Japan is building a 140MW AI factory for robots, and Nvidia is supplying all of it

Japan is building a 140MW AI factory for robots through the FRONTia Project, with Nvidia supplying all compute infrastructure including 13,750 Vera CPUs and 27,500 Rubin GPUs. The facility, established by private consortium Noetra Corp (majority-owned by SoftBank, NEC, Sony Group, and Honda), will produce open multimodal foundation models for AI agents, digital twins, and robots. Japan aims to capture over 30% of the global AI robotics market by 2040, with METI committing up to ¥1 trillion ($6 billion) through fiscal 2030.

Nvidia and a Japanese industrial consortium are building what Nvidia calls the world’s first national AI infrastructure for physical AI, and the specification is unusually concrete for an announcement of this kind. The AI factory will run 13,750 Nvidia Vera CPUs and 27,500 Rubin GPUs across 140 megawatts of data centre capacity, built on the Nvidia DSX platform with Vera Rubin NVL72 racks, Spectrum-X Ethernet, and BlueField DPUs. It is being established by Noetra Corp, and is meant to produce open multimodal foundation models for AI agents, digital twins, and robots. Noetra is the part worth understanding. It is not a government body. It is a private consortium majority-owned by SoftBank, NEC, Sony Group, and Honda, and it is the vehicle through which Japan is routing its physical AI ambitions. It is also the second Nvidia announcement out of Tokyo this week. The first signed most of Japan’s robotics establishment https://thenextweb.com/news/nvidia-japan-cosmos-coalition-robotics up to Nvidia’s open world models. This one supplies the compute those models will be trained on. The state’s role is upstream. METI and NEDO have jointly commissioned Noetra and AIST, the national research institute, to deliver the FRONTia Project, formally titled Development of Multimodal Foundation Models with a View to AI Robotics and Physical AI. The compute is the foundation for that work. The money behind it is large but more conditional than the headline figures suggest. METI has committed up to ¥1trn, roughly $6bn, across fiscal 2026 to 2030. The first tranche is ¥387.3bn, drawn from Japan’s GX Economy Transition Bonds, and only the first two years are locked. The rest runs through annual stage-gate review, which makes ¥1trn a ceiling rather than a promise. “Japan has launched the FRONTia Project, which will serve as the core of the country’s physical AI ecosystem,” said Ryosei Akazawa, Japan’s Minister of Economy, Trade and Industry. He framed it as combining foreign technology with domestic strengths, citing “ Japan’s onsite expertise and manufacturing technology infrastructure”. Jensen Huang reached, as he did in Tokyo the day before https://thenextweb.com/news/nvidia-japan-cosmos-coalition-robotics , for the same historical note. “Japan invented modern manufacturing,” he said. “Now, it is building the AI factories that will power the next industrial revolution.” What nobody has said is what any of it costs. Nvidia’s release contains no purchase price, no deal value, and no dollar figure attached to the chips. It describes Nvidia working with Noetra to launch a facility Noetra is establishing, which is a different thing from a government buying hardware, and the distinction matters for who carries the risk. Noetra’s chief executive, Hironobu Tamba, made the case for pooling. “ Bringing physical AI into the real world requires enormous computing, data and foundational technologies,” he said, “challenges no single company can solve alone.” The strategic target Nvidia cites is Japan’s AI Robotics Strategy, published in March, which aims for more than 30% of the global AI robotics market by 2040, an opportunity it estimates at $133bn. Separately, METI has set a target of 10 million AI-equipped robots across 18 sectors by the same year. Those two numbers come from different documents and should not be blurred together, but they point the same way. Japan is not trying to build a frontier lab. It is trying to put intelligence into machines it already knows how to manufacture, which is a narrower bet and, on its own terms, a more plausible one. The comparison that suggests itself is unflattering. Europe has spent two years on a gigafactory programme that keeps slipping https://thenextweb.com/news/eu-ai-data-centre-gigafactory-delays-funding-stumble , while Japan has a named operator, a named ministry, a bond-funded first tranche, and a rack count. It has also handed the entire stack to a single American vendor, which is the trade Europe has spent those two years arguing about https://thenextweb.com/news/gpuaas-is-reinforcing-the-illusion-of-european-ai-sovereignty . Sovereignty, in this telling, means owning the models and the deployment while renting the silicon. Whether that counts is a question Tokyo appears to have answered by not asking it. The “world’s first national AI infrastructure” framing, it should be said, is Nvidia’s own, in its headline and its body text. No independent party has made the claim, and several countries might contest it. Get the TNW newsletter Get the most important tech news in your inbox each week.