The Institute for Molecular Science (IMS), part of the National Institutes of Natural Sciences, announced on August 24, 2026, that Japan's first full-stack neutral-atom quantum computer, "Shunkai," has begun operating. In its initial stage, it uses roughly 50 qubits. By numbers alone, this isn't the largest system in Japan. Still, this marks a milestone because, for the first time, everything from the software where users input calculations to the quantum processing unit (QPU) that arranges atoms, performs operations, and reads out results has been run as a single integrated system.
Shunkai was developed by a team led by Professor Kenji Ohmori of IMS under the Cabinet Office and Japan Science and Technology Agency (JST) Moonshot R&D Program. Its name comes from Shibukawa Shunkai, an Edo-period astronomer who created Japan's own calendar system. Building on the QPU, laser, and control technologies accumulated in Phase 1, the team has now assembled a testbed for advancing error correction and external access in Phase 2, which began in April 2026.
Why "full-stack" matters more than 50 qubits
Shunkai consists of four layers: software, classical control, quantum control, and the QPU. When a user submits a job from a quantum application, calibration software and a job execution system process the instructions, and the classical control system converts them into signals that drive the QPU. IMS led the overall integration, with Hitachi, Ltd. co-developing the software stack and US-based Infleqtion co-developing the QPU stack together with IMS.
In the QPU, "optical tweezers"—lasers focused through objective lenses—capture individual atoms. The quantum control unit cools the atoms, arranges them into an array, and rearranges them as needed. After manipulating quantum gates with microwaves and laser light, a camera captures the fluorescence of each atom individually to read out the computation results. The term "full-stack" refers to this unbroken chain from input to readout.
Describing this simply as "a US-made QPU placed in Japan" falls short. While Infleqtion's QPU sits at the core, IMS and Hitachi jointly developed the control, operation, and software layers around it. In Phase 2, Hitachi is also developing system performance evaluation software and will participate in discussions on the design and operational policy of the actual hardware. A mechanism for continuously measuring and improving Shunkai has been built into Japan's research plan.
The strength of the neutral-atom approach: moving and connecting atoms
In the neutral-atom approach, a single electrically neutral atom serves as one qubit. Atoms of the same species share identical properties, resulting in less device-to-device variation than artificially manufactured qubits. By moving the optical tweezers, the atoms' arrangement can be changed even mid-calculation. A key feature is the ability to bring separated qubits together to interact, reconfiguring their connectivity to match the algorithm.
Quantum gates work by exciting atoms into high-energy Rydberg states, exploiting strong interactions between nearby atoms. Because this reduces the constraints of fixed wiring, qubit connectivity can be more easily designed to suit error correction. However, being able to freely move atoms is not the same as being able to sustain high-precision operations over long periods. Real hardware must keep the lasers stable and maintain multiple performance metrics simultaneously, from gate operations through readout.
"Room-temperature operation" also requires a caveat. While massive dilution refrigerators like those used in superconducting systems aren't needed, the atoms must still be laser-cooled and confined inside a vacuum chamber. Laser light sources, optical systems, and control electronics remain essential. What operates at room temperature is the overall environment of the apparatus—the cooling process itself doesn't disappear.
The results from Phase 1 show how much foundational technology had been built up before integration. The team constructed a high-power 325 nm light source used to excite ytterbium atoms and recorded a qubit detection fidelity above 99%. The time during which nuclear spin preserves quantum information was more than a million times longer than the operation time of a single-qubit gate. With rubidium atoms, the team automated the shaping and uniformization of 800 optical tweezers, and has advanced non-destructive measurement—reading a qubit's state without destroying it—to the proof-of-principle stage.
However, that above-99% figure refers to "detection fidelity" from an elemental experiment—it does not mean Shunkai's quantum gates operate with over 99% accuracy. This announcement does not disclose single-qubit or two-qubit gate fidelity, achievable circuit depth, continuous operation rate, or computational benchmarks. The start of operations marks the achievement of integration; demonstrations of fault tolerance or quantum advantage still lie ahead.
50, 500, and 10,000: different timelines
The three scales mentioned in the announcement represent the present, the next upgrade, and a long-term goal. Reading them together risks the misunderstanding that Shunkai is already operating with 10,000 qubits.
| Stage | Qubit scale | Status/Goal |
|---|---|---|
| At start of operation | ~50 qubits | Scale used in the initial stage |
| System expansion | ~500 qubits | Planned expansion within the same system |
| By FY2030 | 10,000 physical qubits | Goal of achieving error detection/correction functionality available to external users |
Increasing the number of qubits tenfold or two-hundredfold doesn't necessarily mean computational power scales at the same rate. What matters is how many qubits can be initialized simultaneously, at what precision gates can be applied, and whether they can be used as logical qubits while detecting and correcting errors along the way. IMS has not disclosed in this announcement the exact number of qubits or the atomic species Shunkai actually controls simultaneously in its initial stage.
The interim milestones set by JST focus on function rather than scale. By 2028, the goal is to demonstrate the effectiveness of quantum error correction, with external users within the project running quantum circuits on the actual hardware. By 2030, the plan calls for realizing a universal set of gates on logical qubits, with users outside the project executing logical quantum circuits. The figure of 10,000 physical qubits will only carry research significance once accompanied by that functionality.
Not Japan's largest: how it differs from AIST's 260-qubit machine
Japan already has a neutral-atom machine with more qubits than Shunkai. The National Institute of Advanced Industrial Science and Technology (AIST) operates "System Q," made by US-based QuEra, as part of its quantum-classical hybrid platform "ABCI-Q." This 260-physical-qubit machine uses rubidium-87, is connected via a dedicated line to the GPU supercomputer "System H," and shares its platform with superconducting quantum computers and other systems.
| System | Initial/rated scale | Main goal |
|---|---|---|
| IMS's "Shunkai" | ~50 qubits | Integrating a full stack domestically to research error correction and hardware advancement |
| AIST ABCI-Q's "System Q" | 260 physical qubits | Combining multiple approaches with GPUs to develop applications and use cases |
Accordingly, Shunkai's value cannot be measured by qubit count alone within Japan. While ABCI-Q aims to be a shared platform that mixes and matches multiple computing resources, Shunkai builds a single neutral-atom machine from top to bottom, improving each layer together to support error correction and long-duration operation. That integration capability is what's new here.
What must be shown before external access
Going forward, IMS plans to open part of Shunkai to external users for developing applications and quantum error correction. IMS cites examples such as theoretical and software researchers developing error-correction techniques, and corporate researchers pursuing practical applications. The institute will also collaborate with Yaqumo—where Professor Ohmori serves as founder and executive advisor—on social implementation and hardware advancement. IMS also mentioned a vision of combining Shunkai with its existing shared-use supercomputer to develop a hybrid quantum-GPU computing center.
When external access will begin, who it will target, pricing, and application requirements have not been disclosed. Once researchers can actually use the hardware, a stricter evaluation than qubit count will begin. What matters for usability as a research platform includes the proportion of available time that can actually be used for computation, how long jobs wait in queue, how frequently calibration is needed, and how reproducible results are when circuits are repeated.
Shunkai has moved from the stage of having built its components to the stage of operating the integrated device and identifying its shortcomings. The first test will be whether, by 2028, it can raise its as-yet-undisclosed gate performance and uptime, and demonstrate the effectiveness of error correction on real hardware. Only after that, in 2030, once users outside the project can execute logical quantum circuits, will the significance of possessing a full stack in Japan show up in actual quantum computing performance.