Japan Bets on Physical AI: Jensen Huang's Visit Signals a National Strategy Shift

From Saving NVIDIA to Reshaping Japanese Manufacturing

Thirty years ago, a nearly bankrupt American graphics chip company was rescued by a $5 million investment from Sega; thirty years later, the CEO of that company—NVIDIA's Jensen Huang—struck deals with almost all of Japan's industrial giants in Tokyo within two days, from Toyota to Fanuc, from SoftBank to Sony. Behind this role reversal lies a deep strategic shift in Japan's tech industry: making physical AI a new anchor for national competitiveness.

Three Major Deals Outline the Blueprint

Jensen Huang's Tokyo itinerary on July 15-16 left three key agreements:

Noetra—Japan's National AI Infrastructure. Led by 44 domestic companies including SoftBank, Sony, NEC, and Honda, the Japanese government has committed up to 1 trillion yen (approximately $6.2 billion) over five years to develop an autonomous physical AI foundation model. NVIDIA will build a supercomputing center called the "Vera Rubin AI Factory," equipped with 13,750 Vera CPUs and 27,500 Rubin GPUs, expected to be operational in 2028 with a computing power of 140 megawatts. The plan has three phases: a Japanese-language reasoning model in 2026, multimodal capabilities in 2028, and a "real-world native AI" capable of driving robots in 2030.

Robot Alliance Embraces Cosmos Collectively. Fanuc, Yaskawa Electric, Kawasaki Heavy Industries, Fujitsu, Hitachi, NEC, Sony, SoftBank, Kubota, and other industrial robot and manufacturing giants announced they would develop physical AI based on NVIDIA's Cosmos model. NVIDIA simultaneously launched Cosmos 3 Edge, which can run locally on the Jetson Thor chip. Honda R&D and Omron have begun testing shared control systems. Huang claimed, "Japan invented modern manufacturing; now it has the opportunity to redefine it for the smart industrial age."

Toyota Expands Autonomous Driving Collaboration. Toyota will not only equip its next-generation vehicles with the NVIDIA Drive platform, but also extend the collaboration to manufacturing simulation, automotive software, and traffic perception systems. However, Toyota's approach is relatively conservative, still requiring human drivers to take over at any time, contrasting with the full autonomous driving routes of Waymo and Tesla.

Physical AI: The Logic of Japan's Industrial Strategy

Behind these deals lies Japan's severe demographic reality. According to a plan by Japan's Ministry of Economy, Trade and Industry, Japan aims to introduce 10 million AI robots across 18 industries by 2040, with the government and private sector investing a total of approximately $6.5 billion (public and private combined). The "AI Robot Strategy" released in March this year sets a clearer goal: by 2040, Japan will capture more than 30% of the global AI robot market (approximately 20 trillion yen, or $133 billion).Japan clearly does not want the "brains" of factories and autonomous driving to run on AI from the US and China. The establishment of Noetra is a typical manifestation of sovereign AI—Tokyo wants its own data, its own computing power, and its own controlled AI models. But ironically, the hardware supporting this autonomous ambition still comes from Nvidia. Huang Renxun (Jensen Huang) attended the launch event alongside Japan's Minister of Trade and Industry, Akase Yasuyoshi, while Prime Minister Takayuki Sanae participated via video link, highlighting the government's emphasis on this initiative. The Takayuki administration has designated AI and semiconductors as core pillars of its growth strategy, planning to achieve 370 trillion yen (approximately $2.3 trillion) in public and private investment by 2040.

The Paradox of Dependency and Autonomy

The claim of being the "world's first national AI infrastructure" sounds ambitious, but in essence, it is still a computing pool built by Japan using sovereign capital to purchase American chips. Nvidia firmly controls the entry point to the physical AI ecosystem by providing hardware and the foundational model Cosmos. Japanese companies, meanwhile, compete on the application side through differentiation—using manufacturing data accumulated in domestic factories to refine niche scenarios such as robot control and autonomous driving. This division of labor is effective in the short term, but in the long run, if the foundational model layer remains entirely under US control, the question remains how far Japan's "sovereign AI" can go.

Jensen Huang's global tour—from Taiwan to South Korea and then to Japan—strengthens Nvidia's ecosystem lock-in at every step. In Tokyo, he not only had lunch with CEOs but also enjoyed skewers and whiskey with dozens of supply chain companies at a Kanda izakaya. This deep penetration indicates that Nvidia regards Japan as a key battleground for deploying physical AI, not merely as a chip buyer.

Japan's Choice in the Global Competitive Landscape

While the US leads in cloud-based large models and China catches up in application, Japan has chosen the unique strategic niche of physical AI. Its advantage lies in a century of manufacturing accumulation—precise robot control, high-quality factory data, and a mature supply chain system. But the challenges are equally clear: chip dependency, talent shortages, and whether it can achieve substantial breakthroughs at the AI algorithm and model level.

Noetra's plan for commercial robot AI by 2030 leaves a four-year window. During this period, Nvidia's chip iterations, Japanese companies' engineering transformation capabilities, and the international market's acceptance of Japanese-made AI will determine whether the country can revive the glory of "Made in Japan" in the era of physical AI.

Jensen Huang's Tokyo visit is not just a commercial tour but also a test of the direction of Japan's tech industry. Japan has chosen to bet big, with trillions of yen at stake, while the house is Nvidia.