum sensing already shows value in autonomous navigation and target detection. Quantum ground stations have shrunk from 13 metric tons to under 100 kg for mountains and plateaus, and an Entity-Listed Beijing academy has flight-tested a 38-gram atomic clock on drones. A PLA Information Support Force center says it has led work on more than ten types of experimental equipment, with quantum among its priorities, and trialled results on the mission front line. [S78, S79, S80, S186]

A record of understatement. [FACT] The Pentagon estimates China’s real military spending at 32 to 63% above its announced budget. US agencies found that DeepSeek’s widely cited training cost left out data taken from American models. The Pentagon has long read Deng Xiaoping’s “hide our capacities and bide our time” as a desire to downplay China’s capabilities,

 

Post-Summit Note

[FACT] Xi’s state visit ended on September 25. It extended the trade truce to January 10 and produced a U.S.–China Super Intelligence Dialogue and an incident channel, but no rules on AI development (see About This Report). The items below record what the summit showed; nothing in them changes the report’s findings. Basis: [S87], [S88].

1. China’s readout asks Washington to stop “guarding against” it

[FACT] According to the Chinese readout, Xi said the two countries are both AI powers that compete but can cooperate more, and should each play to their strengths rather than guard against each other (各展所长而不是相互设防). He proposed continuing AI dialogue on risks and benefits, jointly preventing AI’s misuse, and keeping human control over AI. The Chinese readout says Trump favored continued dialogue; no specific rules or enforcement mechanism were announced. Basis: [S51].

[FACT] A People’s Daily front-page commentary on September 22 restated China’s four red lines: Taiwan, democracy and human rights, paths and systems, and development rights. Xi defined the last when he first set out the red lines, at his November 2024 meeting with President Biden in Lima: neither side should deprive the other of its legitimate development rights to keep its own lead, and national security must not be used as a pretext to maliciously restrict and suppress other countries. In June, Premier Li told Summer Davos that “building walls” and “technology blockades” (筑墙设垒、科技封锁) were intensifying and that some leading Chinese firms faced “unfair external suppression.” Basis: [S65], [S66], [S72].

[INFER] Xi’s call not to “guard against each other” sits on a red line Beijing has itself defined as covering US technology restrictions. It was made at the summit where Part One shows those controls already missing the quantum-AI lane. Basis: [S51], [S65], [S66]; Part One, Sections 1 and 3.

1. The President says China shares his hands-off stance

[FACT] Trump wrote “That is China’s position also,” suggesting Xi also opposes new AI regulation, and said the US “guardrail” is the Department of Justice. CNBC read the post as apparently ruling out an agreement to strengthen AI safety precautions. Basis: [S52].

[INFER] Both capitals are now on record declining new limits. Page 1’s “neither agreed to slow down” rests on the President’s own words. Basis: [S52], [S51].

1. Beijing timed a chip push to the summit

[FACT] CNBC reported that Huawei brought forward its next AI chip by three quarters, Alibaba unveiled what it called China’s most powerful AI processor, and DeepSeek reportedly lined up the largest known cluster of Huawei chips. Bloomberg reported on September 4 that DeepSeek plans to deploy at least 160,000 Huawei Ascend 950DT accelerators at a data center in Inner Mongolia, to run its models rather than train them, on a timeline dependent on Huawei’s production. [UNDISC] Neither company had confirmed it as of September 9. Basis: [S53], [S54].

[INFER] China’s domestic AI compute is scaling. That widens the room for methods, like RiverONE, that run on domestic chips and need no controlled item. Basis: [S53], [S24]; Section 3.

4. Both governments now have an AI incident to point to

[FACT] CNN reported on September 18 that in spring 2026, during operations against Iran, a US analyst’s AI tool misidentified a Chinese ship’s cargo as nuclear-program parts, and that troops and aircraft were readied to board it before the error was caught. On September 24 China’s defense ministry spokesman said the reports showed the “huge risks” of unrestricted military use of AI. Basis: [S55], [S56].

[FACT] NATO faces the same problem. Its AI-enabled Maven Smart System reached full operational capability on June 22, 2026, accredited for classified networks; in May a NATO official called for agreed standards and confidence thresholds for AI-enhanced geospatial intelligence; and NATO has tested eight AI models on its classified network for geospatial intelligence, requiring traceability and saying quantum could accelerate the processing. Basis: [S96], [S97], [S98].

[INFER] The episode gives the proposed incident channel a concrete case. It also bears on Part Two’s radar argument: faster machine analysis of overhead sensing has to be matched by verification. Basis: [S55], [S56]; Section 14.

1. Congress has a vehicle for chip controls

[FACT] Senate Democrats used the visit to press for votes on bipartisan bills restricting China’s access to advanced AI chips. Republicans agreed to include them in the defense authorization bill’s manager’s amendment, but that bill is stalled until after the elections. [INFER] That gives Q-2 and Q-3 a legislative vehicle alongside Commerce’s existing authority. Basis: [S57]; Section 10.

Part One — China’s Quantum-AI Lane

Part One stands on its own. It makes no company recommendation.

1. Who Leads the Lane

National policy names the fusion

[FACT] The State Council’s August 2025 Opinions on the “AI+” Action (国发〔 2025〕 11 号) direct stronger collaborative innovation between AI and quantum technology, alongside biomanufacturing and 6G (加强人工智能与生物制造、 量子科技、 第六代移动通信（ 6G） 等领域技术协同创新). Report Two recorded that on July 18, 2026, at the World AI Conference, the president of USTC argued that quantum and space computing answer computing’s physical limits, and that a quantum-intelligence alliance and a space-computing committee were launched the same day.

[INFER] The two national programs point at each other. The fusion of AI and quantum is written into the State Council’s AI program, and an alliance has been organized around it at China’s flagship AI conference. Basis: [S16], [S2].

[FACT] Nationally, the National Venture Guidance Fund — financed by ultra-long-term special government bonds and approaching about $138 billion — names quantum among its focuses, and its three regional funds signed agreements in December 2025 covering quantum technology. SIPRI records announced government quantum investment of $15.3 billion for China against $7.7 billion for the United States as of mid-2025. Basis: [S215], [S216].

 The premier’s record on quantum, March to September 2026

March 5 — policy. [FACT] Li’s Government Work Report pledges to “establish an investment-growth and risk-sharing mechanism for future industries” and to cultivate quantum technology among them (建立未来产业投入增长和风险分担机制， 培育发展…量子科技…等未来产业). It tells government investment funds to “take the lead as patient capital” (政府投资基金要带头做耐心资本), calls for the “new-type whole-nation system” (新型举国体制) for core technologies, and describes China’s quantum research and application as at the world’s forefront. Basis: [S70].

June 5 — the State Council. [FACT] Li chairs an executive meeting on future industries, which resolves to “firmly grasp the initiative in development” (牢牢把握发展主动权). The readout covers future industries generally and does not name quantum; by March’s definition, quantum is among them. Basis: [S71], [S70].

June 24 — abroad. [FACT] At Summer Davos, per the official Chinese transcript, Li says quantum information technology “picked up speed” in the first half (跑出“ 加速度” ) and that China’s large quantum science facilities are open to the world. He denies that subsidies explain the competitiveness of China’s new-energy and connected vehicles, but describes its innovation funding as “enterprise-led, government-guided and society-participating” (企业主导、 政府引导、社会参与), with new risk-sharing policies. Basis: [S72].

September 22 — delivery. [FACT] In Shanghai he views quantum-computing products (察看…量子计算等领域技术产品), reviews the Future Industry Fund, and repeats March’s risk-sharing pledge word for word. Basis: [S63], [S64].

[INFER] The same pledge, word for word, in March policy and at a September fund inspection shows execution, not rhetoric. Beijing distinguishes “subsidy,” which it denies, from “government guidance” and state patient capital, which it announces. And the openness is selective: facilities are offered to the world while specialists are subject to exit controls and core bodies sit on the Entity List. Basis: [S70], [S72], [S64], [S1], [S20], [S21].

The integration layer

[FACT] On March 31, 2026 China Telecom Quantum Group, as controlling shareholder, and nine partners including ZTE, QBoson and 中科弧光 established the Beijing Quantum-Electrical Fusion Industry Innovation Center (北京量电融合产业创新中心) in Beijing E-Town, pooling the shareholders’ capabilities in quantum computing, intelligent computing and high-speed interconnect. The Beijing Municipal Bureau of Economy and Information Technology states that it is “dedicated to building a quantum-electrical full-stack ecosystem benchmarked against IBM Qiskit and NVIDIA NVQLink” (致力于构建对标 IBM Qiskit、 英伟达 NVQlink 的量电融合全栈生态); its first priority is to stake out the software ecosystem, followed by quantum-AI fusion, multi-route infrastructure and applications. Basis: [S213], [S196].

[FACT] Official commentary frames the stakes the same way. Xinhua’s Liaowang weekly wrote in January 2026 that quantum computing will be decided not by any one machine but by whether a country can build a sustainable full-stack ecosystem, and that China must master quantum chips, measurement-and-control systems and operating systems. China’s May 2026 national standard for testing quantum computer performance (GB/T 47468-2026) lists among its drafters the National University of Defense Technology and the Information Engineering University — both PLA institutions — alongside USTC, Hefei National Laboratory, QuantumCTek and China Telecom Quantum Group. The same two PLA institutions are named among the drafters of the national standard on security requirements, testing and evaluation for quantum key distribution (GB/T

47679.2-2026, in force from December 1, 2026), alongside CETC’s 30th and 15th Research Institutes. Basis: [S197], [S198], [S206].

[INFER] China is not only pursuing competitive processors; it has named two American commercial platforms — IBM’s Qiskit software and NVIDIA’s NVQLink interconnect — as its benchmarks, and identified the integration layer — software, control, interconnect and error correction — as strategic, in the same E-Town district as the CAEIT consortium and the Quantum Ten Measures, under a 1260H-listed parent. PLA institutions co-drafting national standards for testing quantum computers and for the security of quantum key distribution is a direct civil-military indicator; it does not by itself show PLA control of the quantum networks. Basis: [S213], [S196], [S198], [S206], [S29], [S18].

[FACT] The integration layer is also operating locally. The Hefei Quantum-Supercomputing Fusion Computing Center went online in January 2026 on the Chaohu Mingyue supercomputer, with two superconducting machines of more than 180 qubits and one 12-qubit trapped-ion machine, and opened quantum computing services to users worldwide. Its supercomputing chips are entirely domestic: Huawei Ascend, Cambricon and Hygon. Basis: [S236].

[FACT] The same ecosystem is working the systems problem. A May 2026 white paper by researchers from China Mobile Research Institute, Xoherence (相干科技, a member of the Beijing consortium) and firms based in Beijing E-Town proposes a six-layer reference architecture for real-time quantum error correction, uses NVIDIA’s NVQLink as its case study for interface standardization, and argues that the bottleneck has shifted to tail latency and end-to-end integration. It is a white paper, not a demonstration. Basis: [S207].

The consortium and its lead

[FACT] In August 2025 Beijing launched a Quantum-AI Collaborative Innovation Consortium, organized by the municipal science commission, the Zhongguancun administration and the Economic-Technological Development Area. It was launched at a quantum industry conference in the zone on August 19, 2025 and is led by CETC’s Electronic Science Research Institute — the China Academy of Electronics and Information Technology (CAEIT). Members named on the BDA’s official preview page include Peking University, the Beijing Academy of Quantum Information Sciences, the Beijing Academy of Artificial Intelligence, the CAS Institute of Computing Technology, the CAS Institute of Software, 相干科技 (English name not confirmed) and QBoson. The stated purpose is civilian: AI-based quantum algorithms and quantum-AI applications in digital finance, smart transport and materials research.

[FACT] At the same event Beijing launched a Quantum Computing Industry Innovation Center with China Telecom, planning a multi-modality software stack, an open-source quantum community and what it calls China’s largest quantum-classical fusion infrastructure and cloud.

[FACT] CAEIT is on the US Entity List in its own right (aliases CAEIT and CETC CAEIT; 11 Shuangyuan Road, Badachu High-Tech Park, Beijing). It was added by the rule published on July 12, 2021, in which it was grouped with the Xinjiang-related entities — a listing tied to surveillance in Xinjiang, not to quantum technology. Canada lists it among research organizations connected to military or state-security entities. On the Department of Defense’s Section 1260H list, published in the Federal Register on June 10, 2026, CETC appears with its 14th, 15th and 32nd Research Institutes, CETC Cyberspace Security and CETC Digital Technology; the Mandarin list confirms CAEIT is not named. Since June 30, 2026, Section 805 of the FY2024 NDAA has barred the Department from contracting with listed entities or entities they control.

[INFER] CETC’s control of CAEIT would appear to bring the institute within that contracting ban even though it is not named on the list. That is a legal question for counsel, not a finding of this report.

Basis: [S29], [S131].

Consortium body	Role	US Entity List
CAEIT (CETC Electronic Science Research Institute)	Lead	Added July 12, 2021 (Xinjiang-related group)
CAS Institute of Computing Technology	Member	Added December 2022 (entities supporting China’s military and defense industry)
Beijing Academy of Quantum Information Sciences	Member	Added May 2024 (quantum-related group)
Beijing Academy of Artificial Intelligence	Member	Added March 25, 2025 (for developing large AI models and advanced computing chips for defense purposes)
Peking University; CAS Institute of Software; 相干科技 (legal name 相干（ 北京） 科技有限公司; brand domain xoherence.com); QBoson	Members	Not listed in the sources reviewed

[FACT] The Beijing Academy of Artificial Intelligence, a consortium member, was listed in March 2025 because, in the Commerce Department’s words, it had “developed large artificial intelligence (AI) models and advanced computing chips for defense purposes.” Basis: [S21].

