Seven organizations, including Nautilus Technologies (hereafter Nautilus), announced on October 26, 2026 that they will jointly develop a next-generation distributed computing platform that connects quantum computers to business data over optical networks. Development began on September 29. By combining multiple quantum computers with a domestically developed database, the project aims to optimize the allocation of resources such as the GPUs used for AI inference and the electricity that powers them. The group also lists scaling up to several hundred thousand qubits as a goal. However, what was announced is a development plan, not measured performance results. Evaluating its practicality will require verifying the entire process, from preparing business data to applying computation results, and not just the scale of the quantum computation.
Four technologies to connect quantum computing to business
The joint development will integrate four technologies: a cluster that links multiple quantum computers, an integration layer for business systems, a database, and an optical network connecting sites. For business data processing, the project will use "Tsurugi," a domestically developed open-source relational database. Tsurugi will extract, transform, and update the data needed for quantum computation, serving as the bridge between quantum computers and existing business systems. Simply installing a quantum computer does not provide a mechanism to pass business data into a computation and feed the results back into operations. This development targets that whole chain of data processing as well.
On the communications side, the plan is to configure the quantum computers and Tsurugi as separate clusters and connect them over an ultra-low-latency optical network. Long-distance links will use "optical NICs," network interfaces that support optical communication. However, no specific latency figures have been published, and the announcement does not claim a demonstration of transmitting quantum entanglement over long distances.
The seven participants each handle a different technical area. Nautilus leads the overall project and the database, Quanmatic handles algorithms and implementation, and Giria is responsible for connecting the systems. 1FINITY develops the network and optical NICs, while IIJ manages everything from procuring to operating the server environment at its Shirai Data Center Campus in Shirai, Chiba Prefecture. On the academic side, Waseda University works on distributed quantum computing algorithms and Nagoya University on distributed indexing technology. The structure integrates development from computational algorithms through data management, communications, and operations, and the work is supported by a grant from NEDO's "Post-5G Information and Communication Systems Infrastructure Enhancement R&D Project."
Optimizing GPU and power allocation for AI inference
One of the main intended uses is the optimal allocation of widely distributed computing, power, and communication resources. The aim is to combine factors such as renewable energy output, electricity prices, network usage, and GPU utilization to allocate each resource efficiently. The "watt-bit collaboration" concept treats power supply and information processing as a single system, optimizing resource allocation while accounting for the constraints of both.
For more localized uses, the group is also considering a mechanism for dynamically assigning GPUs among users and services. Quantum computers are expected to handle the optimization calculations that weigh these multiple conditions. The quantum computers would not run AI inference itself; they would address the problem of how to allocate the GPUs and related resources that perform it.
For practical use, it will be important to distinguish the speed of the quantum computation itself from the processing time of the whole workflow. In real operations, multiple steps occur: retrieving the necessary information from the database, converting it for quantum computation, sending it over the network, and reflecting the results in business systems. Even if the calculation that finds the optimal allocation becomes faster, the overall process will not necessarily shrink by the same proportion. This is an evaluation issue that follows from the system architecture; it is not a measured finding that data preparation or communication became bottlenecks in this development.
When comparing performance with existing optimization methods, the input business data and constraints must be held constant, and both the time to obtain a result and the quality of the resulting allocation need to be verified. Cost reduction and improved power efficiency are among the development goals, but no before-and-after measurements or reduction rates have been published. How far quantum computing can outperform conventional methods remains a question for future verification.
Completing development in fiscal 2028 and offering services around 2030 are separate stages
The plan sets three stages: platform development, application to specific uses, and commercial service. Platform development is targeted for completion by fiscal 2028, application to specific uses from fiscal 2029 onward, and SaaS delivery of the wide-area optimization platform around 2030. Because each stage covers a different part of the process, the timings should be considered separately.
| Timing | Target | Status |
|---|---|---|
| Through FY2028 | Platform development | Development deadline |
| FY2029 onward | Application to specific uses | Planned use |
| Around 2030 | SaaS delivery of the wide-area optimization platform | Delivery goal |
This schedule organizes the future outlook and the description of wide-area optimization use cases given in the joint announcement by the seven parties dated October 26, 2026. The table uses the fiscal-year and calendar-year notation from the announcement as is. All of these are current plans or targets, and the start date of commercial service has not been finalized.
The "around 2030" target in particular corresponds to the concept of offering the wide-area optimization platform as SaaS. It does not mean that all functions of the quantum computing platform, or all the uses under consideration, will be available at the same time. Other application areas mentioned include automotive energy management and financial portfolio optimization. The group also envisions applications such as production planning in manufacturing, and staffing and structural design in construction. However, none of these has a demonstrated deployment so far; they are at the stage of examining future applicability.
How to evaluate the goal of several hundred thousand qubits
The figure of several hundred thousand qubits represents the computational scale the project aims for in the future. However, the announcement does not specify whether it refers to physical or logical qubits, and it does not disclose the quantum computing method or the quantum error correction conditions. The figure therefore cannot be read as showing the realization of a general-purpose, error-corrected quantum computer or a demonstration of quantum advantage over existing computers.
The material available is a joint press release from the seven parties, not a presentation of results from a peer-reviewed paper or preprint. The announcement and its related links give no information such as paper authors, journal, or DOI, and no independent replication or performance comparison is presented. That said, this does not mean that no related papers exist.
Experimental conditions such as the number of actual machines and trials are not disclosed, nor is comparative data on processing time, communication latency, or power consumption. No correlation analyses, intervention experiments, or simulation results are shown either, so the effect of improving power efficiency has not yet been demonstrated. The figure of seven also refers to the number of participating organizations, not a sample size in an experiment. Still, the absence of published evaluation results at this stage does not in itself indicate that the development has failed.
In future evaluations, a key question is whether the qubit configuration, computation method, and experimental conditions are made clear. Beyond that, the benchmark for practicality will be whether the whole process, from preparing business data to applying optimization results, can be compared against existing methods under identical conditions. If processing time and the quality of optimization results are shown quantitatively, along with effects on power consumption and operating costs, companies will be able to judge more concretely whether adopting the platform is worthwhile for their own GPU operations and power management.