Jensen Huang‘s two-day Tokyo visit in July 2025 landed a trifecta of deals that map out Nvidia‘s bet on physical AI in Japan. The core tension is that Japan wants to own the software brain for its factory robots and vehicles—through the Noetra consortium of 44 domestic firms, backed by ¥1 trillion ($6.2B) over five years—but still needs American silicon to build it. Nvidia is building a Vera Rubin AI factory with 13,750 Vera CPUs and 27,500 Rubin GPUs, expected in 2028. Japan’s push for AI sovereignty runs entirely on Nvidia‘s next-generation chips, creating a strategic dependency even as Tokyo aims to capture over 30% of the global AI robotics market by 2040.

Concretely, three projects emerged. Noetra will oversee Japan’s homegrown physical-AI foundation models in three stages: a reasoning model strong in Japanese by fiscal 2026, an omni-modal version by 2028, and “Real-world Native AI” for robots by 2030. A robotics coalition—Fanuc, Yaskawa, Kawasaki Heavy, Fujitsu, Hitachi, NEC, Sony, SoftBank, Kubota and AIRoA—committed to building on Nvidia‘s Cosmos models, specifically the new Cosmos 3 Edge variant running on Jetson Thor chips inside the machines themselves. Toyota extended its existing Nvidia Drive platform work into manufacturing simulation and production-line design, though taking a more conservative approach than Waymo or Tesla with advanced driver assistance that still requires a driver. Huang also courted supply-chain chiefs over dinner in a Kanda izakaya, running the same playbook from his Taiwan and Seoul visits.

The serious takeaway is that Japan’s industrial strategy is now explicitly intertwined with Nvidia‘s hardware roadmap, and Tokyo is spending massively to make it work. Facing a shrinking workforce, Japan targets 10 million AI-equipped robots across 18 sectors by 2040, backed by $65B in public and private physical-AI investment. The Noetra factory, which Nvidia bills as “the world’s first national AI infrastructure,” is the clearest bet yet that Japan wants independence from U.S. and Chinese AI—but for now, that independence runs on American chips. Builders should watch whether the three-stage Noetra model timeline slips, whether Cosmos adoption on the factory floor produces real data advantages, and whether the sovereign-AI play can survive a geopolitically motivated shift in Nvidia‘s chip supply.