Japan is spending $6.2 billion to build a sovereign AI for its factories and robots, and it will all run on Nvidia's chips

Japan is spending $6.2 billion to build an AI it can call its own. The entire project will run on Nvidia's chips.

During a two-day visit to Tokyo in July, Nvidia CEO Jensen Huang stood alongside Japan's trade minister Ryosei Akazawa to launch what Nvidia calls "the world's first national AI infrastructure" 1. The initiative, called Noetra, assembles roughly 44 Japanese firms with SoftBank, Sony, NEC, and Honda at its core, backed by up to 1 trillion yen ($6.2 billion) in government funding over five years 2. Japan's goal is to build its own foundation models for robots, vehicles, and factory floors rather than depend on American or Chinese AI 2.

The sovereignty is real in intent. The dependency is structural in practice.

The data center powering Noetra will be an Nvidia Vera Rubin AI factory: 13,750 Vera CPUs and 27,500 next-generation Rubin GPUs delivering 140 megawatts of compute 1. Noetra will oversee the facility, but the architecture, the silicon, the networking platform (Nvidia Spectrum-X Ethernet), and the reference design (Nvidia DSX) all come from one company in Santa Clara 1. The factory is expected to come online in 2028 2.

Then there is the software layer. Japan's robotics coalition has committed to building on Nvidia's Cosmos models. FANUC, Yaskawa Electric, Kawasaki Heavy Industries, Fujitsu, Hitachi, NEC, Sony Group Corporation, SoftBank Corp., Kubota, and AIRoA intend to join what Nvidia calls the Cosmos Coalition 3. Nvidia unveiled Cosmos 3 Edge in Tokyo, a 4-billion-parameter model designed to run on Nvidia Jetson Thor computers inside the machines themselves 3. Fujitsu is exploring a collaborative control platform integrating Nvidia's physical AI stack, with FANUC, Yaskawa Electric, and Kawasaki Heavy Industries as integration partners; Honda R&D and Omron are already building on the tools 2.

The models Japan trains will be its own. The weights, the training data, the Japanese-language reasoning: all sovereign. But the compute fabric underneath, the development libraries (Nvidia NeMo, Nemotron, Isaac GR00T), the edge inference chips, and the simulation environments all belong to Nvidia 1. Noetra's CEO Hironobu Tamba said the pretrained weights of Noetra's foundation models will be made broadly available to Japanese developers 1. The openness is the sovereign part. The 27,500 Rubin GPUs underneath are the dependency.

Japan is not the first country to wrap a national AI strategy around Nvidia's compute. Huang ran a similar playbook in Seoul last fall, securing a 50,000-GPU deal 2. The structure is consistent across these initiatives: sovereignty at the model and data layer, dependency at the silicon and software layer.

For Japan, the economic logic overrides the contradiction. The country faces a shrinking workforce and wants 10 million AI-equipped robots across 18 sectors by 2040, backed by $65 billion in public and private investment 2. Japan's AI Robotics Strategy, released in March, targets more than 30% of the global AI robotics market by 2040, a market Tokyo values at roughly $133 billion 1. The Takaichi administration has placed AI and semiconductors at the center of a growth plan targeting 370 trillion yen ($2.3 trillion) in public and private investment by 2040 2.

METI, Japan's Ministry of Economy, Trade and Industry, is funding the foundation model program through what it calls the FRONTia Project 1. Noetra's roadmap runs in three stages: a reasoning model focused on Japanese-language skills starting in fiscal 2026, an omni-modal version handling text, images, video, and audio by 2028, and what Noetra calls "Real-world Native AI" for running robots by 2030 2. As the factory scales, it will support training of trillion-parameter models 1.

When Huang appeared with Akazawa on July 16, with Prime Minister Sanae Takaichi joining by video, the framing was partnership 2. The partnership tilts in one direction. Japan brings the factories, the robots, the industrial data, and the demand. Nvidia brings the chips that make it all compute, the models that give the robots their intelligence, and the platform that connects the entire system.

Japan's sovereign AI will be genuinely Japanese in what it knows and who controls the data. It will be entirely Nvidia in how it thinks.