Japan orders 27,500 Nvidia Rubin chips for robotics AI systems

Japan's government is placing one of the largest known sovereign orders for Nvidia's next-generation AI chips.

Japan will purchase 27,500 of Nvidia Corp.'s Rubin chips to build a national robotics AI system, marking one of the biggest government AI hardware procurements to date.

"This order reflects Japan's commitment to building sovereign AI infrastructure for robotics," Jensen Huang, Nvidia's chief executive officer, said at a developer event in Tokyo. Huang also dismissed reports of manufacturing delays for the Vera Rubin platform, calling them "not true" and vowing to deliver "giant amounts" of the systems.

The Rubin architecture, built on TSMC's advanced nodes, succeeds Nvidia's Blackwell line as the company's flagship AI accelerator. The Blackwell B200, Nvidia's current-generation chip, delivers up to 20 petaFLOPS of FP4 performance and 4.5 terabytes per second of memory bandwidth, according to Nvidia's published specifications. The Vera Rubin platform — named after the American astronomer who discovered dark matter — is expected to deliver significant performance gains over the Blackwell Ultra, though Nvidia has not disclosed full specifications. Huang's comments in Tokyo were the first public confirmation that Rubin production remains on schedule.

Japan's robotics push is a strategic priority for a nation facing a shrinking workforce. The country's population aged 65 and older now accounts for about 29% of the total, creating urgent demand for automation in manufacturing, logistics and elder care. The 27,500 Rubin chips will power AI models that enable robots to perceive, plan and act in real time — capabilities that require the massive parallel processing that graphics processing units provide.

The Japanese procurement highlights a broader trend of sovereign nations building domestic AI infrastructure rather than relying solely on US-based cloud providers. For Nvidia, which generated $47.5 billion in data center revenue in fiscal 2025, government orders are becoming a meaningful demand driver alongside hyperscaler customers such as Microsoft Corp., Amazon.com Inc. and Alphabet Inc.

The order comes as Nvidia faces intensifying competition from Advanced Micro Devices Inc., whose MI300X and upcoming MI400 accelerators target the same AI workloads. AMD has secured government AI contracts in Europe and the US, though Nvidia's CUDA software ecosystem — the industry-standard platform for GPU computing — remains a significant competitive advantage. Japan's selection of Rubin over AMD's Instinct line reinforces Nvidia's dominance in sovereign AI deployments.

On the supply chain side, Nvidia relies on Taiwan Semiconductor Manufacturing Co. for chip production and on advanced packaging providers such as ASE Technology Holding Co. for CoWoS (chip-on-wafer-on-substrate) packaging, a critical step in assembling high-bandwidth memory with the GPU die. Huang's confidence that Rubin is on track suggests no disruption to TSMC's capacity allocation for Nvidia's next-generation products.

Nvidia shares trade at about 35 times forward earnings, reflecting investor expectations that AI infrastructure spending will continue to grow. Japan's 27,500-chip order, while modest relative to Nvidia's total annual output of millions of AI accelerators, signals that sovereign demand is becoming a structural revenue contributor. Morgan Stanley analyst Joseph Moore has maintained an overweight rating on Nvidia, citing sustained demand across hyperscale and enterprise customers.

This article is for informational purposes only and does not constitute investment advice.