The same institute, in three regions

[FACT] CAEIT is not only a Beijing consortium lead. On June 26, 2026 Liaoning’s provincial science department, Shenyang city and Hunnan district signed with CAEIT to build a Northeast Asia quantum industry innovation center, led by Academician Lu Jun, a CETC chief scientist, with CAEIT’s party secretary and president present. The center is tied to China’s Fifteenth Five-Year Plan for quantum. Basis: [S89], [S90].

[FACT] Since December 2021 CAEIT has also run a Yangtze River Delta quantum industry innovation center in Suzhou. By CAEIT’s own account it has made and tested several superconducting quantum chips, built a 20-qubit quantum computer test platform and a quantum computing-network cloud, and been told by local leaders to focus on a 100-qubit machine. Basis: [S91].

[FACT] CETC describes itself as holding technical dominance in China’s military electronics and cyber sectors, charged with national defense modernization (在国内军工电子和网信领域占据技术主导地位， 肩负着实现国防现代化…的职责使命). Basis: [S92].

[INFER] The Entity-Listed body that leads Beijing’s quantum-AI consortium builds quantum computers in three regions, and its parent states in its own words that it serves defense modernization. Basis: [S20], [S23], [S89], [S91], [S92].

[FACT] Writing in CAEIT’s own journal, experts from a CETC communications institute’s secure-communications key laboratory state that quantum secure communication has extremely high military and commercial value (具有极高的军事价值和商业价值). Basis: [S203].

[INFER] The body that leads Beijing’s quantum-AI research is already export-restricted by the United States, and so are three of its members, listed on different grounds over four years. The lead’s listing

concerns Xinjiang surveillance; the finding here is about who governs the lane, not why each body was listed. CETC appears in Report Two as the maker of China’s XS1000 dilution refrigerator; in the quantum-AI lane it also sets the research agenda. That is civil-military fusion in the governance of the lane, not only in its supply chain. Basis: [S20], [S21], [S23], [S29].

1. What the District Policy Says

[FACT] On July 14, 2025 the Beijing Economic-Technological Development Area Management Committee issued the Several Measures on Promoting Quantum Science and Technology and Industrial Development (《北京经济技术开发区关于推动量子科技和产业发展的若干措施》 , 京技管发

〔 2025〕 15 号), commonly called the “Quantum Ten Measures.” Effective until December 31, 2027,

it is described as Beijing’s first district-level quantum-industry policy.

[FACT] The operative sentences of the official text read:

• “开放‘ 量子+’多领域应用场景， 在国防、 通讯、 金融、 医疗等领域打造‘ 量子+’示范应用。 ”

— open “quantum+” application scenarios and build “quantum+” demonstrations in national defense, communications, finance and medicine.

• “推进量子通信技术在无人驾驶、 无人机群、 移动终端、 感知侦察等领域应用。 ” — apply quantum communication in driverless vehicles, drone swarms, mobile terminals, and perception and reconnaissance.
• “推动量子精密测量技术在…航天惯性导航等场景应用落地。 ” — deploy quantum precision measurement in scenarios including aerospace inertial navigation.

[FACT] The same text promotes hybrid quantum-classical computing for deep learning and large-model training acceleration. Flagship demonstration projects can receive up to ¥1 million.

[FACT] The Scenario Handshake Plan launched in the same district in September lists proposed projects including underwater sensing and imaging, and quantum-secured communications for vehicles. Basis: [S60].

[INFER] The consortium’s own announcement uses civilian language: finance, transport, materials. The district policy that hosts it names defense, drone swarms and reconnaissance. That gap is the finding. The dual-use intent sits in the district’s published policy, not in the consortium’s description of itself. Basis: [S23], [S18].

1. The Method Beneath the Controls

[FACT] A note on terms: quantum-inspired means classical algorithms that borrow mathematical methods from quantum physics and run on ordinary chips. It is distinct from quantum-enhanced systems, which use quantum hardware. Basis: [S24].

[FACT] Qingxing Yigou (清醒异构), which trades in English as Thewake Systems, was founded in January 2021 by Tsinghua computer-science alumnus Dr. Yu Teng. Its first product was an automatic parallelizing compiler; in 2022 it targeted industrial control, driver assistance, consumer electronics and commercial aerospace. Its chief scientist then was Professor Zheng Wang of the University of Leeds, a Fellow of the Alan Turing Institute.

[FACT] The method is now on the record. RiverONE: Generating Knowledge-Intensive VLM by Simulated Quantum Machines (arXiv:2606.29966, June 29, 2026), from the company’s Thewake RiverYtz Lab with Teng Yu as last author (the founder Yu Teng, in Western name order), describes

quantum-generated parameters produced by simulated quantum computation during model construction and then materialized as classical tensors. The finished model runs entirely on classical GPUs, with no quantum hardware or runtime simulation. The paper reports that the 1.9-billion-parameter model reaches at least 95% of NVIDIA Ising Calibration 1 on quantum-calibration plot understanding with under 10% of its parameters, as stated in the abstract; the paper body states under 5%. The code is public on GitHub.

[CO-STATED] Yu Teng says the project received government funding support and was featured by the Xuhui District Science and Technology Commission. Source: [S45]. The firm itself is small: about

¥100 million raised across its A and A+ rounds. Track One is therefore a documented method plus a small firm industrializing it, not yet a sector.

[INFER] The 95% comparison concerns one task, quantum-calibration plot understanding against a named NVIDIA model. It is not evidence of superior general AI, of a quantum hardware advantage, or that export controls have failed. Basis: [S24].

[CO-STATED] The company moved its research focus to quantum AI in Shanghai in 2025.

 

[CO-STATED] The company’s website says variational quantum circuits replace transformer fully connected layers, with models delivered for precision machining and agricultural robots.

 

1. Quantum Processors Inside AI

[CO-STATED] Origin Quantum and the Hefei Comprehensive National Science Center’s AI institute reported fine-tuning a billion-parameter model on the Origin Wukong superconducting computer, with 76% fewer parameters, 8.4% better training results, and strict math-reasoning accuracy up from 68% to 82%. The claim was carried by Xinhua; it has not been independently replicated. Basis: [S12].

[CO-STATED] QBoson offers its coherent optical machines on Alibaba Cloud with a PyTorch plug-in for training Boltzmann machines, and has open-sourced quantum-native AI training kits. TuringQ says its Gen3 Gaussian boson sampling unit is being explored for world models.

[INFER] These are research demonstrations without independent replication in the sources reviewed. Their policy weight lies in the direction of travel: domestic hardware tied into mainstream AI tooling. Basis: [S10], [S11], [S12].

5. The Distribution Layer

The national network

[FACT] SASAC describes the Tianyan cluster as one 24-qubit, two 176-qubit and one 504-qubit superconducting machine — the largest public quantum resource connected in China. This describes the cluster as of April 2025; Report Two records later additions.

[FACT] China Telecom Quantum Group’s Tianyan cloud joined SCNet, the national supercomputing internet, which links more than 20 centers across 14 provinces and cities. The announcement named AI, code-breaking and drug discovery among the target areas.

[FACT] On October 11, 2025 China Mobile, with CETC, China Information and Communication Technologies Group and about ten other partners, launched the “Light Up 100 Cities” (点亮百城) quantum trial network and a quantum ecosystem alliance including Huawei and ZTE. The network is to link quantum-key distribution across cities through coordinated metro, backbone and satellite-ground (星地) links, pilot government and finance uses, and extend to all 31 provinces. China Mobile issued its first quantum metro-network equipment tender, for Hefei, on October 30, 2025. Basis: [S99], [S100].

[FACT] The carriers building China’s quantum networks are themselves on the Department of Defense’s June 2026 list of Chinese military companies: China Telecom Group, China Telecom Corporation and China Telecom Satellite Communications, which control QuantumCTek and ran the “two satellites, three networks” trial, and China Mobile, which leads the Light Up 100 Cities network. So is CETC, its partner. Basis: [S29].

[FACT] China’s Institute of Commercial Cryptography Standards launched its own call for post-quantum algorithms in 2025, separate from NIST’s process; CSIS warns this could split global cybersecurity infrastructure into US-aligned and China-aligned standards. Basis: [S215].

[INFER] This is quantum communication, not quantum-AI. It matters here because CETC appears in three state quantum lanes — leading the quantum-AI consortium through CAEIT, building refrigerators, and partnering on the national key-distribution network — and a fourth military one, multi-domain drone swarms (Section 8) — and because the network is planned to join satellite and ground links. Basis: [S99], [S23], [S2].

From 100 Cities to the Battlefield: the Network as Defense Infrastructure

[INFER] No source says the national quantum network is military. Its importance comes from three documented facts taken together: a state-run key network is being built across all 31 provinces; it is planned to join satellite and ground links; and China’s defense mobilization law lets the state requisition civilian facilities and equipment and modify them to military requirements. Basis: [S99], [S101], [S2].

Element of the build	Scale	Source
China Telecom city quantum networks	16 cities built by 2024, including Hefei’s, described as the world’s largest	[FACT] [S100]
QuantumCTek quantum-secure infrastructure	40 cities	[FACT] [S2]
“Light Up 100 Cities” trial network	100 cities, extending to all 31 provinces, metro + backbone + satellite-ground	[FACT] [S99]

Element of the build	Scale	Source
Operating fiber network (CN-QCN)	More than 10,000 km of trusted-relay QKD; 145 backbone nodes; 20 metro networks; 17 provinces and 80 cities; 6 satellite-capable ground stations	[FACT] [S133]
In orbit	None publicly identified; the next satellites are in procurement (Section 9)	[FACT] [S132], [S200]
Procurement	China Mobile’s first quantum metro-network tender, Hefei, October 30, 2025	[FACT] [S100]
Users (terrestrial)	China Telecom: 7.2 million quantum-communication users by mid-2026, more than 5,000 government, finance and energy customers; quantum revenue up 65.4% in 2025. These are fiber and mobile users, not satellite users	[FACT] [S137]
QKD with PQC	China Telecom Quantum Group’s commercial cryptographic system combining quantum key distribution with post-quantum cryptography, deployed across 40 key cities	[FACT] [S136]

 [FACT] The PLA created an Information Support Force on April 19, 2024, charged with building and operating a network information system for joint operations. Its mission, per Western analysts of PLA statements, is a network that supports joint and multi-domain operations under any enemy attack; at its founding Xi told it to keep information flowing, integrate information resources and strengthen information protection. China’s National Defense Mobilization Law defines civilian resources as facilities and equipment owned or used by organizations, obliges every organization to accept their requisition, and provides that requisitioned resources may be modified to military requirements at state expense; the revised law, published on August 28, 2026 and in force from October 1, retains requisition of civilian resources. Basis: [S102], [S108], [S103], [S101].

 

[FACT] Allies and US critical-infrastructure operators have reached different conclusions. Romania and Slovakia have deployed national QKD networks under the EU’s EuroQCI program with IonQ’s subsidiary ID Quantique: Romania’s spans 36 links and more than 1,500 kilometers, over 20% of Europe’s terrestrial quantum communications infrastructure, and Slovakia’s connects the Presidential Palace and National Security Authority facilities. In Chattanooga, a Department of Energy-funded project with Oak Ridge and Los Alamos ran QKD between an electric utility’s substations. Basis: [S231], [S232], [S233].

Military significance, ranked

[INFER] The ranking is the authors’ judgment of military weight, applied only to what is sourced. It separates what is operating from what is promised, and quantum keying from the satellites that do other jobs. Basis: Sections 5 and 9.

Rank	Element	Status	Why it matters militarily	Source
High, enduring	CN-QCN fiber network and metro quantum-key networks	Operating	Strategic communications security at home: government, finance and energy links on a network the state can mobilize	[S133], [S101]
High, demonstrated , now latent	Space-to-ground keying (12,900 km) and the portable ground station	Proven, then the satellites re-entered	Shows China can extend keys beyond its fiber to outstations, ships and islands once a satellite flies	[S79], [S132], [S134]
Medium	China Telecom’s 7.2 million quantum-communication users	Operating (terrestrial)	A mass encryption layer on phones and enterprise lines, relevant to continuity of government in a crisis	[S137]
Medium, rising	Three low-orbit satellites and a higher-orbit satellite	Promised; preparation “all-out” since	Would close the current gap in orbit; the higher orbit would give longer passes and more key per	[S135], [S82], [S138]

Rank	Element	Status	Why it matters militarily	Source
		January 2026; pre-IOC or undeclared	pass
Not quantum keying	Qianfan and Guowang broadband; Three-Body and other AI-computing satellites	Operating, different mission	Satellite communications and orbital computing; significant in their own right, but they distribute no quantum keys	[S2], [S74], [S120]

[INFER] Track Three is how Tracks One and Two reach scale: a state carrier distributes quantum resources alongside classical supercomputing and hundreds of open models through one national network. Basis: [S14], [S25].

The classical champion

[FACT] LineShine, at the National Supercomputing Center in Shenzhen, debuted at No. 1 on the June 2026 TOP500 list with 2.198 exaflop/s, displacing El Capitan. It is the first system to exceed two exaflops sustained on CPUs alone and also leads the HPCG ranking. It was China’s first TOP500 submission in nine years. It runs on a domestic Arm-based LX2 processor, which The Next Platform reports was presumably designed with Huawei. Its mixed-precision result ranks fourth, consistent with a CPU-only design.

