On August 24, 2026, Japan’s Institute for Molecular Science announced that Shunkai, the country’s first full-stack neutral-atom quantum computer, had begun operation. It starts with roughly 50 qubits. By the arithmetic that dominates public discussion of quantum computing, that number is unremarkable. By the arithmetic that actually matters strategically, the number is close to beside the point.
Shunkai was built by a team led by Professor Kenji Ohmori under Goal 6 of Japan’s Moonshot Research and Development Program, administered by the Japan Science and Technology Agency. Atomic qubits are held in arrays of optical tweezers — tightly focused laser beams — and manipulated with microwave and laser pulses. The system is named for Harumi Shibukawa, the Edo-period astronomer behind Japan’s first domestically calculated calendar: a reference to precise control, not to scale.
The Institute for Molecular Science, part of the National Institutes of Natural Sciences, led system development and integration as principal investigator. Hitachi developed the software stack. Infleqtion, a United States company, supplied the quantum processing unit stack. Yaqumo Inc., a Japanese neutral-atom startup out of the Ohmori and Takahashi laboratories, is to collaborate on implementation and upgrades. The plan is to grow to roughly 500 qubits, open the system partially to external users, and integrate it with the institute’s shared supercomputer facility as a quantum–GPU hybrid computing centre. The Moonshot second stage, begun in April 2026, targets 10,000 physical qubits with error detection and correction by March 2031.
The interesting claim in the Shunkai announcement is not a qubit count. It is the word “full-stack” — an assertion about who operates, integrates, and upgrades the machine.
The wrong question: how many qubits?
Physical qubit count is the most quoted and least informative figure in quantum computing. It is not meaningless — no useful fault-tolerant machine will be small. But a single integer cannot rank systems built on different physics.
Comparing a neutral-atom array to a superconducting processor or a trapped-ion register by qubit count alone ignores gate and measurement fidelity, coherence relative to gate duration, connectivity, the physical-to-logical overhead of error correction, calibration stability, availability, and usable circuit depth. A smaller machine with better-controlled, better-connected qubits and greater usable depth can outperform a larger one on practically relevant workloads.
Nor has any modality won. Neutral atoms offer operation without dilution refrigeration, reconfigurable geometry, and intrinsically identical atomic transitions — though practical control fidelity still varies across an array. Superconducting circuits offer fast gates and mature fabrication; trapped ions, high fidelity and all-to-all connectivity; photonics, room-temperature operation for many components and a natural path to networking, though leading single-photon detectors remain cryogenic; spin qubits, a route through existing semiconductor manufacturing. Each carries unresolved engineering problems. Anyone declaring a winner in 2026 is forecasting, not reporting.
The strategic error, though, runs deeper than picking the wrong technical metric. Qubit count measures a component. Strategic capability is a property of a system and its operators.
What “full stack” actually means
A working quantum computer is a small industrial plant. Beneath the qubits sit vacuum systems, magnetic shielding, and precisely stabilized lasers. Above them: modulators, waveform generators, FPGA-based real-time control, readout, and calibration that runs continuously because the machine drifts. Above that: a compiler mapping circuits to native gates and geometry, a scheduler and runtime, error-detection logic, a job queue, authentication, telemetry, and the interfaces through which classical HPC and GPU resources exchange work with the processor.
“Full stack” claims that one organization can operate all of that as a single system — and, critically, change any layer of it. That is a different assertion from owning it, and a very different one from having manufactured it.
Sovereignty does not mean autarky
Shunkai is instructive precisely because it is not domestically self-sufficient. Its quantum processing unit stack came from an American vendor; Infleqtion has described itself as the only foreign quantum partner selected by JST for the Moonshot program. Its lasers, optics and vacuum hardware draw on a global supplier base no country reproduces alone.
That does not disqualify the sovereignty claim. It clarifies it. The layers Japan kept are architecture, integration, the software stack, operations, the user-access decision, the HPC coupling, and the upgrade path — the last institutionalized through Yaqumo, a domestic company positioned around implementation and upgrade. Foreign components participate inside an architecture a Japanese public institute defines and runs.
Where a dependency sits geopolitically matters too. Japan’s QPU comes from a United States supplier, and both sit inside the same coordinated regime: the September 2024 BIS interim final rule imposed worldwide licensing on certain quantum items while creating License Exception IEC for allies with equivalent controls — Australia, Canada, France, Germany, Italy, Japan, Spain and the United Kingdom. Allied supply is governed supply: sourcing from an aligned partner lowers the probability of coercion, but the exception exists because the item is controlled, and it holds only while both governments stay aligned. Friend-shoring does not clear the dependency from the ledger.
