Japan seeks to procure 2.75 million NVIDIA Rubin chips, building its own domestic robot “sovereign AI”

Nvidia CEO Jensen Huang personally traveled to Tokyo this week, speaking about how Nvidia can help Japan realize its dream of “sovereign AI” for the yen. According to a report by Bloomberg, a research organization estimates that Japan will procure about 27.5k units of Nvidia’s next-generation Rubin chips to build domestic robot and physical AI foundation models. This is a national-level high-stakes gamble for Japan to break away from dependence on overseas AI technology—Japan wants to take control of the compute lifeline back into its own hands. According to Bloomberg, compiled by 動區動趨.
(Background recap: Samsung and SK hynix’s $1.3 trillion supported by the South Korean government! Analysts: AI is a life-or-death battle for countries)
(Background add-on: How did Sega save Nvidia back then? That’s what made Jensen Huang grateful for a full 30 years)

Table of contents

Toggle

• Jensen Huang’s Tokyo trip is not just about selling chips
• Japan is rushing to grow its own AI
• Sovereign AI becomes an arms race among countries

Key takeaways

• According to Bloomberg, a research organization estimates Japan will purchase about 27.5k Nvidia Rubin chips to build domestic robot AI models.
• Jensen Huang personally went to Tokyo to discuss helping Japan with “sovereign AI,” saying Vera Rubin has entered production and massive production capacity is about to arrive.
• Japan’s five-year domestic AI initiative led by the Noetra alliance; METI’s top investment is 1 trillion yen, targeting 10 million AI robots by 2040.

Nvidia CEO Jensen Huang is in Tokyo this week. This trip isn’t to sell graphics cards—it’s to talk about how Japan can rely on Nvidia’s chips to build its own AI. According to Bloomberg, a research organization estimates that Japan will procure about 27.5k Nvidia’s next-generation Rubin chips to train domestic robot and physical AI (physical AI) foundation models.

This number isn’t small when placed into a single country’s orders to Nvidia. Behind it lies one of Japan’s most aggressive technology-strategy buzzwords in recent years: sovereign AI. The idea is to no longer hand over the compute lifeline of AI to U.S. cloud giants, but to use its own chips, data, and talent to grow an AI that understands Japan’s industries—and keeps Japan’s data as well.

Jensen Huang’s Tokyo trip is not just about selling chips

This time, Jensen Huang’s Tokyo trip centers on how Nvidia can help Japan complete the puzzle of sovereign AI and physical AI. Faced with external doubts about Rubin’s mass-production progress, he directly stated that Vera Rubin is already being produced, and massive production capacity is about to arrive—essentially pre-packaging a promise of satisfaction for big customers like Japan.

Rubin is Nvidia’s next-generation AI platform after Blackwell. Using Rubin GPUs paired with Vera CPUs, it significantly boosts computing performance and also includes a security architecture designed specifically for robots. For a country that needs to feed domestic large models while supporting millions of robots at the same time, this is exactly what it wants.

Japan is rushing to grow its own AI

The answer is hidden in demographics. Japan’s labor force continues to shrink, and factories and the service industry increasingly rely on robots and automation to make up for labor shortages. Old-guard robot makers like Fanuc and Yaskawa are precisely key players.

The Japanese government has already written this into its national plan. The Ministry of Economy, Trade and Industry (METI) has designated a five-year domestic AI development project led by the Noetra alliance (members include SoftBank, Sony, NEC, and Honda). Officially, it will invest up to 1 trillion yen, about $6.2 billion. The foundation model is expected to roll out as early as the current fiscal year, focusing on physical AI built from sensor data, robot operation records, and training in simulated environments. Japan has even set a goal for 2040: to put 10 million AI robots to work.

Sovereign AI becomes an arms race among countries

Japan isn’t the only one thinking this way. From Saudi Arabia and France to South Korea, countries are pouring tens of billions of dollars into building AI infrastructure, worried they might fall behind in this round of compute arms competition. For Nvidia, this is a good business: for any country that wants sovereign AI, in the end it almost always has to place orders with it first to buy chips.

Japan buying 27.5k Rubin chips is only a snapshot of this global contest for computing power. The real question is whether it truly counts as “sovereignty” when compute capability is treated as part of national power and the lifeline is staked on a single American chip supplier—Japan will likely have to answer that for itself next.

Common questions

What is “Sovereign AI” (Sovereign AI)?

It refers to a country using its own infrastructure, data, and talent to build AI capabilities—keeping the compute lifeline in its own hands. Japan, Saudi Arabia, and France have all投入数十亿美元 to build their own, aiming to reduce dependence on overseas AI technologies and cloud services.

What are Nvidia Rubin chips?

Rubin is Nvidia’s next-generation AI platform after Blackwell. It combines Rubin GPU with Vera CPU to greatly improve computing performance and includes a robot security architecture. Jensen Huang says it has entered mass production, and massive production capacity is about to arrive.