[INFER] LineShine is not a quantum-AI machine. Its relevance is that China now has a world-leading classical supercomputer built without US accelerators — the kind of classical capacity any quantum-classical system depends on. The quantum side of China’s national network is Tianyan (Section 5). Its weaker mixed-precision result also shows where China’s AI training still lags GPU-based systems. Basis: [S15], [S14].

[UNDISC] The sources reviewed do not state whether LineShine itself is connected to SCNet.

The integration layer abroad

[FACT] Report Two documented Origin Quantum’s freely downloadable operating system and its global cloud. [INFER] One further step matters for this lane. [FACT] In 2025 Origin signed an agreement with ChinaLink ESGt on what could become a large integrated computing center in Málaga, Spain.

[INFER] A multi-hardware operating system offered free, paired with a European facility, is a bid to make a Chinese integration layer the default outside China — the same layer NVQLink and CUDA-Q occupy on the American side. Basis: [S2], [S13].

1. Research Scale, Capital and Control

The marketplace, in China’s own assessments

[FACT] China’s industry-ministry think tank, CAICT, counts more than 800 quantum-information firms worldwide, about half in computing, and places China and the United States in the first tier. In September 2026 its standards institute head described QKD’s supply chain as basically mature and in scaled pilot deployment, quantum repeaters as not yet achieved, and the domestic industry as a preliminary ecosystem. Basis: [S239], [S238].

[FACT] Capital is accelerating. Domestic quantum-computing financing in the first eleven weeks of 2026 (CNY 2.2 billion) nearly matched all of 2025 (CNY 2.5 billion) on the same count, and Origin Quantum closed a pre-IPO round of nearly CNY 3 billion at a valuation near CNY 24 billion. Hefei alone reports 107 quantum firms with revenue above CNY 6 billion. Scope matters here: the CNY

2.2 billion and 2.5 billion figures are 36Kr and IT Juzi counts of quantum-computing financing only, with 2026 running to March 17; the CNY 3.2 billion first-quarter figure reported by Xinhua and MERICS covers all quantum financing over the full quarter. The two are consistent. Basis: [S240].

 

[ARG] This report does not select a market-size figure: Chinese and Western estimates for 2030–35 are not on a common definition and differ by an order of magnitude.

Research scale

[FACT] ASPI’s Critical Technology Tracker measures each country’s share of the top 10% most-cited papers. For 2019–2023, China led quantum communication (33.6%, against 16.8% for the United States) and narrowly led quantum sensors (24.1% against 23.8%); the United States led quantum computing (33.6% against 15.9% for China). In its later update covering 2021–2025, ASPI finds China leading 69 of 74 technologies, with quantum computing among the five where the United States still leads. Basis: [S46], [S47].

 

Figure 1. Share of high-impact quantum research, 2019–2023. Source: ASPI Critical Technology Tracker, EU top-five snapshot (October 2024). Top-10% most-cited papers; EU shown as a bloc. ASPI’s December 2025 update, covering 2020–2024, finds China leading 66 of 74 technologies and places TU Delft first in quantum computing research; see [S225].

[FACT] By volume rather than citations, MIT’s Quantum Index Report finds China and the United States close in quantum computing publications for 2019–2023, at 23% and 22% of the world total. Basis: [S48].

[INFER] China matches the United States on research volume in quantum computing and trails on impact; it leads outright in quantum communication and is level in sensing. No public dataset yet isolates quantum-AI research specifically, so these figures bound the lane rather than measure it. Basis: [S46], [S47], [S48].

Shanghai: a second quantum-AI hub

[FACT] On September 21, 2025, at a forum co-hosted by the Ministry of Science and Technology and the Shanghai government, Shanghai launched its own Quantum Artificial Intelligence Consortium of 12 research institutions and several firms, including Fudan University, the Shanghai Qi Zhi Institute, the Shanghai Research Center for Quantum Sciences and the Shanghai AI Laboratory. On June 30, 2026 Xuhui district opened a quantum-computing future-industry zone with 26 firms, Qingxing Yigou among them; the Shanghai AI Laboratory and Shanghai Instrument Group have worked with Qingxing to adapt its models to domestic AI chips, with municipal and district support. In September Xuhui issued measures naming “quantum-intelligence fusion” (量智融合) as a priority, with up to

¥100 million for a single project. Basis: [S111], [S112], [S113].

[INFER] Beijing’s consortium is one of at least two municipal quantum-AI vehicles, the second launched with a national ministry as co-host. Shanghai’s members are not the Entity-Listed set in Section 1; the pattern is national policy delivered through several cities. Basis: [S111], [S23], [S16].

Capital

[FACT] Xinhua put China’s first-quarter 2026 quantum financing above ¥3.2 billion, exceeding all of 2025, and called “quantum + AI” fusion the core lever for industrial value. CVSource and STAR Market Daily give ¥3.204 billion; 36Kr counts ¥1.14 billion in January, ¥450 million in February and

¥620 million in March through the 17th.

 

Figure 2. Monthly quantum-computing financing, January to mid-March 2026. Source: 36Kr (eu.36kr.com), March 17, 2026. March is partial.

[FACT] Beyond the Origin Quantum round covered in Reports One and Two, TuringQ raised close to

¥1 billion in the first half at a valuation above ¥7 billion and had its IPO tutoring filing accepted on August 5. QBoson completed a ¥1 billion Series B led by Beijing Financial Holdings and ICBC

Capital. Taiyi Liangsheng, founded in January 2026, raised more than ¥100 million within two months and ¥300 million in June. Fuxi Quantum, founded in April, raised two rounds led by CATL.

 

Figure 3. Selected disclosed 2026 financings. Source: Jiemian/Cailian Press (Origin); Xinhua (QBoson); Tencent News (TuringQ); Tou Zi Jie (Taiyi Liangsheng, Qingxing Yigou). Approximate as reported.

[INFER] The distance from company formation to nine-figure financing is now measured in weeks. State-guided funds are pre-positioned to back quantum teams as soon as the talent is assembled. Basis: [S26], [S9].

The NORINCO-led Origin Quantum round is covered in Reports One and Two and is not repeated. The table covers the rest of the lane.

Investor or lead	Nature	Company or body	Detail
CETC Electronic Science Research Institute (CAEIT)	Defense-electronics group research institute; Entity-Listed	Beijing Quantum-AI Consortium	Lead organization
CETC fund (电科基金)	Fund linked to CETC	Taiyi Liangsheng	Investor in ¥300 million Pre-A
National venture guidance fund, Yangtze River Delta	Central guidance fund	TuringQ	First quantum direct investment
Xuhui Capital; Anhui Hi-Tech Investment	District and provincial state platforms	Qingxing Yigou	A+ round investors
Beijing Financial Holdings; ICBC Capital	Municipal and state-bank capital	QBoson	Co-led ¥1 billion Series B
Shanghai Future Industry Fund	Municipal fund	Taiyi Liangsheng	Co-led angel round
iFlytek venture arm; CATL	Private strategic investors	Taiyi Liangsheng; Fuxi Quantum	Angel-stage leads

The premier and the fund

[FACT] On a Shanghai inspection tour reported by Xinhua on September 22, Premier Li Qiang was briefed on the operation of the Shanghai Future Industry Fund and viewed quantum-computing and

brain-computer-interface products. He called for more venture and patient capital and for mechanisms to share the risks of investing in future industries. Basis: [S63].

[FACT] The fund was launched in September 2024 with ¥10 billion, fully funded by Shanghai’s municipal treasury, for a 15-year term, and was enlarged in September 2025 to ¥15 billion with money from Shanghai’s municipal special-purpose bonds. Its managers named quantum computing among its target fields. By April 30, 2026 it had approved 30 direct investments and 38 sub-funds with a combined target of more than ¥30 billion, with over 90% of direct deals at Series A or earlier. It co-led Taiyi Liangsheng’s angel round. Basis: [S64], [S26].

[INFER] China’s second-ranking leader personally reviewed the state fund behind one of the quantum start-ups in this report. The fund is built to carry early-stage risk that private venture investors avoid, which is the state-direction argument of Part One in a single case. Basis: [S63], [S64], [S107], [S26].

[INFER] The pattern is layered. Local state platforms seed and steer, private strategic investors add scale and demand, and a defense-electronics group leads the body that sets the lane’s research agenda. Basis: [S6], [S10], [S11], [S23], [S26].

[ARG] This is military-civil fusion operating through consortium governance as well as capital, with no military contract ever appearing on the record. Basis: [S1], [S23], [S26].

[UNDISC] Board seats, information rights and side agreements attached to the CETC fund’s investment are not public.

1. US-Origin Touchpoints and Talent

Supplier relationships

[CO-STATED] Qingxing Yigou says it reached a formal cooperation with AMD in 2023 and became an AMD AI system supplier.

[FACT] Its framework tests ran on Biren GPUs. Biren’s Beijing, Shanghai, Guangzhou, Hangzhou and Zhuhai entities were added to the Entity List effective October 17, 2023.

[FACT] Several Chinese firms have reportedly accessed restricted chips’ compute remotely, through data centers in Southeast Asia. Basis: [S58].

[INFER] Testing open-source software on a listed company’s hardware inside China is not by itself an export of US items. The question is whether controlled technology or support moves through a US vendor relationship. Report Two’s R-2 covers the opposite direction — American use of Chinese quantum clouds. Basis: [S6], [S28].

The reverse flow: Western users on Chinese quantum clouds

[FACT] Know-how can also flow toward China. Origin Quantum says its cloud serves users in more than 160 countries and regions; its earlier cloud service reported the United States, Russia and Japan as its top international users. China Telecom’s Tianyan cloud reports more than 12 million visits from over 50 countries and offers adapters for Qiskit, Cirq and PennyLane, lowering the barrier for foreign developers. Basis: [S228], [S229], [S230].

[INFER] Western companies and researchers who run workloads on Chinese quantum clouds may expose problem formulations, data and application know-how to platforms operated by China Telecom, which is on the Pentagon’s 1260H list, or by Entity-Listed firms. That is a customer-side

risk that export controls on hardware do not address, and it belongs in vendor and customer diligence (Q-3). Basis: [S228], [S229], [S29].

Distillation

[FACT] On September 8 the NSA, CISA and FBI issued joint advisory AA26-251A naming six China-based AI companies — DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI — for industrial-scale extraction of capabilities from US frontier models, likely with the Chinese government’s knowledge. The advisory calls distillation the core of how these firms build, describes routing through fraudulent accounts, cloud providers, third-party aggregators and a gray market of API proxies, and says DeepSeek’s widely cited training cost excludes data obtained by distillation. Basis: [S67].

[FACT] China called the advisory a deliberate attack on its AI development. Separately, after Anthropic — a party to the dispute — alleged in a September 10 report that seven China-based labs had covertly routed requests through its Claude models, China’s cyberspace regulator reportedly summoned all seven and narrowed its inquiry to DeepSeek and Moonshot, examining whether sensitive police, military and state-linked data reached US models. Basis: [S68], [S69].

[INFER] Distillation is a fourth route by which American capability reaches Chinese AI, alongside open-source tools, supplier relationships and remote access to chips. Beijing’s response uses the same “development” language as its red line. Basis: [S67], [S68], [S65], [S66].

Software and benchmarks

[FACT] The frameworks validated on Chinese hardware are US- or allied-origin open source: Cirq (Google), Qiskit (IBM) and PennyLane (Xanadu, Canada). The headline benchmark compares against NVIDIA’s open Ising models, released on April 14, 2026.

[INFER] Open US tooling is both the on-ramp and the yardstick for Chinese quantum-AI. Whoever defines the benchmark shapes the field, which argues for publishing more of them. Basis: [S6], [S24].

Talent

[FACT] Taiyi Liangsheng’s chairman and CTO, Dr. Liu Hongbin, was previously Microsoft’s chief quantum architect and led its partnership with Atom Computing. NUDT reported in 2022 that its quantum team’s core members all had overseas visiting experience and more than 20 held doctorates from leading foreign universities.

[INFER] Western-trained talent is a core input to China’s quantum-AI build. Nothing here suggests wrongdoing by any named individual; these career moves are lawful and common. Report Two’s talent-lane proposal (R-7) addresses the policy response. Basis: [S26], [S27].

[FACT] The recruitment mechanism is institutional. The National Natural Science Foundation’s Overseas Excellent Young Scientists Fund, running since 2021, aims to attract accomplished overseas young scientists — including non-Chinese nationals — to return or come to China, with RMB 1–3 million over three years; the Beijing Academy of Quantum Information Sciences, an Entity-Listed member of the quantum-AI consortium, invites global candidates to apply through it. Basis: [S208], [S209].

[INFER] China has an organized, well-funded channel for converting overseas quantum experience into domestic research capacity. That is a mechanism-level finding; it says nothing about any individual and does not establish improper transfer. Basis: [S208], [S209].

8. The Military and Dual-Use Lane

[FACT] The framing is long-standing. A 2014 article in China National Defense News, quoted by SIPRI, compared the significance of quantum computing to nuclear weapons and likened a war between a quantum-equipped and a non-quantum army to a blind man fighting a sighted one. Basis: [S216].

[FACT] NUDT created the PLA’s first quantum-focused research institute in July 2022 and has carried more than 100 major national and military projects. Its quantum interdisciplinary center sits in the same building as the national laboratory that built Tianhe-2, and its QUANTA team develops the Quingo quantum programming language with the Chinese Academy of Sciences’ quantum cloud.

[FACT] The Department of Defense published its updated Section 1260H list on June 8, 2026, naming 188 entities. Analysts note the update reaches into AI, semiconductors and other dual-use sectors, including companies with indirect ties to the defense industrial base.

