IMS says its neutral-atom system will begin at approximately 50 qubits and scale to 500. Its March 2031 goal: 10,000 physical qubits with error detection and correction, no logical-qubit count.
On August 24, the Institute for Molecular Science announced that Shunkai, which it calls "Japan's first full-stack neutral-atom quantum computer," is operational in Okazaki. The IMS-led machine will use approximately 50 qubits in its early stage, the institute says, and will expand to approximately 500. At that scale, it opens partially to external users to develop applications and demonstrate error correction.
The longer timeline sits in the release's closing section. By March 2031, the end of the Ohmori Moonshot Project's second stage, the goal is "a large-scale, high-performance neutral-atom fault-tolerant quantum computer, with 10,000 physical qubits and quantum error detection and correction capabilities, available to external users."
That goal joins a map of dated targets SCN has tracked this year, alongside the US Department of Energy's 2028 initiative, IBM's 2029 Starling system and Microsoft's pulled-forward 2029 timeline — and it is the only one stating a physical-qubit count and no logical one.
The release defines its central term: a full-stack system integrates every layer that turns user input into drive signals for the device and returns a result. "Personal computers and supercomputers are examples of full-stack systems," it notes. In Shunkai, individual atoms are the qubits, held in an array of optical tweezers (tightly focused laser light); microwaves or lasers drive gates, and a camera reads each atom's fluorescence.
The release states the roles precisely: IMS "has taken the lead in developing this full-stack quantum computer," with Hitachi on the software stack and Infleqtion on the quantum processing unit stack. Infleqtion's own release claims no more: it "contributed its quantum processing unit to the program." Superstaq, the software platform in Infleqtion's corporate boilerplate, is not named as a Shunkai component by either institution. Commercialization runs through a planned collaboration with Yaqumo Inc., where Professor Kenji Ohmori, the project's lead, is founder and executive advisor.
The name is an astronomer's. Harumi Shibukawa, whose given name can also be read "Shunkai," built Japan's first original calendar in the Edo period; a note in the release runs the line from the celestial sphere to the Bloch sphere.
Shunkai is the hardware face of a program that has been running since fiscal 2022. Under the Cabinet Office/JST Moonshot Program's Goal 6, "Realization of a fault-tolerant universal quantum computer," Ohmori led Stage 1 (FY2022–FY2025) and was selected in October 2025 to lead Stage 2 (FY2026–FY2030, matching the release's April 2026 start and March 2031 end), retitled "Neutral atom-based fault-tolerant quantum computer."
The record above the release is consistent, and consistently number-free. The Cabinet Office's formal 2030 milestone is a quantum computer with "small-scale or partial fault tolerance," as translated from the Japanese; no qubit count. JST's Stage 2 project page and its 2026 project-summary PDF (team chart dated April 2026) set two capability milestones: quantum error correction demonstrated to be effective by 2028; a universal gate set with logical qubits, run by users outside the project, by 2030. Neither states a qubit-count target. The 10,000 figure appears in the August launch releases: IMS gives "with 10,000 physical qubits," Infleqtion "up to 10,000."
The figure surfaces elsewhere in the Goal 6 record: a progress report from the program's Koashi theory project treats the "10,000-qubit scale" as where near-future machines are expected to land, a theorist's resource estimate on a page that never mentions the Ohmori project. No program document states 10,000 physical qubits as a hardware target for the Ohmori project.
Stage 2 also redraws the roster: JST's R&D institutions are IMS, RIKEN, Kyoto University, Hitachi, the National Institute for Fusion Science, AIST and Yaqumo. Infleqtion, a Stage 1 collaborator credited in the launch release with the QPU stack as built, is not among them; neither document addresses the difference.
The same layering shows up in the US record. The most-quoted number in the DOE's 2028 push, fault-tolerant systems with logical qubits "numbering in the low hundreds," comes from DOE's June 2026 Quantum Genesis announcement; Executive Order 14413 above it establishes the QC-ADDS effort with no qubit numbers and no 2028 date. There, the quotable figure lives in the announcement; the authority documents stay qualitative.
Hence the labels on the map: a dated public target, the authority behind it, and the scale exactly as stated, which in one case means not at all.
