Japan Commits $2.4 Billion to Buy 27,500 Nvidia Rubin Chips for Homegrown Robot AI

Japan is putting real money behind a bet that it can build its own robot AI instead of importing one from the U.S. or China.

The government has allocated ¥387.3 billion, about $2.4 billion, to a newly formed company called Noetra Corp. through March 2027, according to Bloomberg reporting carried by the Japan Times. Noetra will use that money to buy 27,500 of Nvidia's next-generation Rubin chips and build a roughly 140-megawatt data center. Stocktwits reported the buildout also includes 13,750 Vera CPUs, forming what Nvidia calls an AI factory.

The data center is slated to go online in June 2028, per the Japan Times. That's a two-year runway before Japan has its own large-scale compute for this project up and running.

Who's Actually Building This

This isn't a government project bolted onto one company. Sony Group, SoftBank, NEC, Fujitsu, and Toyota-backed Preferred Networks are all involved in setting up and operating Noetra, according to ZeroHedge and the Japan Times. Each of these firms already built its own AI model: SoftBank's Sarashina, Preferred Networks' PLaMo, NEC's cotomi, according to a Prismmarketview report. Noetra is meant to pool that scattered effort into one national push instead of five companies duplicating work.

Noetra President Hironobu Tamba previously led development of SoftBank's large language model, per Prismmarketview. Tamba described the goal plainly: build "a third option" beyond U.S. and Chinese AI platforms that Japan and other countries could choose to use, according to Prismmarketview's paraphrase of his comments.

Stocktwits quoted Tamba separately making the practical case for why this needs to be a joint national effort rather than a single company going it alone: "Bringing physical AI into the real world requires enormous computing, data and foundational technologies, challenges no single company can solve alone."

The Numbers, In Context

Noetra plans to release its first general AI model by March 2027, then follow with a version tailored specifically for industrial robots, according to both ZeroHedge and Prismmarketview. The strategic target sitting behind all of this: Japan wants to capture more than 30% of the global robotics market, which Prismmarketview pegs at roughly ¥60 trillion (ZeroHedge cites the dollar figure as $370 billion) by 2040.

It's an ambitious target for a country whose population is shrinking and whose labor force needs exactly the kind of automation humanoid robots promise. Japan already builds a huge share of the world's industrial robots. Turning that hardware advantage into an AI-software advantage is the actual strategic logic here, not just chasing a trend.

Noetra's 27,500-chip order, while large in absolute terms, is modest compared with what Microsoft alone plans to deploy: hundreds of thousands of Nvidia's Vera Rubin systems across its own data centers. Prismmarketview and the Japan Times both note this reality. Japan's national project is a fraction of one hyperscaler's ambitions.

Nvidia's Read on This

Nvidia CEO Jensen Huang, quoted by Bloomberg via ZeroHedge, called the Noetra project "just the beginning" for Japan. "We're going to be building a lot more infrastructure here," Huang said. Nvidia is also expanding partnerships with Mitsubishi Heavy Industries and Toyota separately from the Noetra deal, according to Stocktwits, and Mitsubishi Motors has separately signed on for humanoid robot production, per ZeroHedge.

The strongest skeptical case here isn't that Japan is wrong to want AI independence. It's a fair concern that a government-funded, multi-company consortium risks the same coordination problems that plagued past Japanese industrial-policy pushes, from semiconductors in the 1990s to earlier AI initiatives that struggled to keep pace with faster-moving American and Chinese rivals. Pooling five competing corporate AI teams under one roof does not guarantee they'll out-innovate Nvidia's own customers or OpenAI's compute-rich backers. The 2028 data center timeline also means Japan is betting on a two-year window where Rubin-generation chips remain competitive rather than superseded by whatever Nvidia ships next.

Nvidia stock traded slightly lower, down about 1%, in Thursday's premarket session on July 16, according to Stocktwits, even as the outlet noted the stock was on pace for a third straight weekly gain. That's a minor blip against a broader run driven by AI infrastructure demand across multiple countries, not just Japan.

Noetra has a budget, a chip order, and a roster of corporate partners. What it doesn't have yet, as of this writing, is a working data center, a shipped model, or a track record. Japan's METI-backed FRONTia initiative, which Stocktwits reports is the policy umbrella for this effort, will be judged on whether Noetra actually ships that first AI model by its stated March 2027 target, and whether the robotics-specific model that follows can compete with what American and Chinese labs will have built by then.