These distinctions collapse easily in policy conversation, so they are worth separating. Manufacturing is who fabricates a part; ownership, who holds title; intellectual property, who may reproduce or modify the design; integration, who assembles the system; operation, who runs it day to day; control, who can change it, restrict it, or refuse a change under pressure; dependency, what happens when a supplier withdraws. A country can answer these seven questions differently for every layer of the same machine.
Maple Quanta has argued in the AI context that sovereignty is not independence but strategic control, and that it is a spectrum rather than a binary. Quantum infrastructure makes the same point with harder physics.
Residency versus sovereignty
Residency tells you where a system sits. Sovereignty tells you who can ultimately decide what it does.
A quantum computer inside national borders is not meaningfully sovereign if its calibration software is a vendor black box, its firmware arrives on the vendor’s schedule, its scheduler and identity layer run in a foreign-controlled cloud, its maintenance needs an engineer who cannot be replaced domestically, and its telemetry leaves the country. That is a physical asset and a service contract.
The operative measures are the ones this firm has applied to sovereign AI procurement: exit time and exit cost. How long, and at what price, could the operator substitute a critical layer if the supplier were lost to export control, acquisition, sanction, or commercial failure? Shunkai’s answer for its QPU is not public, and should not be assumed favourable.
What limits the damage is encapsulation. Because the Institute for Molecular Science holds the compiler, the runtime, the HPC coupling and the operating knowledge, losing the QPU supplier would invalidate a component rather than the system around it. Substituting another vendor’s hardware would still force real re-engineering — control, calibration and the lower software layers are never fully vendor-neutral — but retained integration knowledge makes that work bounded rather than institutionally impossible. That is the difference between a long exit time and an unbounded one — and why the layers above the hardware carry disproportionate strategic weight.
The quantum sovereignty stack
A single sovereignty label conceals more than it conveys. The following framework decomposes the claim into ten layers, each with one diagnostic question. It is intended to be scored per layer — controlled, shared, dependent, or undocumented where public evidence is insufficient — never summed into a rank.
| Layer | The diagnostic question |
|---|---|
| Compute | Who designed the processor, and could its physics be reproduced domestically? |
| Control | Are lasers, electronics and calibration routines inspectable and modifiable by the operator? |
| Software | Who controls the compiler, runtime and scheduler — and can they be forked? |
| Integration | Who couples the quantum processor to classical HPC, GPU and storage infrastructure? |
| Access | Who decides which users, workloads and jurisdictions are admitted? |
| Data | Where are inputs, results and system telemetry processed and retained, and under whose law? |
| Operations | Can the system be run, calibrated and repaired without foreign personnel? |
| Supply chain | For each critical dependency, what is the exit time and exit cost? |
| Knowledge | Does a domestic workforce understand the system well enough to modify it? |
| Upgrade authority | Who approves the next architecture, and who can veto it? |
What Japan’s approach reveals
Three features of the Shunkai arrangement are institutional rather than technical, and they are the transferable part.
First, a public research institute holds the integrator role. Integration knowledge — how the layers interact, why calibration fails, what breaks when a supplier changes a part — is the least documented and least transferable asset in the system. Japan placed it in a national institution rather than a vendor relationship.
Second, the machine is being attached to existing shared supercomputing infrastructure. Near-term quantum utility is hybrid: classical resources prepare, verify and post-process quantum results. A processor that cannot be tightly coupled to national HPC is a laboratory instrument, not infrastructure.
Third, an upgrade path was institutionalized alongside first operation, through a domestic company. Sovereignty over the machine you have is temporary; sovereignty over the machine you get next is durable.
None of this is yet proven. Shunkai is early-stage, roughly 50 qubits, partially open, with fault tolerance nearly five years out on the program’s own timeline. The claim here is about the shape of the strategy, not about demonstrated computational advantage.
The Canadian question
Canada is not behind on quantum technology. It is arguably ahead of most of the world on the layers Shunkai imported.
The National Quantum Strategy, launched in January 2023 with $360 million, was reinforced by Budget 2025 with $334.3 million over five years. In December 2025, ISED launched Phase 1 of the Canadian Quantum Champions Program: up to $92 million, up to $23 million each to four Canadian-headquartered firms — Anyon Systems, Nord Quantique, Photonic and Xanadu Quantum Technologies — with progress assessed through a new National Research Council Benchmarking Quantum Platform initiative and an explicit link to the forthcoming Defence Industrial Strategy. The release ties the program to “protecting national sovereignty”; Quantum Industry Canada’s chief executive describes the aim as “scalable, sovereign capability with long-term value.”