Target holder | Authority | Stated target | Date |
|---|---|---|---|
US DOE (Quantum Genesis) | government initiative | fault-tolerant quantum systems with logical qubits "numbering in the low hundreds" (DOE announcement, June 2026) | 2028 |
IBM | vendor roadmap | Starling: 200 logical qubits, roughly 100 million gates | 2029 |
Microsoft | vendor roadmap | "a scalable, practical quantum computer"; no qubit count stated | 2029 |
IonQ | vendor roadmap | 2 million physical qubits, 80,000 logical qubits (roadmap, retrieved August 31) | 2030 |
Ohmori project (Japan) | government-funded project goal | 10,000 physical qubits with "quantum error detection and correction capabilities," open to external users (IMS release) | March 2031 |
There is no reliable way to convert the 10,000-physical-qubit goal into logical qubits without knowing the code, physical error rates, and target logical error rate. Vendor ladders show the spread: IonQ's roadmap pairs 10,000 physical qubits, the scale Japan's project targets for 2031, with 800 logical qubits as early as 2027, a projection built on its trapped-ion hardware and error-correction assumptions. SCN's survey of error correction beyond the surface code examines how the choice of code moves that ratio. The project's own 2030 milestone makes logical qubits the destination; it does not say how many.
Gate fidelities are not reported in the launch materials, and the only number above 99% in the project's official record is Stage 1's "qubit detection fidelity exceeding 99%," a readout result from an elemental experiment that says nothing about gate performance on Shunkai. XenoSpectrum's enumeration, checked against both language versions of the release, adds circuit depth, continuous-operation rate, computational benchmarks, the exact number of qubits running at launch, and the timeline and terms of external access. The atomic species also goes unnamed.
The platform's reference points sit outside the project. QuEra's Gemini product page specifies 99.9% single-qubit and 99.2% two-qubit gate fidelity on 260 rubidium-87 qubits (a vendor spec, retrieved August 30). In published research, a Caltech team reported entangling gates at 99.62% fidelity, averaged over symmetric input states, on optical-clock qubits in a tweezer array (Nature, 2024), and in June 2026 a JILA team reported an error-detected two-qubit gate fidelity of 99.78% on metastable ytterbium-171 qubits, a figure that excludes error events flagged by erasure detection and so is not directly comparable to raw gate fidelities.
Whether error correction pays off at a given fidelity is not a one-number question: thresholds depend on the code, the full error model including leakage, measurement performance, qubit connectivity, and the decoder. In general, overhead falls as fidelity climbs past a code's threshold, at a code-dependent rate. Where Shunkai sits, the launch does not say. IMS says the 500-qubit phase will support application development and "the demonstration and improvement of quantum error correction."
The "full-stack" qualifier does real work: a US-built neutral-atom machine already runs in Japan. QuEra announced in late May 2025 that it had installed a quantum computer at AIST's G-QuAT center in Tsukuba, its first deployment outside its own labs. On March 24, 2026, AIST opened ABCI-Q's quantum computers, including the 260-qubit QuEra system, to external users: domestic organizations first, international slated for FY2026. Shunkai's distinction, as IMS frames it, is where the stack was built: under the institute's lead, inside a national program. Its October announcement described the Stage 1 result as Japan's first, and one of the world's few, full-stack neutral-atom quantum computer systems.
"It is also expected that Shunkai will be integrated with the existing shared supercomputer facility at the IMS to develop into a quantum-GPU hybrid computing center," Ohmori said. The AIST pairing points in the same direction, a direction SCN has tracked as hybrid quantum-classical work moves into working chemistry pipelines. Asked about Shunkai at her August 25 press conference, Minister of State Kimi Onoda, whose portfolio includes science and technology policy, said that achieving it "as a system spanning software through the computing device" is "a major step forward in Japan's quantum computer development" (translated from the official summary).
Three markers will show how much weight the 2031 goal can bear. The first is any fidelity or benchmark disclosure as the machine grows toward 500 qubits, the phase IMS has set aside for demonstrating error correction. The second is paperwork: whether 10,000 migrates from launch material into JST's Stage 2 record, which today carries capability milestones and no qubit target. The third sits on the US side of the map: EO 14413 gives the Energy Secretary until September 20, 90 days from the order, to identify QC-ADDS technical specifications and "publicly release a summary of those specifications, as appropriate." That is the next dated marker on the DOE's 2028 fault-tolerance push.
The SCN Staff is a small AI editorial squad working under human direction. Each agent owns one job.
Scout does the research. It runs down primary sources and checks what's already been published, on SCN and everywhere else, before a story gets written. If a claim can't be traced back to a real document, Scout flags it.
Forge writes. It takes what Scout found and turns it into a draft, argument and sentences and all. Every SCN piece starts here, then gets sharpened.
Cipher handles search: the titles, descriptions, and keyphrase work that decides whether a good article ever gets found. Least glamorous job on the squad. Also one that matters more than it looks.
Pixel makes the visuals. Images, charts, the occasional diagram, all built to SCN's brand instead of pulled from a stock library. When something's easier to see than to read, it goes to Pixel.
Editorial judgment and the final call stay with the humans. So does the fact-checking.
