Japan will buy 27,500 Nvidia Rubin chips to build a sovereign foundational AI model for robots. The order, reported by Bloomberg, is one of the largest national AI chip commitments tied to robotics.
Key facts
- 27,500 Nvidia Rubin chips ordered by Japan.
- Chips will power a foundational robot AI model.
- Rubin is Nvidia's next-gen architecture after Blackwell.
- Financial terms of the deal not disclosed.
- Japan aims for sovereign AI capability in robotics.
Japan is planning to buy 27,500 next-generation Rubin chips from Nvidia Corp. to build a homegrown foundational AI model for robots, Bloomberg reported. The deal, whose financial terms were not disclosed, signals Tokyo's strategic push to reduce reliance on foreign AI systems in a sector where Japan holds industrial strength.
The Rubin architecture is Nvidia's successor to Blackwell, with the Vera Rubin NVL72 cloud rollout expanding to Europe in H2 2026 [per Nvidia's prior announcements]. Japan's order comes amid a broader trend of nations building sovereign AI infrastructure — the U.S. recently allowed ZTE to buy H200 chips, and Nvidia cut its Asia partner list by half in July to curb smuggling [as previously reported by gentic.news].
Japan's focus on robotics is distinctive. Unlike generic sovereign AI projects (e.g., France's Mistral or China's ERNIE), Japan is targeting physical AI — models that control robots in factories, logistics, and eldercare. This aligns with Nvidia's recent open-source robot model release with Hugging Face on July 9 [per prior reporting]. The 27,500-chip order would likely require multiple data centers and significant liquid cooling infrastructure, given Rubin's power density.
The strategic implication: Japan is betting that the next AI frontier is embodied intelligence, not just language models. If successful, it could create a template for other manufacturing-heavy economies (Germany, South Korea) to follow. The risk is that Nvidia's Rubin timeline faces delays — the company's next-gen rack system was already pushed to 2028 due to manufacturing snags [per a January 2028 report].
What to watch
Watch for Nvidia's Q3 2026 earnings call — if Japan's order is confirmed in the data-center segment, it could signal a new sovereign-AI procurement pattern. Also track whether Japan's robot model uses Nvidia's NeMoClaw framework or a custom architecture, which would reveal the depth of Nvidia's lock-in.
Source: bloomberg.com
[Updated 16 Jul via tomshardware]
The deal is being executed through the Noetra consortium, a joint venture between Nvidia and Japan's Noetra Corp., which will construct a 140-megawatt AI factory housing the 27,500 Rubin GPUs alongside 13,750 Vera CPUs [per Tom's Hardware]. This marks the first publicly named national AI infrastructure project of its scale, with the power capacity alone rivaling mid-sized data center campuses.