[INFER] In the quantum-AI lane the documented defense links run through governance (the consortium led by Entity-Listed CAEIT), capital (the CETC fund) and institutions (NUDT), and local incentives add stated defense, drone-swarm and reconnaissance scenarios. Report Two documented the orbital end of the same trend: Chinese satellites that identify targets on board.

Quantum-AI is the next layer that capability will draw on. Basis: [S2], [S18], [S23], [S26], [S27].

[FACT] The same pattern appears in robotics. A Reuters review of more than 100 Chinese military procurement notices, studies, patents and company materials, published on September 7, found China’s defense establishment accelerating research into humanoid robots’ military uses and planning for eventual wartime deployment. NORINCO’s August materials describe its full-size Fuxi (伏羲) humanoid for sentry duty, all-weather reconnaissance and patrol, paired with remote teleoperation; in June the PLA budgeted about $300,000 for a system to collect and label humanoid training data. Reuters found no evidence of an armed humanoid with an operational PLA unit. The robot is unrelated to Fuxi Quantum (伏曦). Basis: [S61].

[INFER] A defense conglomerate marketing dual-use robots while the PLA buys the data to train them follows the same pattern as CAEIT in quantum-AI: defense institutions at the center of a nominally civilian technology. Basis: [S61], [S20], [S23].

Quantum sensing moves toward the field

[FACT] CAICT describes quantum precision measurement as the fastest-landing quantum segment, with a distinctly dual military-civilian character: magnetometers, gravimeters and photonic quantum radar are in commercial iteration and used in defense, aerospace and precision detection, while quantum gyroscopes remain early. That is CAICT’s characterization; this report’s assessment that quantum radar claims are contested and largely impractical is unchanged. Basis: [S238].

[FACT] In December 2025 an engineer at a PLA Information Support Force center told Science and Technology Daily that a miniaturized quantum communication prototype had received radio signals from tens of kilometers away in a field test on the northern steppe, offering a new means of communication for new-type combat forces in valleys and dense forest. Researchers at the PLA’s Information Engineering University publish on Rydberg-atom receivers in Acta Physica Sinica. Basis: [S220].

[FACT] Marine quantum gravimetry is moving to standardization. On August 29, 2026 China registered a national standardization guidance technical-document project, 20264591-Z-469,

“Quantum measurement — Technical specifications for marine dynamic absolute gravimeter,” under the national quantum technology standards committee (TC578), drafted by the Changsha Quantum Measurement Industrial Technology Research Institute over ten months. Its official explanatory note says no international standard covers marine dynamic use, that the standard addresses ship-motion compensation, sea-state adaptability, shipborne-system integration and specialized data processing, and that it serves the maritime-power strategy and China’s national positioning, navigation and timing system. Basis: [S211], [S212].

[INFER] Shipborne absolute gravimetry supports gravity-aided navigation and seabed survey. The standard shows engineering maturity and intent; it does not establish a deployed military capability. Basis: [S211], [S212].

[FACT] Pan Jianwei said in March 2026 that quantum precision measurement shows initial application value in autonomous navigation, environmental monitoring and target detection (目标探测), and that China’s strontium optical clock, with 100-km free-space time transfer, can safeguard national strategic security (保障国家战略安全). Basis: [S78].

[FACT] The Beijing Academy of Quantum Information Sciences — an Entity-Listed member of the quantum-AI consortium in Section 1 — reports a 38-gram micro atomic clock that has completed rotor-drone flight tests and ground simulation of space conditions, and develops atomic magnetometers, atomic antennas and diamond NV sensors. Basis: [S80], [S21].

[FACT] China’s own literature records limits. A 2026 Chinese Science Bulletin review finds that drone-mounted quantum magnetometry achieves only nanotesla-level precision in practice, far below its theoretical picotesla level, because of unsolved interference. Basis: [S81].

[FACT] The field is moving. The Quantum Insider and the South China Morning Post report a 2026 Acta Physica Sinica paper from USTC, Hefei National Laboratory and a China State Shipbuilding Corp. marine navigation laboratory describing a compact atomic magnetometer, with a 7-cubic-centimetre probe, that tracked underground trains and located a buried magnet in field tests; the South China Morning Post reports the researchers intend it for anti-submarine warfare, though the paper tested no submarines. Japan is preparing ship-launched drones with quantum magnetic sensors by fiscal 2031 to watch the straits Chinese submarines use. Basis: [S167], [S168].

The PLA’s “ten-plus” experimental systems

[FACT] China’s official Science and Technology Daily, reprinted by People’s Daily Online on January 9, 2026, reported that a center under the PLA Information Support Force, working with a supercomputing laboratory at the National University of Defense Technology, turned in 2025 to cloud-native computing, AI and quantum technology. It placed several research projects in quantum and related fields, led research on more than ten types of experimental equipment (牵头开展十余型试验装备攻研), and organized trial use of several results on the mission front line. Its themes are “quantum plus supercomputing” and a “cyber-information plus perception” system for real-time detection, precise identification, cross-domain integration and rapid support. Basis: [S186].

[FACT] Western coverage recast this as “more than ten quantum cyber-warfare tools” tested at the front. The Chinese text does not say all ten are quantum or that they are weapons, and it names no individual system. Basis: [S177], [S178], [S186].

[UNDISC] The PLA has not listed the systems. The table below maps the reasonable quantum possibilities against what US defense assessments, allied analysts and Chinese literature say China is pursuing; it is the authors’ mapping, not an identification. Basis: [S176], [S180], [S181].

Possible tool	Evidence	Military meaning	Status
1. Quantum-accelerated intelligence processing	[FACT] The stated purpose: processing battlefield and open-source cyber data within seconds, and extracting high-value intelligence from public cyberspace. Basis: [S176], [S177].	[INFER] The only function the PLA has named. Speed of fusion, not a new sensor, is the claimed edge.	Official claim; no technical detail published
2. Quantum-secured command links	[FACT] The Pentagon assesses the PLA may use integrated quantum networks and quantum key distribution to reinforce command, control and communications; the Information Support Force runs that network mission. Basis: [S181], [S102].	[INFER] Hardened keys for the joint command network described in Section 5.	Plausible; infrastructure exists
3. Jam-resistant navigation	[FACT] The US Defense Intelligence Agency reports China investing in quantum inertial navigation; a cold-atom gyroscope operates on China’s space station; Western coverage of the announcement cites navigation resistant to jamming. Basis: [S180], [S153], [S178].	[INFER] Navigation for missiles, drones and submarines that satellite jamming cannot deny.	Demonstrated in orbit; fielding unconfirmed
4. Stealth-detecting radar	[FACT] Western coverage of the announcement cites air-defense radars able to detect stealth aircraft. The Pentagon’s own science board judged in 2019 that quantum radar offers no real benefit, and CSIS analysts advise skepticism, expecting hybrid niche systems at most. Basis: [S178], [S182], [S179].	[INFER] If real, pressure on stealth aircraft; more likely a quantum-enhanced component in a classical radar.	Contested; treat claims with caution
5. Magnetic anomaly detection of submarines	[FACT] The Defense Intelligence Agency reports Chinese investment in quantum magnetometers; a USTC, Hefei National Laboratory and China State Shipbuilding team reports a compact field-tested magnetometer intended for anti-submarine warfare. Basis: [S180], [S167].	[INFER] Pressure on the survivability of US and allied submarines in shallow and coastal waters.	Field-tested; no submarine trials published
6. Gravimetry	[FACT] The Defense Intelligence Agency reports Chinese investment in quantum gravimeters; CSIS notes they can reveal tunnels, shafts and large objects around the seabed. QuantumCTek reports its cold-atom gravimeter upgraded and placed on a first-unit procurement catalog in 2026, and a national standard for marine dynamic absolute gravimeters is in drafting. Basis: [S180], [S179], [S193], [S211].	[INFER] Detecting underground facilities and aiding navigation by gravity maps.	Research and early fielding
7. Quantum imaging	[FACT] The Pentagon reports Chinese industry and universities developing quantum imaging for intelligence, surveillance and reconnaissance; Chinese analysts cite imaging that is fast and resistant to interference. QuantumCTek reports a compact single-photon imaging radar in trial operation, a non-line-of-sight imaging prototype, and ultra-long-range single-photon imaging in testing. Basis: [S181], [S184], [S193].	[INFER] Better detection and identification through clutter and at low light.	Products in trial; research
8. Precision timing and	[FACT] The Pentagon includes positioning, navigation and timing among China’s	[INFER] Tighter coordination of sensors, jammers and	Research; strategic clocks

Possible tool	Evidence	Military meaning	Status
synchronization	quantum ISR applications; Chinese analysts note entanglement-based clock synchronization for C4ISR, radar, electronic warfare and weapons. Basis: [S181], [S185], [S78].	weapons, and timing that survives the loss of satellite navigation.	tied to national security
9. Swarm coordination	[FACT] Chinese analysts cite quantum protocols to keep drone and autonomous-vehicle swarms synchronized regardless of their number; CETC reports multi-domain swarm trials. Basis: [S185], [S169].	[INFER] Faster, harder-to-disrupt coordination of the swarms described in this section.	Concept; swarms themselves are in trials
10. Cryptanalysis and the harvest-now threat	[FACT] A “quantum cyber warfare” program implies interest in breaking encryption; NATO’s air-power center warns capable services already archive encrypted traffic for later decryption. Basis: [S176], [S150].	[INFER] Stored Western traffic becomes readable if China reaches a cryptographically relevant quantum computer first.	No evidence of such a computer; the threat is to archives

[FACT] US assessments frame the stakes. The Defense Intelligence Agency’s 2025 threat assessment reports China and Russia expanding quantum communication networks while investing in quantum magnetometers, gravimeters and inertial navigation. CSIS analysts warn that quantum sensing, if fielded at militarily relevant performance, could erode the survivability of submarines and stealth aircraft; a former US deputy national security adviser notes that BeiDou lets China jam GPS across a theatre while its own forces keep navigating. Basis: [S180], [S179], [S183].

[ARG] Read with the disclosure-gap thesis, a PLA statement that it is trialling more than ten experimental systems at the front, with quantum a priority, is the part Beijing chose to publish. The prudent assumption is that the program is broader than the announcement. Basis: [S186]; Section 8.

Fielded systems beyond quantum

[FACT] In an article on CETC’s own social-media account in September, its Intelligent Technology Research Institute called drone swarms an intelligent, distributed combat system, and set the longer-term goal as a combat network coordinating different uncrewed platforms across multiple domains; CETC’s Atlas system launches up to 96 drones under one operator. A 2026 Chinese Academy of Engineering study sets out the national research agenda for cross-domain swarms across air, space, ground, sea and underwater, including game-theoretic adversarial decision-making, and Shanghai state media expects the GJ-11 combat drone’s upgrade to add AI autonomous decision-making. Basis: [S169], [S170], [S171].

[FACT] State television showed a National University of Defense Technology swarm test in which one soldier directed some 200 drones. The Pentagon’s 2025 report assesses the WZ-9 drone as semi-operational and records two sixth-generation fighter prototypes in flight test since December 2024, known publicly only by analysts’ tentative designations. NORINCO markets a military humanoid robot. China launched a 26th Guowang satellite-internet batch on September 23. Basis: [S75], [S76], [S77], [S61], [S74].

Understatement as practice

[FACT] The Pentagon estimates China’s real military spending at 32 to 63% above its announced budget. The NSA, CISA and FBI advisory states that DeepSeek’s widely cited training cost excludes data obtained by distillation from US models. The Pentagon’s 2007 report read Deng Xiaoping’s “24-character strategy” — including “hide our capacities and bide our time” (韬光养晦) — as reflecting a

short-term desire to downplay China’s capabilities alongside a long-term build-up of power. In 2023 Meta attributed “Spamouflage,” the largest known cross-platform covert influence operation, to individuals associated with Chinese law enforcement. Basis: [S76], [S86], [S67], [S83], [S84].

[INFER] Official Chinese military channels are silent on quantum deployment, while budgets and costs are understated and prototypes go undisclosed and a state-linked influence operation has operated across more than 50 platforms. The silence on quantum is therefore not evidence of absence. Basis: [S76], [S67], [S77], [S84], [S2].

[INFER] A new documented case: in September 2026 China’s foreign ministry opposed the weaponization of space while Shanghai state media credited the PLA with anti-satellite equipment for a “space defense barrier,” and the ministry answered reports of Chinese imagery behind American deaths with a formula rather than a denial. Basis: [S174], [S171], [S162].

[ARG] The disclosure-gap thesis, stated as analysis rather than assumption: Chinese disclosures on quantum, AI and military readiness are an incomplete indicator of total capability. What we can observe is already compelling; what we cannot is a collection requirement, not a premise the argument depends on. Basis: This section; Report Two, KJ-3.

What Russian-language sources show

[FACT] Leaked documents from an unpublicized China–Russia military-technical forum, published in July 2026 by The Insider with Der Spiegel and Le Monde, concern air defense, drones and Starlink, not quantum technology. They are summarized in Appendix D rather than here. Basis: [S114].

[FACT] On May 20, 2026 Rosatom’s director-general Alexei Likhachev and Chinese Academy of Sciences president Hou Jianguo signed a memorandum covering fusion, nuclear medicine, accelerators, and photonic and quantum technologies. Likhachev later described the quantum cooperation as operating at the level of research institutes and of practical application, algorithm development and building the corresponding installations, expanding on the instruction of Russia’s leadership. Basis: [S115], [S116].

