China’s Neutral-Atom Sprint: Three Generations in Nine Months — But Where Are the Benchmarks?

July 18, 2026 — Chinese startup Zhongqi Wuliang (众擎无量) unveiled the Qinghe No. 1 (清河一号) at the World Artificial Intelligence Conference (WAIC 2026) in Shanghai. The company, a spinoff from the Chinese Academy of Sciences’ Shanghai Institute of Optics and Fine Mechanics (CAS SIOM) and founded by Lu Xudong, described the machine as a neutral-atom quantum computer engineered to fit inside a single standard data center server rack.

The Qinghe No. 1 is the third neutral-atom quantum computer to emerge from China’s CAS-linked ecosystem in nine months. In October 2025, a separate CAS-affiliated company, Zhongke Kuyuan, delivered the Hanyuan-1 to a China Mobile subsidiary: a 100-qubit room-temperature system spanning three standard equipment racks, with reported single-qubit gate fidelity of 0.999 and two-qubit gate fidelity of 0.98. The Hanyuan-1 also secured China’s first neutral-atom export order, a sale to Pakistan for its national quantum computing center, bringing total orders to over 40 million yuan (approximately 5.6 million USD), according to the CSIS. In May 2026, Zhongke Kuyuan unveiled the Hanyuan-2: a dual-core system using 200 qubits split across two rubidium isotopes (Rb-85 and Rb-87), housed in a single cabinet with total power consumption below 7 kW.

Three systems from China’s CAS-linked neutral-atom ecosystem in nine months. October 2025 to July 2026. Each in a smaller package than the last: three racks, one cabinet, one rack. All operating at room temperature with laser cooling instead of dilution refrigerators. Two came from the Wuhan-based Zhongke Kuyuan team (backed by Hubei province); the third from Shanghai’s Zhongqi Wuliang (backed by the city’s dual quantum hub initiative). Different cities, different companies, same CAS research lineage, same modality.

Several outlets covered the Qinghe No. 1 as the “first data-center-deployable quantum computer.” That claim does not hold up. Austrian company Alpine Quantum Technologies (AQT) demonstrated a trapped-ion quantum computer in two standard 19-inch server racks in 2021, operating at room temperature and consuming under 2 kW, and published the design in PRX Quantum with full technical specifications. AQT has since commercialized the design; its IBEX Q1 system became available on Amazon Braket in November 2025. IonQ’s Forte Enterprise, a trapped-ion system in rack-mounted form factor, has been commercially available since April 2025.

The Qinghe No. 1 may be the first neutral-atom system in a single server rack, though even that distinction is blurred by the Hanyuan-2’s “single cabinet” form factor two months earlier. Zhongqi Wuliang has not published the system’s dimensions, weight, or power consumption.

On performance, Zhongqi Wuliang told WAIC attendees that the Qinghe No. 1 achieves “internationally competitive” metrics in qubit count, readout fidelity, and gate fidelity, according to Baidu’s encyclopedia entry. The company has not published specific numbers for any of these metrics in any venue, peer-reviewed or otherwise.

Zhongqi Wuliang’s supply chain partner, Xuanxiang Technology, has separately commercialized what it describes as the first million-level metasurface optical tweezers array chip, designed to address the optical alignment bottlenecks that have historically constrained neutral-atom scaling. The chip completed system adaptation and verification with Zhongqi Wuliang in June 2026.

My Analysis

Nine months. Three neutral-atom systems. Each one in a smaller package. Two different CAS-linked teams in two different cities, backed by two different provincial governments, converging on the same modality. I am not aware of any Western neutral-atom vendor that has publicly demonstrated a comparable form-factor compression within a comparable timeframe. And I am unable to tell you whether any of these machines are any good, because none of them have published the data that would let me assess that.

Those two facts sit in tension.

Nine Months, Two Ecosystems

What makes this sprint striking is that it emerged from two independent CAS-linked regional clusters, not one company’s roadmap. Zhongke Kuyuan operates out of Wuhan, backed by the CAS Innovation Academy for Precision Measurement Science and Technology and Hubei province’s “Pioneer” program. Zhongqi Wuliang operates out of Shanghai, backed by CAS SIOM and the city’s dual quantum hub initiative. When two separate provincial ecosystems converge on the same modality and produce progressively more compact hardware within months of each other, it tells you something about national-level coordination.

I traced the roots of this alignment in my 10-part series on China’s quantum ambitions. The 15th Five-Year Plan (2026–2030) placed quantum technology first among seven “future industries.” The National Venture Guidance Fund allocated approximately 17.5 billion USD across three regional funds covering strategic technologies including quantum computing. Shanghai designated neutral-atom computing as a core route in its quantum roadmap, established two purpose-built hubs (the Xuhui Cultivation Zone and Zhangjiang Quantum Bay), and committed up to 100 million yuan per project for foundational research through the Xuhui zone.

