Jensen Huang Just Spent Two Days In Tokyo And Walked Away With A Huge Chunk Of Japan's AI Future

Taiwan a few weeks back, South Korea months before that, and now Tokyo, where he spent July 15 and 16 courting basically every major name in Japanese industry and chip supply. He didn't leave empty-handed. He left with deals spanning a national AI factory, partnerships with Japan's biggest robotics companies, and agreements with chip-material suppliers feeding directly into Nvidia's next-generation chips.

The underlying message from Huang was pretty explicit: AI's next chapter isn't happening in chatbots or cloud software; it's happening on factory floors, in robots, in physical machines. And he wants Japan to build that future using Nvidia's hardware underneath it all.

There's a nice historical wrinkle here too. Thirty years ago, a $5 million investment from Sega helped keep a near-bankrupt Nvidia afloat. Now the relationship's flipped: Nvidia and Japan's industrial giants need each other again, this time to build what's being called the "physical-AI era." Three specific projects are anchoring that push.

Noetra: Japan's bet on owning its own AI brain

Japan doesn't want to run its factories and robots on American or Chinese AI models. That's the whole premise behind Noetra, a coalition of roughly 44 domestic companies, anchored by SoftBank, Sony, NEC, and Honda, built specifically to develop homegrown "physical AI": foundation models designed to actually run machines, not just generate text.

Tokyo's putting real money behind this up to 1 trillion yen, or about $6.2 billion, over five years. The goal is clear: Japan wants to own the software brain driving its industrial future. But here's the catch that runs through this entire story: the hardware powering that ambition still comes from Nvidia.

Specifically, Nvidia is building what it's calling a "Vera Rubin AI factory" a massive data center packed with its next-generation chips, expected to launch in 2028. We're talking 13,750 Vera CPUs and 27,500 Rubin GPUs, delivering 140 megawatts of compute. Noetra will oversee the whole effort and handle building the actual data center.

Noetra's roadmap runs in three distinct stages. First, a reasoning model heavy on Japanese-language capability, starting in fiscal 2026. Then an omni-modal version handling text, images, video, and audio by 2028. Finally, what's being called "Real-world Native AI" built specifically to run robots targeted for 2030, with plans to release it to outside Noetra developers in phases as it matures.

The robotics coalition: Japan's manufacturing giants all lining up behind one model

This is where the sheer scale of Huang's Tokyo trip becomes clear. Fanuc, Yaskawa, Kawasaki Heavy, Fujitsu, Hitachi, NEC, Sony, SoftBank, Kubota, and robotics group AIRoA have all signed on to build using Nvidia's Cosmos models an open-model initiative Nvidia launched back in May with a handful of global AI labs.

To sweeten that commitment, Nvidia unveiled Cosmos 3 Edge while in Tokyo: a version of the model specifically built to run on its Jetson Thor chips embedded directly inside machines, rather than needing to phone back to a data center for every decision. Some companies in the coalition are already testing a shared control system built around this. Others, including Honda R&D and Omron, are actively building on the tools right now.

Huang framed the opportunity in genuinely ambitious terms: "The next frontier of AI is in the physical world, and this is a once-in-a-generation opportunity for Japan. Japan invented modern manufacturing. Now, it has the opportunity to reinvent it for the age of intelligent industries."

Toyota: extending Nvidia deep into how cars actually get built and driven

Toyota's relationship with Nvidia already runs deep; the company committed its next-generation vehicles to Nvidia's Drive platform back at CES in January 2025. What's new from this Tokyo visit is how much further that relationship now extends: into Toyota's actual manufacturing process, where simulations are used to design production lines, into the software running inside its vehicles, and into systems that read and interpret road traffic in real time.

Worth noting is that Toyota's taking a notably more conservative approach here than some competitors. Its vehicles will run advanced driver assistance steering and braking support that still requires an actual human driver behind the wheel rather than pursuing the kind of reduced-human-dependence systems Waymo and Tesla are developing. That's a real strategic distinction, and it says something about how Toyota's thinking about risk and rollout pace compared to companies racing toward fuller autonomy.

Japan is betting this big

Behind all three deals sits a genuinely urgent demographic problem: Japan's workforce is shrinking, and the country needs a real solution. The target is 10 million AI-equipped robots deployed across 18 different sectors by 2040, backed by $65 billion in combined public and private investment specifically in physical AI.

The ambition goes well beyond just solving Japan's own labor shortage, though. Japan's AI Robotics Strategy, released back in March, aims to capture more than 30% of the global AI robotics market by 2040, a market Tokyo values at roughly ¥20 trillion, or about $133 billion. Japan's trade ministry (METI) is funding the domestic foundation model meant to run all these machines, and Noetra's Nvidia-powered factory is specifically where models at that scale reaching into the trillions of parameters would actually get trained.

The sovereignty angle underneath the industrial one

There's a deeper strategic logic running beneath all of this that's worth pulling out explicitly. As the US and China continue pulling ahead in large-scale AI development, Tokyo wants its own data, its own compute infrastructure, and meaningfully less dependence on infrastructure it doesn't directly control.

The political weight behind this was on full display during Huang's visit; he appeared on July 16 alongside Japan's trade minister Ryosei Akazawa at the government's official physical-AI launch event, with Prime Minister Sanae Takaichi joining by video. This isn't just corporate deal-making; it's genuinely tied to national strategy. The Takaichi administration has made AI and semiconductors the centerpiece of a growth plan chasing ¥370 trillion, roughly $2.3 trillion, in combined public and private investment by 2040.

Nvidia itself is billing Noetra's planned factory as "the world's first national AI infrastructure," which is a striking claim on its own. But there's a real irony sitting at the center of Japan's sovereignty push: for now, that push toward independence still rests entirely on American chips. Japan wants to own the software brain running its industrial future, but the hardware body underneath it is, at least for the foreseeable future, still Nvidia's.

The playbook, repeated across three countries

What's genuinely striking is just how consistent Huang's approach has been across this whole tour. In two days in Tokyo, he sat down with nearly every name that matters in Japanese tech CEOs from Toyota, Fanuc, Yaskawa, Fujitsu, and Kawasaki over lunch, and dozens of supply-chain executives over skewers and whisky at a Kanda izakaya.

It's the same basic playbook he ran weeks earlier in Taiwan (a homecoming keynote), and the one before that in Seoul last fall (fried chicken and a 50,000-GPU deal). This time it was Tokyo's turn robots, supply chain, and the chips underneath all of it, wrapped into deals that position Nvidia as the essential infrastructure layer beneath multiple countries' national AI ambitions simultaneously.

What emerges from this trip is a pretty clear pattern worth watching closely: Nvidia isn't just selling chips anymore it's embedding itself as the foundational infrastructure layer underneath entire countries' industrial and AI strategies, one high-profile regional tour at a time. Japan wants sovereignty over its AI future badly enough to commit trillions of yen toward it. But for now, and for the foreseeable next several years, that sovereignty is being built on top of Nvidia's hardware, with Huang himself sitting at the table for nearly every major deal that makes it happen.