[INFER] The Chinese Academy of Sciences’ own account says the memorandum, signed on May 20, 2026 in the presence of Xi Jinping and Vladimir Putin, provides for joint research and experiments, personnel exchanges and academic workshops across controlled fusion, nuclear medicine, accelerators, and new photonic and quantum technologies (组织联合研究和实验、 开展人员交流). [S205] Report Two recorded the memorandum with implementation unobserved. It is now a stated intent to build quantum hardware together; implementation remains unobserved. Basis: [S115], [S116], [S2].

[FACT] Bloomberg reported on August 25, in coverage relayed in Russian-language outlets, on the journal China Military Science, describing Major General Zhu Xiaoqian, head of the PLA Army Engineering University, proposing that AI collect battlefield data, build the operational picture, generate options and automate some decisions in real time, and a companion paper with the National University of Defense Technology on human–machine military decision-making. A February 2026 CSET study of thousands of PLA procurement requests from 2023–2024 finds the PLA pursuing AI for decision support, sensing, targeting and data fusion across command, control and intelligence. Separately, a paper by an NUDT-affiliated team in Command Control & Simulation reports an AI command agent integrated into a platform supporting battalion-level operations. The PLA Daily has also argued that AI must remain a tool, with final decisions resting with human commanders. Basis: [S117], [S118], [S119], [S130].

[FACT] Russian outlets track China’s AI satellites closely: the Three-Body computing constellation’s first 12 satellites in May 2025 toward a planned network of more than 1,000; ADA Space’s 2,800-satellite plan, whose representative said future satellites could directly control autonomous systems, from humanoid robots to drones, where communications are absent; the state aerospace corporation’s plan for gigawatt-class orbital computing in 2026–2030; and a satellite launched in September 2026 whose onboard AI recognizes fires, storms and ships. Basis: [S120], [S121], [S122], [S123].

[UNDISC] No Russian-language source reviewed independently establishes operational Chinese military quantum-AI integration. The AI-computing constellations above are distinct from China’s planned quantum key-distribution constellation, and Russia’s own quantum research is separate from both. Basis: [S114]–[S123].

[UNDISC] Public sources show no military program work at Qingxing Yigou, QBoson, TuringQ or Fuxi Quantum. Report One’s caution applies: absence of public evidence is not evidence of absence in a managed-disclosure system.

1. Orbit and Counterspace

This section covers China’s quantum satellites past and planned, the ground segment built for them, the capabilities a deployed network would bring, and the contested orbit in which it would operate.

Where China’s quantum satellites stand

[FACT] A national standard for space quantum key distribution is under review: project 20251077-T-491, “Technical requirements for space quantum key distribution (QKD) systems” (空间量子密钥分发

（ QKD） 系统技术要求), issued on April 30, 2025 under the Chinese Academy of Sciences. The plan record lists five drafters — USTC, the Jinan Institute of Quantum Technology, the Shanghai Institute of Technical Physics, QuantumCTek and CAS Quantum Network — and no PLA institution. It is not yet a published standard. Basis: [S210].

[FACT] NUDT and the Information Engineering University are named on GB/T 47468-2026 and GB/T 47679.2-2026. [UNDISC] The space-QKD project’s original roster did not name them; whether PLA institutions were added at the review stage is a collection item, not a finding. Basis: [S198], [S206], [S210].

[FACT] China has flown two dedicated quantum-key satellites, and both have left orbit. Jinan-1, a sub-100 kg microsatellite with a design life of two years, fell into the South Pacific on October 17, 2025 after more than three years in orbit; Micius re-entered in January 2026. As of September 2026 the operating system is terrestrial: CN-QCN, a trusted-relay network of more than 10,000 km with 145 backbone nodes, 20 metro networks and six ground stations built to work with Jinan-1. Basis: [S132], [S2], [S133].

[FACT] The official project account published after Jinan-1’s re-entry said a constellation of three low-orbit satellites had entered a critical stage of development and was expected to launch the following year — that is, in 2026 — to form the prototype of a “quantum constellation,” and that the first medium-to-high-orbit quantum satellite was in development. Basis: [S135].

[UNDISC] As of September 27, 2026, no launch naming a quantum-key payload has been announced. Basis: [S135], [S74].

[FACT] China National Radio reported on June 30, 2026 that a quantum constellation of medium- and high-orbit quantum satellites and low-orbit microsatellites is being built at an accelerating pace, to

link with ground fiber into a global, round-the-clock network. Pan Jianwei said in March that his team is building the medium-to-high-orbit satellite with the National Space Science Center. Basis: [S73], [S78].

[UNDISC] The high-orbit satellite’s date has moved: a 2024 Qiushi article said it would be launch-ready by end-2026, while a 2024 Shanghai government headline said 2027. The same Qiushi article said two or three more low-orbit quantum microsatellites might launch in 2025; this report found no confirmation that any did. Basis: [S82].

[INFER] Three orbital stacks should not be confused. The dedicated quantum-key satellites: none declared in orbit as of September 2026; the Qianfan and Guowang constellations are classical broadband; the Three-Body and similar satellites are AI computing. Only the first distributes quantum keys. In orbit, what is publicly identified is empty; what is in procurement is not. Basis: [S132], [S133], [S2], [S120].

[FACT] The official Chinese account of the 2025 Jinan-1 result says quantum ground stations shrank from about 13 metric tons to under 100 kg, can be deployed rapidly in cities, mountains and plateaus, and can in principle support mobile quantum communication (原理上已可支撑移动量子通信). Basis: [S79].

[FACT] When QuantumCTek launched the first portable satellite ground station in 2021, it described the target users as those with high secrecy needs who cannot build fiber quantum networks, such as overseas missions, near- and far-sea islands, ocean-going ships and offshore oil and gas fields. (Those uses appear in the 2021 launch coverage, not on the current product sheet.) Basis: [S134].

[FACT] QuantumCTek’s current portable station, the QCGS-S280, is rated at 100 kg or less, at least 1 kbps of key in good weather (elevation above 20°, link loss up to 35 dB), a 20U cabinet, under 1 kW, and 12 hours to deploy with one operator after setup. Basis: [S188].

[FACT] In a late-2024 trial published by China Telecom and reprinted by SASAC, China combined “two satellites” — Jinan-1 for keys and a high-throughput communications satellite — with “three networks”: 5G mobile, the quantum-satellite key network and high-throughput satellite broadband, using ChinaSat-26 and APSTAR-6D. It claimed the first quantum-encrypted 5G voice calls, messaging and data, delivering keys to portable stations where fiber does not reach. Basis: [S187].

[INFER] Keys from a quantum satellite, traffic over a communications satellite and 5G at the edge is the architecture a military would use to secure field and maritime communications; the trial proved it, and its quantum satellite has since re-entered. Basis: [S187], [S132].

[INFER] No military user is named. Rapidly deployable stations for mountains, plateaus, islands and ships are, however, a field-use profile. Basis: [S79], [S134].

The next quantum constellation

[FACT] Parts of the next orbital quantum layer are visible in procurement. In December 2025 the CAS Innovation Academy for Microsatellites advertised an open tender for a solar-array drive assembly for a medium-high-orbit quantum science experimental satellite (中高轨量子科学实验卫星); on December 23 it terminated the tender because fewer than three firms bought the documents and said it would proceed by competitive negotiation. No award notice was found. On September 23, 2026 the Aerospace Information Research Institute of the Henan Academy of Sciences posted an open tender for the development of a “quantum remote-sensing fusion” microsatellite platform and payload integration; as of September 26 the listing showed only the invitation, with bid opening, award and contract blank, and the public page carries no payload specification; Henan’s own 15th

Five-Year Plan calls for laying out quantum communication and satellite internet ahead of demand.

Basis: [S200], [S201], [S204].

[INFER] The pipeline is being built in the open. That is stronger evidence than any assumption about undeclared payloads, though neither tender by itself establishes a launch date or mission. Basis: [S200], [S201].

[FACT] The program is being rebuilt at pace. In January 2026, in a Xinhua piece on the national-leaders page, Peng Chengzhi of the Micius team said his team would continue to prepare “all-out” for the launch of a new medium-high-orbit quantum satellite (新一年， 我们将继续为新型中高轨量子卫星的发射全力准备). The project account after Jinan-1’s re-entry said three low-orbit satellites would launch in 2026. No launch has been announced as of September 26. Basis: [S138], [S135].

[FACT] Premier Li’s 2026 Government Work Report named quantum technology a priority future industry, made aerospace a new pillar industry and, for the first time, ordered faster development of satellite internet. China has filed ITU applications for 203,000 satellites and plans more than 10,000 low-orbit satellites by 2030. Jinan-1’s scientists have said China is considering integrating quantum encryption with its satellite-internet plans, and the CAS institute that built Jinan-1’s platform is led by the chief commander of the Qianfan constellation. Western analyses report that the 15th Five-Year Plan calls for expanding the ground-to-space quantum network, and Pan has said commercial quantum services with China Telecom could arrive as early as 2027. Basis: [S157], [S142], [S79], [S143], [S151], [S144].

[FACT] Officially, the 15th Five-Year Plan puts quantum technology first among the future industries to become new growth points, and the industry ministry’s September 2026 plan for the communications sector calls for building an integrated space-ground network and accelerating the fusion of quantum encryption with communications networks (量子加密与信息通信网络融合). Basis: [S190], [S191].

[UNDISC] Western analyses, and one reviewer of this report, place the specific phrase “build an integrated space-ground quantum communication network” in the Plan’s quantum section; this report could not locate it in the published text, so it remains attributed. Basis: [S151].

[FACT] China Telecom — itself on the Pentagon’s 1260H list — has made the commitment in its own name. Its Quantum Science and Industry 2030 Action Plan sets ten tasks to the end of the 15th

Five-Year Plan period, including to basically complete a nationwide space-ground integrated quantum communication network (基本建成覆盖全国的天地一体量子通信网络) and to build a new cryptographic system fusing quantum key distribution with post-quantum algorithms. SASAC, which oversees China Telecom, has told state enterprises to target quantum technology to create “asymmetric” and “assassin’s mace” (杀手锏) technologies. Basis: [S194], [S195], [S29].

[FACT] China launches communication payloads it describes only vaguely. TJS-25 (June 2026) was announced as a communication-technology validation satellite with no images or further information; a pair, TJS-27A and TJS-27B, launched from Taiyuan on a Long March 6A on July 30, 2026 into an orbit of roughly 1,160 km, described officially only as communications-technology demonstration satellites and not declared as quantum payloads. And Chinese quantum space work has been disclosed late before: a 2013 satellite-to-ground quantum experiment was published years after the satellite used had deorbited. Basis: [S139], [S140], [S141].

[INFER] The mandate, the builders, the stated schedule and the disclosure pattern all point the same way: China is rebuilding its orbital quantum capability faster than it is disclosing it. Basis: [S138], [S157], [S143], [S139], [S141].

System	Status	Key specifications	Source
Micius (2016–2026)	Re-entered	More than 600 kg; entanglement, key distribution and teleportation; keys processed over days; ground stations of about 13 metric tons	[S2], [S79]
Jinan-1 (2022–2025)	Re-entered	Payload about 23 kg, an order of magnitude lighter than Micius; light source about six times faster; real-time keys; 250,000 to 1 million secure bits per pass; about 3,000 bits per second on average; ground stations under 100 kg	[S79], [S132]
Three low-orbit microsatellites	Promised for 2026; none announced	Prototype of the “quantum constellation”; candidates to ride on satellite-internet buses	[S135], [S142]
Medium-high orbit (~10,000 km)	Team preparing “all-out” since January 2026; launch undated; component tender December 2025	Key distribution through the whole day; entanglement-based keys between stations 10,000 km apart; micro-vibration control the main engineering hurdle	[S138], [S146]
“Dawn” (geostationary)	Announced, about 2027	Above 35,000 km; entanglement distribution beyond 10,000 km; continuous 24-hour coverage	[S145]
Direct quantum messaging (QSDC)	Components flight-tested	Sends the message itself through the quantum channel, not just a key; tested by an Entity-Listed consortium member	[S147], [S21]
Device-independent keys and repeaters	Demonstrated on the ground, 2026	Keys over 100 km that do not depend on trusting the hardware; a building block for scalable repeaters	[S148]

The team behind China’s quantum satellites

Scientist	Role	What the profile shows	Source
Pan Jianwei	Chief scientist of the quantum satellite program; USTC executive vice president	Announced the “Dawn” geostationary satellite; described by a Yale analysis as having links to Chinese defense contractors. USTC is Entity-Listed.	[S145], [S149], [S21]
Peng Chengzhi	Chief of Micius’s science application system and deputy chief of the satellite system; led its payload	Said in January 2026 his team was preparing “all-out” for the next launch. Formerly chairman of QuantumCTek (Entity-Listed since 2021), which he took public and which Zhang Wenqiang now chairs; now China Telecom’s chief scientist, and elected to the Chinese Academy of Sciences in November 2025.	[S138], [S155], [S189], [S192]
Liao Shengkai	USTC; space electronics and quantum communication systems; Jinan-1 lead	Said China is considering integrating quantum encryption with its satellite-internet plans.	[S160], [S142]
Wang Jianyu	CAS academician; executive deputy chief designer and satellite chief commander of Micius; former director, Shanghai Institute of	An optical remote-sensing specialist; named low-orbit key networking and medium-high-orbit platforms as the main directions.	[S156], [S146]

Scientist	Role	What the profile shows	Source
	Technical Physics
Hu Haiying	President, CAS Innovation Academy for Microsatellites	Chief commander of the Qianfan satellite system; his academy built Jinan-1’s platform.	[S143], [S79]
Zhan Mingsheng’s team	CAS precision-measurement institute, Wuhan, with Hefei National Laboratory	Built the cold-atom gyroscope operating on China’s space station.	[S153]

[INFER] The people who built Micius and Jinan-1 are preparing the next launches, and they sit at the junctions of the national quantum laboratory, the constellation builder and an Entity-Listed company. That is a team positioned to move fast. Basis: [S138], [S143], [S155], [S21].

