Jensen Huang's Tokyo Trip Just Reshaped Japan's Robot Future

Key points

• Nvidia CEO Jensen Huang visited Tokyo on 15 and 16 July 2025, securing deals across Japan's robotics, automotive, and chip-supply industries.
• A Japanese government-backed consortium called Noetra, drawing on roughly 44 domestic firms, is committing up to 1 trillion yen (about 6.2 billion US dollars) over five years to build homegrown AI for robots and factories.
• Nvidia will build what it calls "the world's first national AI infrastructure" for Noetra, a giant data centre packed with next-generation chips, due to open in 2028.
• Ten robotics and manufacturing companies, including Fanuc, Kawasaki Heavy, Sony, and SoftBank, agreed to build products on Nvidia's Cosmos AI models, which are designed to teach machines how to move in the real world.
• Japan wants 10 million AI-equipped robots working across 18 industries by 2040, backed by 65 billion dollars in public and private investment.

Jensen Huang, the chief executive of Nvidia, the American company whose chips power most of the world's serious AI work, touched down in Tokyo on 15 July. He left 48 hours later with handshakes from almost every major name in Japanese industry. The visit, first reported by TechCrunch AI, produced three concrete outcomes worth understanding.

The national AI factory

Japan does not want its factories and robots running on American or Chinese software brains. So Tokyo pulled together roughly 44 domestic companies, with SoftBank, Sony, NEC, and Honda at the core, into a single consortium called Noetra.

Noetra's job is to build Japan's own AI for machines. The government is putting up to 1 trillion yen behind it over five years.

The catch: building AI this large requires enormous amounts of computing power, and Japan currently has to buy that hardware from Nvidia. Nvidia will construct a massive data centre for Noetra, packed with its upcoming "Vera Rubin" chips (its next generation of specialised AI processors), delivering enough power to run the most complex AI training imaginable. Target date: 2028.

Noetra plans to release its AI in stages. A Japanese-language reasoning model first, then one that handles text, images, video, and audio together, then, by 2030, an AI built specifically to run robots.

Ten robotics giants back Nvidia's Cosmos

Nvidia has an open AI model called Cosmos, designed to help robots and machines understand the physical world around them. Think of it as a shared starting point that companies can adapt for their own machines, rather than building everything from scratch.

In Tokyo, Nvidia announced a new version called Cosmos 3 Edge, which runs directly inside the machine itself rather than needing a distant server. Fanuc, Yaskawa, Kawasaki Heavy, Fujitsu, Hitachi, NEC, Sony, SoftBank, Kubota, and the robotics group AIRoA all committed to building on it. Honda R&D and Omron are already testing the tools.

Toyota's cars and factory floors

Toyota, which already uses Nvidia chips widely, is now extending that relationship into its manufacturing simulations (digital rehearsals used to design production lines before a single bolt is tightened) and into software that reads live road traffic. Its next-generation vehicles will use Nvidia's Drive platform for advanced driver assistance, which handles steering and braking but still needs a human in the seat.

What does this mean for ordinary people?

For most of us, nothing changes tomorrow. But Japan is placing a very large bet that AI-controlled robots will fill the gap left by a shrinking workforce. If it works, the products those factories make, cars, electronics, industrial equipment, could get cheaper and more reliable over the next decade. Japan's plan targets over 30 percent of a global AI robotics market Tokyo values at roughly 133 billion dollars by 2040.

The sovereignty angle is real too. Japan wants its own data and its own computing infrastructure rather than depending entirely on foreign-controlled systems. For now, though, that independence still runs on Nvidia silicon.