The pattern I described when analyzing Hefei’s Quantum Avenue is now replicating in Shanghai. Hefei took USTC research and produced Origin Quantum, QuantumCTek, and CIQTEK through a combination of state capital, university pipelines, and municipal industrial policy. Shanghai is running the same playbook on neutral atoms: CAS SIOM provides the research base, Xuanxiang Technology provides the optical tweezers supply chain, startups like Zhongqi Wuliang and Taiyi Quantum attack different hardware approaches, and the city government provides the coordination and capital. Meanwhile, Wuhan’s Zhongke Kuyuan has been executing its own neutral-atom program through Hubei’s parallel funding channels.

Taiyi Quantum illustrates the pace. Founded in January 2026, the company raised over 400 million yuan (59 million USD) in six months, built a 1,000-square-meter cleanroom in Xuhui, and assembled a team of roughly 50 researchers, including hires with experience at institutions such as MIT, JILA, NIST, and the Centre for Quantum Technologies in Singapore, according to company and municipal announcements. Its roadmap calls for demonstrating logical qubits on ytterbium atoms by the end of 2026. I have not been able to verify that timeline against published experimental results.

The form-factor compression across these systems is worth examining, even though it spans two separate companies rather than one iterating roadmap. Going from three racks (Hanyuan-1) to a single cabinet (Hanyuan-2) to a single rack (Qinghe No. 1) in nine months required sustained progress in miniaturizing the laser systems, vacuum chambers, and control electronics that make up a neutral-atom quantum computer. Neutral atoms avoid the large dilution refrigerators that dominate the footprint of superconducting systems, but they still require substantial optical infrastructure: trap lasers, Rydberg excitation lasers, spatial light modulators or acousto-optic deflectors, vacuum pumps, and imaging cameras. Fitting all of that into a single server rack is a legitimate engineering achievement regardless of which team accomplished it. Whether the performance survived the compression is the question nobody has answered.

Speed Without Data

Here is what has been published about the performance of these three systems:

Zhongke Kuyuan reported single-qubit gate fidelity of 0.999 and two-qubit gate fidelity of 0.98 for the Hanyuan-1. These numbers appeared in Chinese state media (Hubei Daily) and were subsequently reported by the Quantum Computing Report. They have not been published in a peer-reviewed journal. The qubit count (100) and the room-temperature operating claim are consistent with what the international neutral-atom field has demonstrated, so these figures are plausible. But plausible and verified are different standards.

For the Hanyuan-2, no gate fidelity figures were disclosed. Its 200-qubit dual-core architecture (two arrays of different rubidium isotopes operating in parallel or in a main-core/auxiliary-core configuration) is a novel design claim, but no performance data accompanied the announcement.

Zhongqi Wuliang disclosed nothing for Qinghe No. 1. Qubit count: unpublished. Two-qubit gate fidelity: unpublished. Coherence time: unpublished. Readout fidelity: unpublished. Chinese-language coverage describes the system’s performance as reaching “world-class level” (国际一流水准), a stronger assertion than the English-language “internationally competitive” that circulated through trade press. Either way, it is a marketing claim. As I noted in my Shanghai hubs coverage before the WAIC debut: “I have not been able to independently verify either claim from primary technical publications.”

Compare this with the Western neutral-atom ecosystem. QuEra’s logical qubit demonstrations are published in Nature with full experimental detail. AQT’s rack-mounted trapped-ion system was published in PRX Quantum in 2021 with qubit count, gate fidelities, and system specifications. Pasqal’s deployments at GENCI, Jülich, CINECA, and OVHcloud include published performance data accessible through cloud platforms and partner documentation. The field’s convention is that hardware claims accompanied by published benchmarks carry weight. Claims without benchmarks are announcements.

This is not a Western-exceptionalism argument. Several Chinese quantum teams publish world-class results through conventional academic channels. Lu Chaoyang’s photonic boson sampling demonstrations, Pan Jianwei’s quantum satellite experiments, and the Zuchongzhi superconducting processors from USTC have all appeared in Science, Nature, and Physical Review Letters with full reproducibility data. The neutral-atom commercial ecosystem in China has not followed that pattern. Whether this reflects early-stage commercial secrecy (common in Western startups too), a deliberate strategy to avoid revealing capabilities, or a gap between the marketing claims and the underlying performance is something I cannot determine from the outside.