Quantum sensors already in orbit

[FACT] China already operates quantum sensors in space. Its space station has carried, since November 2022, the first operational cold-atom gyroscope in orbit, which its builders describe as a step toward a space quantum inertial measurement unit for navigation when satellite navigation is unavailable. The Mengtian module carries an ultracold-atom physics cabinet, a high-precision time-frequency cabinet and a space cold-atom clock. Basis: [S153], [S154].

What a fully deployed network could do

[INFER] Each capability below is grounded in a sourced fact; its meaning for China’s national security, defense and intelligence community is the authors’ inference. Basis: Section 9 sources as cited.

 

 

 

   

 

 

   

 

Orbit is already contested

 

1. Part One Actions

[ARG] Report One proposed five charter asks for the AI Force and Report Two nine actions (R-1 to R-9). This report proposes eight more, numbered Q-1 to Q-8; none repeats the earlier fourteen.

Throughout this report, R-numbers refer to Report Two’s actions for the Administration, not to the NATO recommendations in Report One’s annex. Six belong to Part One and name no company (Q-2, Q-3, Q-4, Q-6, Q-7 and Q-8); two belong to Part Two (Q-1 and Q-5). Basis: [S1], [S2].

Q-2. Name quantum-inspired AI. Commerce and the Department of Defense should track quantum-inspired AI as a named category and screen the companies, funds and consortium members in this report against the Entity List and the June 2026 Section 1260H list. The consortium lead, CAEIT, is already Entity-Listed but is not named on the 1260H list, which lists its parent; the remaining members and funds around it have not been screened in public. Owner: Commerce (BIS) and the Department of Defense. 90-day artifact: a one-page screening note on quantum-inspired AI that sets out CAEIT’s Entity List status against its parent’s 1260H listing and covers every consortium member. Authority: the Export Control Reform Act of 2018 and EAR §744.11 (Entity List); Section 1260H of the FY2021 NDAA. Success measure: every consortium member and fund screened, and a published determination on whether CAEIT should be named in the next 1260H update.

Q-3. Vendor diligence and customer vigilance. US chip, cloud and software vendors should review supply relationships with Chinese quantum-AI companies, including relationships those companies report themselves, framed as diligence rather than accusation. Owner: Commerce (BIS) outreach, with the vendors. 90-day artifact: an industry advisory on quantum-AI end-use questions. Authority: the Export Control Reform Act of 2018 and BIS “Know Your Customer” guidance (Supplement No. 3 to EAR Part 732). Success measure: the advisory issued, and the major US chip and cloud vendors confirming they have reviewed quantum-AI relationships in China.

Customer vigilance. Commerce (BIS) and CISA should also advise US companies, universities and laboratories on the risks of running sensitive workloads on quantum cloud services operated by 1260H-listed or Entity-Listed Chinese firms, since problem formulations and application know-how can leave with the job. [S228, S229, S29]

Q-4. Keep the yardstick American. NIST and DOE should fund open quantum-AI benchmarks and datasets — calibration, sensing and optimization tasks among them — so that the standards Chinese firms certify against remain American-set. This differs from Report Two’s R-6, which benchmarks Chinese machines. Owner: NIST with DOE. 90-day artifact: public calibration-plot and radar change-detection benchmark sets, each with classical baselines, reproducibility rules and cost reporting, open to any vendor’s submission. Authority: NIST’s measurement and standards mission (15 U.S.C. 272) and the National Quantum Initiative Act of 2018. Success measure: two benchmark sets public within 90 days and cited in independent published results within a year.

Link to Quantum First 2030. Q-1 and Q-4 complement the Commission’s Quantum First by 2030 recommendation and its proposed Quantum Software Engineering Institute. Beijing’s integration center puts the software ecosystem first, which raises the Institute’s urgency. [S214, S213]

Q-6. Update the baseline. Congress should ask the U.S.-China Economic and Security Review Commission for a staff update, or a hearing, on China’s quantum program that draws on the documentary sources identified in this report: standards-agency rosters, procurement listings, municipal notices and the Mandarin 1260H list. Owner: the Commission, at congressional request. 90-day artifact: an updated staff assessment. Success measure: the outdated statements identified in “The Baseline Has Moved” are corrected in the record Congress uses, and the Quantum First goal is extended from computational advantage to quantum networks, space and sensing readiness — the domains where China’s documented lead lies — in time for the National Quantum Initiative reauthorization. [S214]

Q-7. Watch the standards. NIST and the State Department, with NATO, should track Chinese national quantum standards — including their drafting rosters — and Chinese submissions to ISO/IEC JTC 3 and ITU-T, since PLA institutions appear on the national QKD-security drafting roster

and SIPRI identifies standards as a strategic battleground. 90-day artifact: a standing allied watch list. [S206, S210, S216]

Q-8. Secure the anomaly data. The Department of Defense, with NATO allies, should protect and complete the gravity and magnetic anomaly datasets on which quantum navigation depends, now that China has registered a national standard for marine dynamic absolute gravimeters tied to its positioning, navigation and timing system. 90-day artifact: an allied inventory of anomaly-data holdings and gaps. [S211, S212, S216]

Part Two — The American Answer: IonQ and NVIDIA

What This Means for IonQ

[ARG] Part One showed what China is building. This page sets out what that means for IonQ: the openings that exist now, the result that shows its acquisitions working as one company, and the numbers that carry the case. Basis: Part One; [S2].

 

The openings

Opening	What it is	Why IonQ fits	Timing
QC-ADDS (EO 14413)	[FACT] A federal effort to deliver a scientifically useful quantum computer to a DOE facility. Basis: [S49].	[ARG] The NVAQC configuration is the closest existing model of a QPU installed beside AI supercomputing. Basis: [S30].	National Quantum Strategy update due within 180 days of June 22, 2026
DOE Quantum Genesis Q Competition	[FACT] Up to $215 million, milestone-based, for systems of at least 100 logical qubits. Basis: [S2].	[ARG] IonQ’s published decoder and Walking Cat design address the fault-tolerance milestones directly. Basis: [S36], [S2].	Applications close October 19, 2026
AI Force infrastructure	[FACT] Report One’s five charter asks for the infrastructure beneath AI. Basis: [S1].	[ARG] A US-fabricated QPU on an open link to AI racks fits a charter built on trusted infrastructure. Basis: [S1], [S31].	Charter not yet published
Allied quantum-GPU centers	[FACT] NVQLink adopters in Japan, Korea, Taiwan, Singapore, Australia and Europe. Basis: [S41].	[CO-STATED] IonQ already works with KISTI and SDT in Korea and Cambridge in the UK; Report Two records its Q-Alliance ties in Italy. Basis: [S39], [S35], [S40], [S2].	Now
Allied quantum-safe space links	[FACT] Europe’s first quantum-key satellite has slipped to late 2027 or early 2028, while China prepares new quantum satellites. Basis: [S152], [S138].	[ARG] IonQ has announced a space-to-ground key network and fields post-quantum products alongside it. Basis: [S2], [S105].	Now

Opening	What it is	Why IonQ fits	Timing
Quantum-AI for overhead sensing	[FACT] IonQ’s Capella unit holds a National Reconnaissance Office radar contract. Basis: [S50], [S2].	[ARG] The radar result shows a quantum-AI analytic on the same data stream. Basis: [S38].	Now

Integration made real

[INFER] Report Two set IonQ a second-stage test: whether its acquired businesses produce, together, something none could produce alone. The radar paper is the first documented pass. Capella’s engineers supplied the satellite radar and co-authored the work; IonQ’s quantum team supplied the generative model; the result ran on IonQ’s own trapped-ion hardware. The Superion 256 chip is a second: Oxford Ionics’ on-chip qubit control, fabricated in SkyWater’s trusted foundry. Basis: [S38]; [S31]; Report Two’s integration test [S2].

[ARG] For IonQ’s investors and partners, this is the more important signal of the month. Assembling a stack takes capital; integrating it takes execution, and the evidence of execution is now on the record. Basis: [S2], [S38].

1. The Announcement and the Stack

[FACT] What the announcement does and does not say. IonQ’s NVAQC announcement describes the first on-premise QPU at NVIDIA’s quantum research center; it does not publicly disclose a sale, ownership of the system, payment terms, IP allocation or exclusivity — unlike the explicit commercial language for SDT and FIU. Quantinuum, also a trapped-ion company, is already inside NVIDIA’s quantum ecosystem. IonQ’s established distinction is being first on premise, not trapped-ion exclusivity. Basis: [S30], [S35], [S37], [S41].

 China is building quantum-AI integration as a national stack. On September 23 the United States put its answer on a calendar: a US-fabricated quantum processor wired directly into America’s leading AI rack.

[FACT] IonQ will be the first on-premise quantum deployment at NVIDIA’s Accelerated Quantum Research Center (NVAQC). A Superion 256 system will be linked directly to an NVIDIA GB200 NVL72 through NVQLink, with workloads orchestrated by the open CUDA-Q platform, and installation scheduled for 2027. The program covers hybrid software development and large-scale system prototyping, and will publish open results and a guide for AI use cases and quantum-GPU co-design. Research targets include portfolio optimization and risk modeling, materials science, and chemistry for drug discovery.

[CO-STATED] NVIDIA’s quantum general manager, Timothy Costa, said every supercomputer will become a quantum supercomputer.

Layer	Component	What the record shows
Quantum processor	IonQ Superion 256 (sixth generation)	[FACT] 256-qubit trapped-ion QPU with electronic qubit control on the chip, fabricated at SkyWater. First customer deployments early 2027.
Interconnect	NVIDIA NVQLink	[FACT] Open architecture over RDMA on commercial 400 Gb/s Ethernet. QPU–GPU round trip under 4 µs (3.96 µs maximum measured). Real-time callbacks let one C++ program address

Layer	Component	What the record shows
		CPU, GPU and FPGA subsystems.
Classical AI rack	NVIDIA GB200 NVL72	[FACT] 36 Grace CPUs and 72 Blackwell GPUs in one liquid-cooled rack; a 72-GPU NVLink domain with 130 TB/s of GPU communication.
Orchestration software	NVIDIA CUDA-Q	[FACT] Open platform exposing NVQLink’s real-time APIs for calibration, error correction and hybrid applications.
Error-correction decoding	IonQ real-time decoder	[FACT] Decodes simulated workloads of up to 408 logical qubits on a single CPU with under 0.3% delay (analyzed in Report Two).
Scaling path	Superion 10K	[CO-STATED] CMOS integrated onto the chip for about 10,000 qubits; cost per qubit expected to fall more than 300-fold across the roadmap.

Competing QPU on NVQLink	What the record shows
Quantinuum Helios (trapped ion)	[FACT] Used NVQLink to demonstrate real-time decoding with a 67-microsecond reaction time.

[ARG] If Q-1 is the architecture, IonQ is an early anchor of it, not the architecture itself. Basis: [S30], [S41].

1. The Timing Budget

[FACT] Trapped-ion error-correction cycles run at about 1 to 5 milliseconds. NVQLink’s round trip is under 4 microseconds. Superconducting systems work to a cycle budget of about 1 microsecond.

 

Figure 4. Link latency against error-correction cycle times (log scale). Source: NVIDIA and HPCwire (NVQLink); IonQ, arXiv:2608.25027 (trapped-ion cycle); arXiv:2605.12046 (superconducting cycle budget).

[INFER] The arithmetic is the argument, and it is arithmetic: it compares a published interconnect latency with a published cycle range, not a measured correction loop on an installed system. The engineering proof will be end-to-end measurement of acquisition, decoding, communication, feedback and jitter on the NVAQC installation. A 4-microsecond round trip uses 0.4% of a 1-millisecond trapped-ion cycle and 0.08% of a 5-millisecond cycle. On a trapped-ion machine the link

is effectively free, so the whole GPU rack is available inside every error-correction cycle. On a superconducting machine the same round trip exceeds the whole cycle, which pushes the fastest decoding loops toward FPGA or custom hardware beside the chip. Basis: [S42], [S36], [S44].

[FACT] China is working the same systems problem. The industry ministry’s 2025 national “unveil the list” tasks (工信厅高新函〔 2025〕 21 号) call for quantum control systems with low signal latency, fast feedback and real-time error correction, and for real-time partial measurement and feedback control in error-correcting codes, targeting surface-code distance of at least 5 by 2026. In December 2025 Pan Jianwei’s team reported below-threshold error correction at surface-code distance 7 on the Zuchongzhi 3.2 processor, and in May 2026 Chinese industry researchers published a six-layer real-time QEC reference architecture that uses NVQLink as its interface case study and includes a trapped-ion implementation. Basis: [S199], [S202], [S207].

[INFER] China has identified the same end-to-end loop — acquisition, transport, decoding, feedback

— that IonQ and NVIDIA are attacking. That is evidence of the same problem being worked, not of equivalent maturity in QPU–GPU integration. Basis: [S199], [S202], [S42].

[INFER] Report Two showed that IonQ’s decoding fits on one conventional CPU. Put that together with the timing budget and the consequence is new: the 72 Blackwell GPUs are not needed for decoding. They are free for the work that makes the pairing valuable — AI-generated circuits, learned calibration, and hybrid algorithms that alternate between QPU and GPUs inside each cycle. Basis: [S2], [S36], [S43].

1. Three Weeks of Quantum-AI Evidence

[FACT] IonQ’s releases between 8 and September 24 show the quantum-AI business taking shape around the NVIDIA deployment.