That is serious, well-aimed industrial policy. It anchors companies, funds fault-tolerance roadmaps across four modalities, and creates an independent benchmarking function — an assurance instrument rather than a subsidy.
Canada also already operates quantum hardware — and the two systems on Canadian soil demonstrate the framework better than an argument could. MonarQ, a 24-qubit superconducting machine built by Anyon Systems, hosted at ÉTS in Montréal and operated by Calcul Québec, was inaugurated in September 2024 and appears in the Digital Research Alliance of Canada catalogue. PINQ², a non-profit founded by the Université de Sherbrooke and Québec’s economy ministry, has since 2023 operated a 127-qubit IBM Quantum System One at Bromont, which it describes as the only one in Canada.
| Layer | MonarQ — 24 qubits Anyon Systems / Calcul Québec | IBM Quantum System One — 127 qubits IBM / PINQ², Bromont |
|---|---|---|
| Compute | Controlled — Anyon-designed and -built processorc | Dependent — IBM architectured |
| Control | Controlled — in-house cryogenics and control electronicsc | Dependent — IBM control stackd |
| Software | Controlled — Snowflurry, open source and forkablec | Shared — open-source SDK; IBM operates the service runtimed |
| Integration | Controlled — inside national research computinga,b | Controlled — PINQ² couples it to its HPC and AI servicesd |
| Access | Controlled — Alliance allocation to Canadian institutionsb | Controlled — PINQ² sets service termsd |
| Data | Controlled — operator and vendor both Canadiana,c | Shared — operator commits user data stays in Canada; vendor telemetry and diagnostic flows undocumentedd |
| Operations | Controlled — Calcul Québec operates; any vendor support is domestica,b,c | Shared — PINQ² operates; IBM’s maintenance role not publicd |
| Supply chain | Undocumented — major subsystems in-housec; component-level imports not public | Dependent — vendor-integrated systemd |
| Knowledge | Controlled — design and operating knowledge domestica,c | Dependent — deep-modification knowledge is IBM’s |
| Upgrade authority | Controlled — Anyon and the host decidea,c | Dependent — IBM roadmapd |
Neither reading is a criticism. PINQ² gives Canadian users a class of machine no domestic vendor yet builds, operates it itself, and makes a real residency commitment — more than most countries have. The point is narrower: the larger system is not the more sovereign one, and no single number would have told you that.
The NQS quantum computing roadmap is candid about supply-chain gaps — semiconductor fabrication, helium-3, high-purity silicon, neon — and about the need to capture more of the chain domestically. What remains absent is a designated federal operator: a single institution with a national mandate, security accreditation for sensitive or classified workloads, and standing authority over stack integration and upgrades. Regional hosts and a milestone-based industrial program are both valuable; neither is that. In the Japanese arrangement, that role has one address.
| Emphasis | Japan (Moonshot / Shunkai) | Canada (NQS / CQCP) |
|---|---|---|
| Primary instrument | National institute as system integrator and operator | Milestone funding to anchor domestic companies |
| Hardware source | Imported QPU stack inside a domestic architecture | Four domestic modalities under domestic IP |
| Software layer | Domestic corporate partner (Hitachi) | Strong domestic SDK and software ecosystem |
| Assurance | Internal to the program | Independent NRC benchmarking function |
| Named operator | Institute for Molecular Science | Regional hosts (Calcul Québec, PINQ²); no federal mandate |
Canada’s position is close to the inverse of Japan’s. Japan imported at the component layer and built at the institutional one; Canada has extraordinary depth at the component and IP layers and has built its operational layer regionally rather than nationally. The Canadian gap is arguably the cheaper to close, because the hard part — the underlying technology — already exists domestically.
From quantum science to quantum infrastructure
Possessing a technology and operating strategic infrastructure are different achievements requiring different institutions. Excellent research produces papers, patents, companies, and talent. Infrastructure requires service-level commitments, calibration discipline, on-call staff, access governance, security accreditation, and an owner who is accountable when it is unavailable.
Quantum computing is crossing that line. When Ottawa frames quantum as strategic infrastructure with defence applications — as the December 2025 announcement does — operational questions follow immediately: which sensitive workloads run on which system, under whose authorization, with results retained where, maintained by whom, and upgradable on whose decision. Those are not research questions.
Closing that gap would not require Canada to abandon its pluralistic hardware strategy or crown a winning modality. It requires naming an integrator, and at least three routes are available:
- An expanded NRC mandate. Grow the National Research Council’s Benchmarking Quantum Platform role from assessment into full-stack operation, with an accreditation path for defence and other sensitive workloads.