Smaller Boxes, Unknown Contents

The emphasis on form factor in the coverage of Qinghe No. 1 deserves scrutiny. A quantum computer that fits in a server rack is easier to deploy than one that requires a dedicated laboratory. That is a genuine advantage for commercial adoption. But form factor and computational capability are independent variables. A single-rack neutral-atom system with 20 noisy qubits would be less useful than a room-sized system with 1,000 high-fidelity qubits running quantum error correction.

For assessing progress toward a cryptographically relevant quantum computer (CRQC), what matters are the capabilities I mapped in my CRQC Quantum Capability Framework: physical error rates below the fault-tolerance threshold (B.3), demonstrated logical qubit operations (C.1, C.2), real-time decoder performance (D.2), and the ability to sustain computation over extended periods (D.3). Shrinking the physical enclosure addresses none of these. It addresses E.1: Engineering Scale and Manufacturability, which is important for eventual deployment, but late in the capability chain.

I want to be precise about this because the “data-center-ready” narrative can mislead enterprise buyers and policymakers who are not tracking the technical details. A quantum computer that plugs into an existing server rack is attractive to a procurement officer. Whether that machine can run a useful circuit, hold a logical qubit, or sustain an error-corrected computation for longer than a few seconds requires a completely different set of evidence. Zhongqi Wuliang has provided the first kind of claim and none of the second.

Why Neutral Atoms Matter Despite the Benchmarks Gap

None of this skepticism about the current announcements changes my assessment of the neutral-atom modality itself. In my CRQC Scorecard, I mapped every major quantum computing modality against the three executive metrics (Logical Qubit Capacity, Logical Operations Budget, and Quantum Operations Throughput) that define the path to a CRQC. Neutral atoms held the strongest overall position: the smallest LQC gap of any modality (approximately 15x), demonstrated magic state distillation, massive parallelism through reconfigurable tweezer arrays, and uniquely, demonstrated continuous operation exceeding two hours. No other modality had demonstrated anything comparable on that last point.

Shanghai’s quantum planners appear to have reached a similar conclusion. Instead of extending Hefei’s superconducting and photonic research base, Shanghai carved out a neutral-atom niche backed by dedicated innovation zones and municipal subsidies. Google’s decision to launch a parallel neutral-atom research program in Boulder, Colorado in March 2026, alongside its decade-old superconducting effort, provides independent confirmation of the modality’s trajectory.

The question is whether China’s neutral-atom commercial ecosystem is building genuine capability behind the form-factor headlines, or whether the hardware cadence reflects primarily an industrial policy push for visible milestones. The two possibilities are not mutually exclusive. A government that funds three hardware generations in nine months is also funding the engineering teams, laser physicists, and vacuum specialists who will eventually produce competitive performance numbers. The talent pipeline from CAS, which I analyzed in my China series, is real. The publication gap may close as these commercial systems mature and the companies seek international customers who demand verified specifications.

The Comparison Worth Making

The form factor and performance axes tell different stories depending on which company you examine.

AQT demonstrated a rack-mounted (two standard 19-inch racks) trapped-ion quantum computer in 2021, at room temperature, under 2 kW, with peer-reviewed benchmarks and subsequent deployment at European computing centers. Its commercial IBEX Q1 offers 12 fully connected qubits with two-qubit gate fidelity reported between 97.7% (Amazon Braket at launch) and 98.85% (AQT partner materials). Small, verified, modest in qubit count.

Pasqal’s Orion series operates at room temperature with no cryogenic cooling and a peak power consumption of 10 kW per its technical specifications. The Orion is not a rack-mounted system; it spans five modules in a dedicated hosting room, with a total system weight of approximately 2,500 kg. But Pasqal has deployed systems ranging from 100+ qubits at GENCI (France) and Jülich (Germany) to 140 qubits at CINECA (Italy), with published performance data accessible through cloud platforms. Its March 2026 QRMI integration with NVIDIA CUDA-Q made the neutral-atom QPU Slurm-schedulable, a genuine HPC integration milestone.

The Chinese neutral-atom systems occupy a different quadrant: rapid iteration on form factor, with the Hanyuan-1 providing the only published (if not peer-reviewed) fidelity numbers in the group. The comparison is not flattering to the Chinese systems on the benchmarks axis. It is striking on the deployment speed axis.

For CISOs and organizations tracking the CRQC timeline, neither axis can be ignored. The benchmarks tell you how far the physics has come. The deployment speed tells you how fast the engineering ecosystem is moving. I have consistently argued that the engineering capacity to build, test, and iterate quantum hardware rapidly is an underappreciated factor in predicting Q-Day. Most CRQC timeline models treat hardware capability as a smooth curve. The Hanyuan-1 to Qinghe No. 1 arc suggests the curve in China could have a steeper slope than those models assume, once the benchmarks catch up to the form factor. Whether the benchmarks will catch up is the open question.