Date	Announcement	Why it matters for the quantum-AI lane
Sep 8	Superion 256 launched; orders open	[FACT] The QPU that goes to NVAQC; first integrated 256-qubit chips fabricated at SkyWater.
Sep 14–15	Nine IEEE Quantum Week papers; four Best Paper Awards from 857 submissions	[FACT] Peer-reviewed quantum fine-tuning of foundation models with lower energy use and a 24% error reduction in AI classification (with QuantumBasel).
Sep 16	With ORNL, NVIDIA and the University of Tennessee: generative AI writes quantum optimization circuits (DQAOA-GPT)	[FACT] AI for quantum — the workload type the GPU rack will run beside the QPU.
Sep 17	Computer-aided engineering workloads accelerated by up to 14.6% (with Synopsys)	[FACT] Meshes of up to 35 million elements; simulated in CUDA-Q to 150 qubits and validated on the 36-qubit Forte.
Sep 21	Strategic partnership with SDT, South Korea	[CO-STATED] Superion 256 and a quantum memory module for a planned hybrid quantum-classical data center; an assembly and integration facility in Gumi.
Sep 22	Real-time error decoder on one CPU	[FACT] Frees the GPU rack for AI work (Section 12); full analysis in Report Two.
Sep 23	First on-premise QPU at NVIDIA NVAQC	[FACT] The reference architecture this paper proposes.
Sep 24	Florida International University buys a Superion 256	[CO-STATED] Installation expected late 2027; research planned in AI, cybersecurity, grid management and

Date	Announcement	Why it matters for the quantum-AI lane
		materials.
Sep 24	Quantum generative modeling for radar change detection (with Capella Space)	[FACT] Quantum-AI on Capella satellite radar, run on IonQ hardware (Section 14).

1. The Radar Result

[FACT] The preprint SAR and InSAR Change Detection with Quantum Generative Models (arXiv:2609.05313, Sekwao et al., September 4, 2026) replaces the classical background estimator used in radar change detection with a quantum circuit Born machine (QCBM). Its authors are drawn from IonQ and from Capella Space, described as an IonQ company, and it runs on Capella satellite acquisitions. The SAR test site is Marine Corps Air Station Miramar in San Diego.

Setting	QCBM	Classical baselines	What it shows
Ideal simulator, original data	0.30	NLCD 0.14; Copula 0.20	Advantage on sparse, heavy-tailed pixel statistics
Ideal simulator, median-filtered (Med3)	0.41	NLCD 0.16; Copula 0.24	The largest simulated gap
Ideal simulator, Yeo–Johnson median-filtered (near-Gaussian) data	0.56	NLCD 0.55; Copula 0.56	Advantage confined to non-Gaussian data
IonQ hardware: training and inference on the QPU (Med3-style)	0.32	0.16; 0.24	The abstract’s headline: a feasibility result on Forte and Forte Enterprise
Simulator training, QPU inference	0.37 (vs 0.40– 0.41 ideal)	—	Modest hardware noise penalty
QPU training, ideal inference	0.32	—	Training on hardware costs more than inference on it

Reading the scores. [FACT] The main comparison table uses ideal-simulator training and inference. The 0.32 filtered F1 in the abstract is the full-hardware figure. The paper presents it as a feasibility demonstration that the estimator runs on trapped-ion hardware with a modest noise penalty. IonQ’s September 24 release did not distinguish the simulator table from the QPU 0.32 figure, and did not name Capella; this section does.

[INFER] Two points matter. First, this is a documented cross-unit result of the kind Report Two set as the test of integration: Capella’s radar and IonQ’s quantum models, with authors from both. Second, it is the closest American analytic counterpart to the on-orbit target identification that Report Two records China demonstrating on September 20. Basis: [S38], [S2].

[INFER] The honest bound: one preprint, one airport site, and an advantage confined to sparse, non-Gaussian data; on the volcanic InSAR dataset all three methods reached about 0.66. On hardware the model still doubled one baseline and beat the other by a third. It supports feasibility and a task-specific gain, not an operational ISR advantage. The consequential next tests are a second site, an evaluation protocol set in advance, and a full cost and latency comparison. Basis: [S38].

15. The Operational Case and Allies

• Customers before installation. [FACT] Superion 256 has been sold to the University of Cambridge. [CO-STATED] IonQ reports sales to SDT for a Korean customer and to Florida International University. NVAQC and first customer installations are scheduled for 2027.
• Korea, on the government record. [FACT] Korea’s science ministry has announced the Hangang–Tempo link; the full sequence is in the playbook table above. Basis: [S93], [S94].
• Japan, already running. [FACT] RIKEN installed ROQUO in June 2026: 540 NVIDIA Blackwell GPUs linked through an interface built on CUDA-Q to Fugaku, IBM’s ibm_kobe and Quantinuum’s trapped-ion Reimei. [INFER] Japan’s trapped-ion hybrid path runs on Quantinuum hardware, not IonQ’s: the allied architecture is spreading, and IonQ must compete within it. Basis: [S95].
• An allied quantum-GPU footprint. [FACT] KISTI is among NVQLink’s international adopters, alongside RIKEN and AIST in Japan, Taiwan’s NCHC, Singapore’s national hub, Australia’s Pawsey center and CINECA in Italy.
• Security fit. [INFER] A QPU fabricated in a US trusted foundry, on an open allied interconnect, meets the provenance tests Reports One and Two set for the infrastructure layer and the Trusted Quantum Stack. Basis: [S1], [S2], [S31], [S41].

The IonQ playbook: sovereign quantum-AI nodes

[INFER] IonQ’s announcements follow a repeatable pattern. IonQ has not described it this way; the pattern is the authors’ reading, and each step below is on the record. IonQ places its processor inside a nation’s own AI supercomputing, through that nation’s lead institute, with a local partner, on NVIDIA’s open interconnect. Basis: [S30], [S93], [S94], [S35], [S41].

 

Date	Korea, step by step	Source
May 2025	IonQ–KISTI memorandum: infrastructure access, talent, and integration with KISTI’s supercomputing	[FACT] [S104]
Dec 2025	Agreement to deliver the 100-qubit Tempo into KISTI-6 (Hangang); IonQ named primary provider with Megazone Cloud	[FACT] [S94]

Date	Korea, step by step	Source
Mar 2026	NVIDIA–IonQ–KISTI memoranda announced by Korea’s science ministry; Hangang–Tempo link planned by year-end	[FACT] [S93]
Sep 2026	SDT partnership: Superion 256, quantum memory, Gumi assembly, hybrid quantum-classical data center	[CO-STATED] [S35]

 

Figure 5. The sovereign quantum-AI node, as assembled in Korea and the United States. Source: [S30], [S35], [S41], [S93], [S94], [S104], [S105]. The pattern is the authors’ reading.

[ARG] This is the democratic counterpart to China’s model. Beijing builds one national stack by state direction; IonQ builds allied national stacks by partnership, each owned by the host nation and joined by an open interconnect. Basis: Part One; [S41], [S93].

Both layers of quantum-safe security

[FACT] IonQ, through ID Quantique, has deployed national QKD networks in Romania (36 links, more than 1,500 kilometers) and Slovakia under the EU’s EuroQCI program. Basis: [S231], [S232].

 

Why it matters against China

 

Figure 6. The two integration paths, layer by layer. Source: Parts One and Two of this report; [S14], [S15], [S17], [S18], [S23], [S24], [S30], [S31], [S41], [S43].

[ARG] The American advantage is not that it has a stack and China does not. China has one. The advantage is that the American approach is open, allied, fabricated at home and built around the AI hardware the world already runs. Deployed first, it becomes the default everyone else must connect to. Basis: [S14], [S15], [S31], [S41].

Where IonQ leads, and where it does not yet

[FACT] Superion is designed as an upgradeable platform: IonQ says all future compute products are expected to be built on it, and Superion 10K is being developed alongside Superion 256. A system occupies a standard server rack, draws less power than a rack of GPUs and cools through ordinary data-center plumbing. Its processors are fabricated at SkyWater, a DMEA Category 1A Trusted Foundry. Basis: [S234].

[INFER] For a sovereign or federal buyer, that changes the decision: not a one-off machine that may be a dead end, but an entry point to a line that upgrades on a semiconductor roadmap in a trusted US foundry. Basis: [S234].

Dimension	IonQ	Quantinuum
Position on NVQLink	[FACT] First on-premise QPU at NVAQC;	[FACT] Already demonstrated real-time

Dimension	IonQ	Quantinuum
	installation 2027. Basis: [S30].	decoding over NVQLink, with a 67-microsecond reaction time. Basis: [S41].
Fabrication	[FACT] Owns SkyWater, an accredited US trusted foundry, where Superion 256 chips are made. Basis: [S31], [S2].	[FACT] Traps from Honeywell, adding GlobalFoundries. Basis: [S2].
Quantum-AI evidence on own hardware	[FACT] Radar change detection on Capella data; peer-reviewed quantum-AI results with national laboratories. Basis: [S38], [S32], [S33].	Not assessed in this report.
Allied deployments	[FACT] Korea (KISTI–Hangang, on the Korean government record); UK (Cambridge). [CO-STATED] SDT; FIU. Basis: [S93], [S94], [S40], [S35], [S37].	[FACT] Japan: Quantinuum’s Reimei is linked to RIKEN’s ROQUO and Fugaku. Basis: [S95].
Compute benchmarks	Leads on some; not on all.	[FACT] Report Two records that Quantinuum, IBM and Google each lead on particular compute benchmarks. Basis: [S2].

[INFER] IonQ’s lead is in owned, trusted fabrication, the NVAQC position, the Capella data and allied reach. Quantinuum was first to show decoding over the same link. An informed reader should weigh both. Basis: [S2], [S30], [S38], [S41].

What it is not

• [FACT] IonQ is not NVQLink’s only quantum partner. Quantinuum’s Helios used NVQLink for real-time decoding with a 67-microsecond reaction time. IonQ’s distinction is the first on-premise QPU at NVAQC.
• [FACT] Installation is scheduled for 2027. No production workload has yet run on this configuration.
• [INFER] CUDA-Q and NVQLink are open, so Chinese groups can study and emulate them. The durable US advantages are the hardware on each side of the link and the allied deployments around it. Basis: [S41], [S42].

1. The Tests Ahead, and What IonQ Should Do Next

[ARG] The authors set these tests in advance. Any one of them would weaken the case for Q-1 and for IonQ’s role in it.

 

What IonQ should do next

[ARG] Each of the four tests can be turned into a milestone IonQ controls. Basis: Section 16 tests; [S30], [S35], [S37], [S38].

1. Publish a hybrid result that uses the NVIDIA GPU rack inside an error-correction cycle, on IonQ hardware, before NVAQC installation.
2. Replicate the radar result at a second site and on a larger QPU, with an outside co-author.
3. Put the SDT and FIU contracts on the record in SEC filings, so they move from company statements to documented fact.
4. Submit results to the open benchmark sets proposed in Q-4 as soon as they exist.
5. Publish NVAQC installation milestones, so slippage or progress against the 2027 schedule can be tracked.

What the reviewed assessments mean for the American answer

The assessments reviewed in “The Baseline Has Moved” were written to describe China. Read together, they also describe what an American answer has to cover. This subsection sets out where IonQ’s announced position aligns with that answer, and where it does not.

[ARG] The Commission’s Quantum First goal asks for scalable quantum computing, secure quantum communications and faster migration to post-quantum cryptography, plus a Quantum Software Engineering Institute. IonQ is the one American company whose announced portfolio spans all four: trapped-ion processors fabricated in the United States, quantum key distribution and post-quantum cryptography through ID Quantique, and an open integration path to NVIDIA’s AI systems through NVQLink. Extending Quantum First to networks and space, as Q-6 proposes, would widen that alignment rather than create it. Basis: [S223], [S30], [S105], [S106], [S41].

[INFER] RAND identified America’s alliance network and the strength of its private industry as advantages China lacks, and flagged two risks: the financial health of small quantum firms and a supply chain dependent on European and Japanese components such as dilution refrigerators and specialized lasers. IonQ’s allied sovereign-node deployments are a working instance of the alliance advantage; its scale reduces the small-firm risk; and trapped-ion systems do not need the millikelvin dilution refrigerators that superconducting machines depend on — although they do depend on precision lasers, which keeps that part of RAND’s concern open. Basis: [S222], [S93], [S94], [S95].

[INFER] ASPI finds China now leads research on small satellites and that the United States still employs the largest share of top-tier technology talent. Converting that talent into fielded capability is the American answer’s job; IonQ’s ownership of Capella Space gives it a US small-satellite radar constellation, the platform for the early quantum-enabled radar result in Part Two. Basis: [S225], [S38].

[INFER] MERICS records CETC helping build Pakistan’s national quantum computing center — China exporting quantum capacity to a partner military. The allied counter-model is a trusted, sovereign quantum node inside an ally’s own supercomputing, the pattern Korea has announced with IonQ and that RIKEN runs in Japan with Quantinuum hardware. NATO’s own trends report places the AI–quantum race at the center of Alliance competition; quantum–GPU integration on an open interface is where the two meet. Basis: [S224], [S93], [S95], [S226].

[INFER] China is backing trapped ions too. The Hefei fusion center includes a trapped-ion machine, and Hefei Yaozheng Quantum, incubated by Guo Guangcan’s team, describes itself as China’s only

QCCD trapped-ion company, with six systems built; Anhui investors say China cannot risk a strategic miss on the trapped-ion route. That validates the architecture IonQ is pursuing, and it means the competition on that route is real. Basis: [S236], [S237].