- A federated Alliance consortium. Formalize a multi-site national integration network under the Digital Research Alliance of Canada — MonarQ already runs inside that federation.
- A chartered infrastructure authority. Establish a federally chartered public-private body on the PINQ² model, bridging domestic QPUs into national HPC and carrying accreditation a regional non-profit cannot.
Each carries different costs and constituencies, and the choice is a political one. It is also fair to ask whether Canada needs a dedicated federal operator yet: formalizing a federated integration function first, and letting workload maturity and security requirements drive anything more, is a defensible posture while pre-commercial infrastructure is expensive to hold. What matters analytically is that today the choice is implicit — and implicit arrangements do not survive contested conditions.
The Maple Quanta view
The useful reading of Shunkai is not that Japan has taken a lead in quantum computing — on measured performance it has not, and its own timeline says so. It is that Japan chose an operational rather than proprietary definition of quantum sovereignty, and built an institution around it before the technology matured.
For Canadian executives and policymakers, the framework above converts a slogan into a checklist. A sovereignty claim about any quantum — or AI — system should be met with the same demand: name the layer, name the controller, and state the exit time.
The countries that lead the quantum era are unlikely to be those that manufacture every component or announce the largest qubit number. They will be those that institutionalized control early — deciding which layers had to remain their own, and building the institutions to hold them before the technology forced the question.
Sources
Source note. Technical and programmatic facts about Shunkai are drawn from the Institute for Molecular Science announcement of August 24, 2026, the JST Moonshot Goal 6 program pages, and partner press releases; roadmap figures including 500 qubits and 10,000 physical qubits by March 2031 are stated program targets, not demonstrated results. Canadian program facts are drawn from ISED, Government of Canada, Calcul Québec, Digital Research Alliance of Canada and PINQ² publications; MonarQ and IBM Quantum System One specifications are as published by their operators. The ten-layer sovereignty framework, the layer-by-layer scoring of MonarQ and the Bromont IBM system on the controlled / shared / dependent / undocumented scale, the two-model comparison, and the residency–sovereignty argument are Maple Quanta analysis, derived from the operator and vendor documentation cited below.
- Institute for Molecular Science, National Institutes of Natural Sciences. “Japan’s first full-stack neutral-atom quantum computer ‘Shunkai’ is now operational.” August 24, 2026.
- Japan Science and Technology Agency. Moonshot Goal 6: Realization of a fault-tolerant universal quantum computer that will revolutionize economy, industry, and security by 2050.
- Moonshot Goal 6 — Cold-atom quantum computing project (OHMORI Kenji PM). Project overview.
- Hitachi, Ltd. “Hitachi to participate in Phase 2 research projects of JST Moonshot Goal 6.” March 2026. Hitachi’s Phase 2 scope is described as system performance evaluation software for the neutral-atom system.
- Infleqtion, Inc. “Infleqtion collaboration with Japan Moonshot Program achieves major milestone.” August 2026.
- Yaqumo Inc. Company overview.
- Innovation, Science and Economic Development Canada. Canada’s National Quantum Strategy.
- Innovation, Science and Economic Development Canada. National Quantum Strategy roadmap: Quantum computing.
- Government of Canada. “Minister Solomon announces major new quantum initiative.” December 2025.
- Calcul Québec. “Official inauguration of MonarQ, a quantum computer dedicated to research and innovation.” September 2024.
- Digital Research Alliance of Canada. MonarQ service description and access conditions.
- PINQ². Quantum service offering, IBM Quantum System One, Bromont.
- Anyon Systems. Products: full-stack superconducting quantum systems. In-house processor, cryogenics, control electronics and the Snowflurry open-source software library.
- Bureau of Industry and Security, U.S. Department of Commerce. “Commerce Control List Additions and Revisions; Implementation of Controls on Advanced Technologies Consistent With Controls Implemented by International Partners.” 89 FR 72926, September 6, 2024. Establishes License Exception IEC at 15 CFR § 740.24.
- The Quantum Insider. “Japan’s full-stack neutral-atom quantum computer is operational.” August 24, 2026. Secondary reporting; used to locate the announcement.
Disclosure: Maple Quanta Inc. provides quantum technology advisory, AI evaluation, assurance, governance, and technical auditing services. Readers should consider this commercial context when assessing the analysis. This Insight is for general informational purposes only and does not constitute legal, procurement, cybersecurity, investment, or regulatory advice.