Noetra, an AI initiative supported by 44 major Japanese companies, is accelerating the development of domestic foundation models for robots and industrial machinery.

SoftBank, Sony, NEC and Honda are among the participating companies. The project is also expected to receive more than ¥380 billion—approximately $2.5 billion—in government support during its first year.

Noetra is not trying to build a Japanese alternative to ChatGPT. Its objective is physical AI: systems that combine language, images, video, audio and sensor data so machines can understand and act in the physical world.

The company plans to begin developing foundation models capable of Japanese-language understanding and logical reasoning in fiscal 2026. It is targeting an “omnimodal” model capable of processing multiple forms of information by fiscal 2028, followed by what it calls “real-world native AI” that understands space and physical properties by fiscal 2030.

American and Chinese companies currently lead generative AI in foundation models, cloud infrastructure, semiconductors and access to capital.

Japan, however, retains a major industrial base connected to the physical world. Its strengths include automotive manufacturing, robotics, factory equipment, image sensors and logistics.

Noetra’s advantage may therefore extend beyond the capital supplied by its corporate members. Participating companies could potentially connect data from factories, vehicles, sensors, machinery and operational processes to a shared AI foundation.

Instead of competing directly with the US and China in text-generation models, Japan is attempting to create a different market around manufacturing and robotics. That is what makes Noetra strategically significant.

The project’s computing infrastructure will operate at a scale closer to a national programme than a conventional startup.

NVIDIA and Noetra plan to build a 140-megawatt AI infrastructure system using 13,750 Vera CPUs and 27,500 Rubin GPUs. Construction is scheduled to begin in April 2027, with operations expected to start in June 2028.

The infrastructure is intended to allow Japanese companies to use industrial data without relying entirely on foreign AI services.

There is, however, an obvious contradiction. While Noetra is presented as a domestic AI project, its computing capacity will remain dependent on NVIDIA, an American company.

Rather than complete technological independence, the project represents a more pragmatic form of sovereign AI: foreign semiconductors supporting models and industrial data managed within Japan.

Bringing together 44 major companies could provide Noetra with broad access to technology, customers and industrial data.

It could also make execution considerably more difficult.

Decisions about data sharing, intellectual property, commercial rights and the distribution of future revenues will require coordination among companies with different interests. If members are unwilling to contribute their most valuable data, the size of the alliance will not automatically translate into a competitive advantage.

Noetra must also do more than complete a large foundation model. It needs to produce systems that companies can deploy in manufacturing, logistics, healthcare, mobility and elderly care—and demonstrate that they can reduce costs or address Japan’s labour shortage.

Three questions will determine the project’s future.

First, how much valuable industrial data will participating companies actually share?

Second, can Noetra convert research and infrastructure into products and services that generate revenue?

Third, will the gap with international competitors continue to widen while the project waits for its computing infrastructure to become operational in 2028?

Noetra is not a conventional startup. It is closer to a national industrial project connecting government funding, major corporations and a foreign semiconductor provider.

If it succeeds, Japan could establish a distinctive position in physical AI by building on its strengths in manufacturing and robotics. But funding and corporate participation alone do not guarantee commercial success.

Noetra’s value will ultimately depend on whether it can transform the technology and industrial data it gathers into products used beyond the companies and institutions that created it.

Adrià Mas Rodríguez
Editor, Sakime

Covering East Asian startups, technology, investment and business developments.

Connect on LinkedIn
linkedin.com/in/adrià-mas-rodríguez-453279205/