[ARG] None of this makes IonQ the whole answer. Other American and allied firms lead in other modalities and applications, Quantinuum is already inside NVIDIA’s ecosystem, and IonQ’s 2027 milestones are still ahead. The case is narrower and stronger: the gaps the reviewed assessments leave open are ones IonQ’s announced portfolio is positioned to fill. Basis: [S37], [S30].

1. Part Two Actions

[ARG] The two actions that name the American architecture. Competing platforms should be able to qualify under both.

Q-1. A quantum-GPU reference architecture for federal AI infrastructure. DOE and national-laboratory hybrid work should adopt the NVAQC pattern — a QPU co-located with AI supercomputing over an open interconnect — as the reference architecture, and AI Force infrastructure planning should reserve a place for it. This complements Report Two’s R-4 by specifying how the compute element of an integrated stack connects to AI supercomputing. This paper proposes IonQ as an early anchor, given its NVAQC deployment, trusted-foundry fabrication and peer-reviewed quantum-AI record with national laboratories; competing platforms should qualify against the same architecture. Owner: DOE Office of Science with the national laboratories. 90-day artifact: a published reference-architecture note and a call for hybrid pilot proposals that sets representative workloads, classical baselines, total operating costs, reproducibility rules and security requirements, and lets multiple QPU vendors qualify. Authority: the National Quantum Initiative Act of 2018 and DOE Office of Science program authority, consistent with EO 14413 of June 22, 2026, which creates the QC-ADDS effort to deliver a scientifically useful quantum computer to a DOE facility [S49]. Success measure: at least two national-laboratory hybrid pilots running on the architecture within twelve months, with at least two qualified QPU vendors and published results.

Keep it open. [ARG] Q-1 and Q-5 should be GPU-agnostic as well as QPU-agnostic, so that the reference architecture and allied interface apply to AMD- as well as NVIDIA-based federal systems and to every qubit modality. Speed matters, but so does keeping options open.

Q-5. An allied quantum-GPU interface. State and DOE should agree a common quantum-GPU interface and security baseline with NVQLink adopters in Japan, Korea, Taiwan, Singapore, Australia and Europe, delivered through companies already building there — IonQ’s KISTI and SDT work among them. Japan’s RIKEN already runs the pattern, linking Blackwell GPUs to Fugaku and to IBM and Quantinuum processors [S95], and Korea’s science ministry has announced the KISTI–Hangang link [S93]. This differs from Report Two’s R-8, which concerns cryogenic chokepoints.

Owner: State with DOE. 90-day artifact: an allied technical working group convened with NVQLink adopters and a draft interface and security baseline. Authority: State Department foreign-affairs authority and DOE international science agreements, consistent with EO 14413’s direction to align quantum policy with allies [S49]; Report Two. Success measure: at least four allied adopters in the working group and a published baseline within twelve months.

18. Honest Concessions

• Most Chinese technical claims here originate with the companies or their investors. No Chinese quantum-AI result in Part One has a third-party replication in a non-affiliated venue as of September 24, 2026; RiverONE’s method is published, but only as a preprint.
• Nobody, IonQ included, has demonstrated quantum advantage for AI at production scale. IonQ’s cited results are peer-reviewed research or company preprints; the radar result is one preprint whose largest gains are simulated; its hardware figure (0.32 filtered F1) is a feasibility result.
• Quantum-inspired methods may owe little to quantum physics. If so, Track One is a model-efficiency story, and its competition is classical compression.
• State capital and state-led consortia are pervasive in China. Their presence does not by itself prove military purpose.
• LineShine’s CPU-only design is a strength for independence and a limitation for AI training. It should not be read as parity in AI compute.
• Beijing rejects the reading of “hide our capacities and bide our time” as deception, and documented understatement of budgets and costs does not by itself prove a campaign to understate quantum readiness. The disclosure-gap thesis is offered as a planning assumption, not a finding of hidden capability. [S85], [S83]
• China has no publicly identified operational quantum-key satellite in orbit; the next are in procurement, and whether any undeclared payload flies is a collection requirement, not a finding. [S132], [S200], [S201], [S139]
• The PLA has not identified its “ten-plus” experimental systems, and the Chinese text does not say all are quantum; the table in Section 8 maps quantum possibilities, and claims of stealth-detecting quantum radar in particular are contested by US defense scientists. [S176], [S182]
• The State Department finding names imagery companies, not the Chinese state; the separate Jordan attribution rests on unnamed US officials and is a collection item, and Beijing disputes it. The HQ-29’s anti-satellite role is not confirmed, and the microwave-driver paper claims no satellite use. [S217], [S161], [S162], [S173], [S172]
• AI can mislead intelligence work: a reported US near-boarding of a Chinese ship in spring 2026 rested on an AI misidentification. Faster ISR analysis, including quantum-AI, needs verification; NATO has made the same demand for traceability and confidence thresholds. [S55], [S56], [S97], [S98]
• The authors hold long positions in IonQ, and the actions name IonQ.

1. What Would Change This View

C1. Independent tests show quantum-inspired models give no efficiency advantage over leading classical compression on domestic Chinese GPUs.

C2. The Chinese quantum-AI results in Part One fail independent replication, or the companies step back from AI claims in IPO prospectuses.

C3. Disclosures show CETC’s consortium role and fund position are administrative or purely financial, with no information or program links.

C4. The NVAQC installation or first Superion 256 deployments slip materially, or the radar and fine-tuning results fail to hold on larger hardware. The case for Q-1, and for IonQ’s role, would weaken.

20. Conclusion — Closing the Series

 Three reports, one argument: the lead will be kept in integration, or it will not be kept.

What the series has shown. Report One showed that the infrastructure beneath the AI transition can be contracted now, and that China runs a conversion system that retains its specialists, converts its firms and finances its champions through defense capital. Report Two showed what that system is building — a national quantum stack, carried into orbit — and that IonQ has assembled the American equivalent inside one accountable company. This report has shown where the system is pointed next: the fusion of AI and quantum on three tracks, governed in Beijing by a defense-electronics institute, running on a domestic supercomputer that tops the world list.

What the summit did not change. [FACT] Both leaders raised AI on September 24; Xi asked for human control, the President said he wanted to leave AI where it is, and the visit ended on September 25 with a new AI dialogue and incident channel but no rules on AI development. The trade truce runs to January 10. [INFER] Nothing agreed at the summit constrains either side’s AI or quantum program. The terms of the competition will be set in laboratories, foundries and machine rooms, not in the East Room.

What the United States should do. [ARG] It does not need to match China company for company. It needs to integrate first: American quantum processors beside American AI supercomputers, on an open standard allies already share, measured against benchmarks America sets. The fourteen actions of Reports One and Two build and secure the stack. The eight in this report connect it to AI and update the baseline it rests on. Focus decides the outcome: backing paths that can be manufactured and upgraded at scale is the difference between keeping pace and falling behind, and Part Two sets out one such path. Basis: [S1], [S2], [S30], [S41]; Sections 10 and 17.

[ARG] In September 2026 IonQ put a QPU on the calendar beside NVIDIA’s leading AI rack, sold the same system into Korea and Florida, published peer-reviewed quantum-AI results with national laboratories, and applied quantum generative models to defense-grade radar. It should have to compete for every federal role this paper proposes, and it should have to win. The next markers are the NVAQC installation and the first Superion deliveries in 2027 on the American side, and the TuringQ and Origin Quantum prospectuses on the Chinese side. Basis: [S30], [S32], [S35], [S37], [S38].

What Beijing has not shown. [ARG] The premier’s record shows the state putting its own money behind quantum and saying so. What it does not show is the full extent: a quantum satellite prepared for launch and not announced, anti-satellite equipment its diplomats say China opposes, and imagery from Chinese companies reaching a war against Americans. This report has counted what Beijing chose to reveal. Beijing fields drone swarms, develops military humanoid robots and flies sixth-generation fighter prototypes it has not presented, builds its quantum constellation with little public detail, spends well beyond its announced budget, and has been linked to the largest covert influence operation Meta has documented. What is visible is not the whole; the prudent course is to act on what can already be seen and to treat the rest as a collection requirement. Basis: [S63], [S64], [S70], [S71], [S76], [S84]; Sections 1, 8 and 9; Report Two.

 Whoever integrates AI and quantum first will hold a moat that is commercial and military at once. The window is open now. It will not stay open by default.

Appendix A. Confirmed US Designations

Designations relevant to the quantum-AI lane, confirmed in this report’s sources or in Reports One and Two.

Entity	Role in this report	Designation	Source
China Academy of Electronics and Information Technology (CAEIT; CETC CAEIT)	Lead of Beijing’s Quantum-AI Consortium	Entity List, July 12, 2021 (Xinjiang-related group); not named on the June 2026 Section 1260H list (CETC parent is)	Federal Register 2021-14656; DoD 1260H list
CAS Institute of Computing Technology	Consortium member	Entity List, December 2022	Federal Register 2022- 27151
Beijing Academy of Artificial Intelligence	Consortium member	Entity List, March 25, 2025; grounds: developed large AI models and advanced computing chips for defense purposes	Federal Register 2025- 05427
Beijing Academy of Quantum Information Sciences	Consortium member	Entity List, May 2024	Federal Register, May 14, 2024
Biren entities (Beijing, Shanghai, Guangzhou, Hangzhou, Zhuhai)	Qingxing Yigou adaptation partner	Entity List, effective October 17, 2023	Federal Register, October 19, 2023
QuantumCTek Co., Ltd.; Shanghai QuantumCTek	Controlled by China Telecom Quantum Group (Tianyan)	Entity List, November 2021	Commerce Department; Report Two
Origin Quantum	Track Two (AI fine-tuning)	Entity List, May 2024	Report Two; The Quantum Insider
University of Science and Technology of China	Its president set out the quantum and space computing case at WAIC 2026	Entity List, May 2024	Report One; Federal Register, May 14, 2024
CAS Center for Excellence in Quantum Information and Quantum Physics	Academy quantum institute	Entity List, May 2024	Report One; The Quantum Insider
Hefei National Laboratory for Physical Sciences at Microscale	USTC quantum research base	Entity List, November 2021	Commerce Department; eeNews Europe

Appendix B. Primary-Source Log

Tier is a guide to source type: US/allied gov; CN official (Chinese government and state media); Institution/company (primary corporate, laboratory and institutional pages); Research (peer-reviewed papers and preprints); Secondary (press, analysis and relays). For the most consequential claims, the report cites the highest tier available. Web links for every source are in the companion Link Register (342 links across 247 sources), which is kept current; this log gives each source’s title, tier and date so any [S#] can be resolved from the printed page.

The leaked documents

Described in the authors’ words from published reporting. The originals are not reproduced here; slide images appear in The Insider’s investigation (theins.press/en/inv/294635).

Document	What it contains	Source
Guangzhou deck on Starlink (November 2023)	[FACT] Presented by Huang Hui and Ren Jie of China Aerospace Science and Technology Corporation and marked “internal” in Chinese. It surveys Starlink’s civilian and military uses, scale and performance, then its role in Russia’s war in Ukraine, and sets out a three-tier escalation: legal and diplomatic pressure and occupying the frequencies and orbital positions Starlink needs; jamming and cyberattacks through civilian terminals; and low-cost weapons to destroy satellites in orbit.	[S114], [S124], [S125]
The other three Guangzhou decks (November 2023)	[FACT] Together they propose joint development of integrated air and missile defense; autonomous loitering-munition swarms combining Russia’s frontline experience with Chinese AI and manufacturing; next-generation armored vehicles with AI-driven active protection, uncrewed turrets and integrated drone swarms; and technology exchanges tied to material supplies.	[S114], [S124], [S125]
Moscow working protocol (June 5, 2023)	[FACT] Ten pages, bilingual and signed. A Chinese delegation led by Colonel Tong Xiaofeng of the technical-cooperation bureau in the Central Military Commission’s Equipment Development Department met Rosoboronexport, Almaz-Antey and NPO Almaz. Reporting on the protocol says it commits the two sides to jointly build a hypersonic-missile defense system using technology Russia had not previously shared.	[S114], [S126], [S127]
Fourth forum, Yekaterinburg (December 2024)	[FACT] Chinese military researchers proposed that data on Russian drone strikes in Ukraine be shared with Beijing.	[S114]
Provenance	[FACT] An anonymous source passed the documents to Der Spiegel more than a year before publication. The Insider says errors in the Russian text show the decks were drafted by the Chinese side; travel records reportedly corroborate named delegates; and a forum notice prohibited any disclosure or interviews.	[S114], [S128]

[INFER] The documents describe a division of labor: Russia brings combat experience from Ukraine, China brings technology and manufacturing scale, and the two plan systems neither could build alone. They sit outside anything Beijing has publicly acknowledged. Basis: [S114], [S125].

 

[FACT] Leaked Russian defense-industry documents published by StateWatch show that the Nudol anti-satellite missile’s construction uses materials of foreign origin, including Chinese-made Centralite II. Basis: [S175].

Disclosure

Friends of IonQ is a group of researchers and analysts focused on IonQ activity, the quantum ecosystem at-large, and how quantum technology can enhance the human condition while fortifying the national security of the United States and its allies. Its members hold a long position in IonQ. No individual nor Friends of IonQ receives any compensation or future considerations from IonQ. When the group sees challenges the company needs to address or issues of concern, it raises them in its reports.

This report is not investment advice and is not a solicitation. Forward-looking statements by companies are cited as made and are not endorsed. Chinese-language material has been translated and summarized by the authors; readers should consult the originals listed in Appendix B.

ALL SOURCES AVAILABLE UPON REQUEST.