Quantinuum RIKEN Upgrade Puts 56 Qubits on Reimei

Quantinuum has completed an upgrade of its Reimei quantum computer at RIKEN to the company’s System Model H2, the two organisations said on 10 November 2026. Quantinuum’s release says the machine, installed at RIKEN and operated as part of the NEDO-funded JHPC-quantum project, “has successfully completed its upgrade to System Model H2.” RIKEN’s Japanese announcement adds that the upgraded Reimei now has 56 fully connected qubits. An English RIKEN note points readers to the Quantinuum release and records a ceremony on the Wako campus the same day.

Alongside the completed upgrade, the partners announced two new memoranda of understanding. Those documents are statements of intent, not build contracts or a timetable for a fault-tolerant machine.

The machine is upgraded; the next computer is only a memorandum

Quantinuum and RIKEN say they will use the upgraded platform to combine quantum computing, supercomputing and AI, with work aimed at electronics, energy and life sciences. RIKEN’s Mitsuhisa Sato, division director of the Quantum-HPC Hybrid Platform Division at the RIKEN Center for Computational Science, said the partners would now be able to tackle “more complex and larger-scale scientific problems that were difficult to address with the previous H1 system,” and that a newly signed memorandum would “pave the way for expanding our research into the use of quantum computing for AI for Science.”

One memorandum covers work on a “Made with Japan” fault-tolerant quantum computer, involving Japanese universities and domestic suppliers. The other, involving RIKEN, Quantinuum and SoftBank Corp., is to explore connecting quantum computers to AI data centres in Japan. Quantinuum’s wording is that the partners “will collaborate with SoftBank Corp. to develop a path to connect quantum computers directly with AI data centers.” RIKEN’s English note says they will “explore the connection.” Neither text gives a construction schedule, a site, a qubit target or a commercial commitment for that next machine.

Dr Rajeeb Hazra, Quantinuum’s chief executive, said: “Japan’s quantum ambitions require advanced capabilities today and a sustained commitment to developing what comes next.” He described the Reimei upgrade as expanding what the partners can explore, and the fault-tolerant machine and SoftBank work as what they are “building on this work to develop” and “exploring.”

Two early projects, including a preliminary magnetism result

Quantinuum says two research projects have already begun on the upgraded system. In one, the partners are using RIKEN’s JHPC-quantum GPU supercomputer, ROQUO, to train an AI model on Reimei-generated quantum data for magnetism simulations. “Preliminary results suggest the model can make more accurate predictions than AI models trained on classical simulations alone and more precise predictions than the quantum computer itself,” the company says. That is a preliminary, partner-reported result, not a finished, independently reviewed finding, and the releases do not publish the underlying measurements.

In the second project, the partners are combining Reimei, RIKEN’s Fugaku supercomputer and NVIDIA’s evolutionary AI agent running on ROQUO to simulate a photoactive biomolecule. Quantinuum presents that work as support for RIKEN’s AI-for-science effort and as a possible route to better understanding of how biomolecules respond to light. It is a stated research plan, not a completed materials or energy result.

The upgrade and the existing Reimei operation sit inside the JHPC-quantum project, which both parties say is funded by NEDO, Japan’s New Energy and Industrial Technology Development Organization. That funding line explains how the machine is being run today. It is not, in the texts published this week, a pledged budget for the “Made with Japan” fault-tolerant computer.

An upgraded machine is not the same as a committed next computer

The fact that can be stated plainly is the upgrade itself. Reimei is now an H2 system, and RIKEN says it has 56 fully connected qubits, a larger and more connected machine than the previous H1 installation. That is a completed hardware change at an existing national laboratory, paid for through a project that was already running. It is the sort of step TSN has already seen Quantinuum claim on other machines, including a very high logical fidelity result on its Helios system, and it sits in the same Japanese research landscape as RIKEN’s own silicon spin-qubit work.

The memoranda should be kept in a different box. An MOU can be a real political and scientific signal, especially when it names universities, suppliers and SoftBank. It does not, by itself, commission a fault-tolerant computer or wire a quantum machine into a commercial AI data centre. The SoftBank text is about exploring a connection, and the “Made with Japan” text is about joint research and development towards a future machine. No date is attached to either. Other Japanese industry groups have also begun joint quantum and AI materials research; those efforts, like this one, are easier to over-read as a finished industrial plan than the documents actually support.

The magnetism study needs the same restraint. If a model trained on Reimei data eventually outperforms models trained only on classical simulations, that would be a useful hybrid-computing result. The partners themselves call the present findings preliminary, and they have not put a paper alongside the announcement. Until those results are reported in enough detail to be checked, they are a reason to watch the project, not a reason to treat the upgrade as a proven materials breakthrough.

The next facts that would change the story are a detailed account of the 56-qubit H2 system in use, a published magnetism result, and any later agreement that turns either memorandum into a funded machine or a dated data-centre link. Until then, the Quantinuum RIKEN news is a completed upgrade and two exploratory documents, not a new, committed quantum computer.