Japan’s 27,500-chip Nvidia deal challenges China’s grip on robotics
Japan’s government has committed ¥1 trillion ($6.3 billion) over five years to build a state-backed AI infrastructure project centered on 27,500 Nvidia Rubin chips, directly positioning itself to compete with China’s near-total dominance in robotics manufacturing. This represents the first major institutional commitment to “physical AI”, the software layer that enables robots to perceive and act in real-world environments, potentially reshaping the global supply chain for automated manufacturing and logistics.
- Japan will purchase 27,500 Nvidia Rubin chips for the Noetra project, with a 140-megawatt AI factory launching operations in June 2028.
- China shipped approximately 97% of the world’s humanoid robots last year (about 19,000 units) and operates nearly 2 million industrial robots globally.
- A consortium of 44 Japanese companies including Honda, Sony, and NEC will develop physical AI software, with Fujitsu building an operating system specifically for robotic applications.
- 27,500 Nvidia Rubin chips allocated to Japan’s Noetra AI infrastructure project
- ¥1 trillion ($6.3 billion) government funding commitment over five years versus China’s $300 billion robotics subsidies
- 97% share of global humanoid robot shipments manufactured in China last year
Japan’s trade ministry announced a strategic pivot away from competing directly with the United States in general-purpose artificial intelligence or with China in broad robotics manufacturing. Instead, the country is concentrating resources on a narrower but potentially more defensible segment: the software and systems that allow robots to understand and respond to physical environments.
The Noetra consortium, led by SoftBank and backed by the government, will build this infrastructure around Nvidia’s next-generation hardware, with construction beginning in April 2027 and full operations expected by June 2028.
The scale of Japan’s commitment underscores the strategic urgency. The government is investing ¥387.3 billion (approximately $2.4 billion) this fiscal year alone, with the five-year total reaching ¥1 trillion. That investment will establish a 140-megawatt data center on the grounds of a former Sharp factory in Sakai, Osaka Prefecture.
Thirteen research institutions, including Tokyo University of Science, Cambridge, and Oxford, have been formally organized into a national institute to conduct foundational research alongside the corporate partners.
The 44-company consortium includes not only established robotics manufacturers like Fanuc, Yaskawa Electric, and Kawasaki Heavy Industries, but also automotive firms Honda and Mitsubishi Heavy Industries, electronics makers NEC and Sony, and technology companies like Fujitsu.
Nvidia Supplies Foundation Models and Operating System as Japan Targets Robotics Gap
The hardware-software pairing is deliberately structured to lock in Nvidia’s ecosystem while building Japan-specific capabilities on top. Nvidia will provide its Nemotron and Cosmos foundation models plus chip technology optimized for robotic applications.
Fujitsu, Japan’s largest domestic computing company, will develop an operating system specifically designed for physical AI workloads, the first major effort to create a robotics-focused OS layer independent of open-source alternatives.
This division of labor reflects a pragmatic assessment of where Japan can add value. Nvidia CEO Jensen Huang explicitly framed Japan’s post-war manufacturing heritage as the country’s competitive advantage. Speaking at the Fujitsu roundtable, Huang noted that the manufacturing industry’s accumulated “know-how” represents Japan’s enduring asset in an AI-driven economy.
Nvidia is already collaborating with Toyota on manufacturing robotics and has indicated plans to expand partnerships to Honda, Mitsubishi Heavy Industries, and Hitachi. The foundation models and chip technology Nvidia contributes will accelerate Japanese companies’ ability to deploy robots in real factories rather than research labs.
The physical AI approach sidesteps direct confrontation with Chinese mass production and American foundational model dominance, instead focusing on the integration layer where Japanese manufacturing expertise remains world-leading.
China’s 4.5x Industrial Robot Lead Built on EV Supply Chain and $300 Billion Subsidy Program
China’s grip on robotics did not emerge by accident. The country’s dominance in humanoid and industrial robot manufacturing flows directly from its earlier dominance in electric-vehicle supply chains, which created large-scale production expertise and component ecosystems that robotics manufacturers could leverage.
Chinese factories shipped approximately 19,000 humanoid robots in the most recent year tracked, representing roughly 97 percent of global humanoid robot shipments. Industrial robot deployment tells an even starker story: China operates an installed base of nearly 2 million industrial robots, a figure roughly 4.5 times larger than Japan’s installed base.
The Chinese government has committed approximately $300 billion in planned subsidies to robotics and AI development under its 2026-2030 five-year plan. That subsidy level dwarfs Japan’s ¥1 trillion ($6.3 billion) commitment on a nominal basis, though the comparison becomes more nuanced when accounting for purchasing power and the relative size of each economy’s industrial base.
China’s strategy has emphasized breadth: subsidizing robot manufacturers to deploy systems across manufacturing, logistics, and service sectors. Japan’s strategy, by contrast, emphasizes depth in a single critical layer: the software that translates sensor input into physical action.
The fundamental question for Japan is whether leadership in the operating system and foundation models for physical AI can generate sufficient returns on ¥1 trillion investment to justify the commitment against China’s entrenched manufacturing scale.
Labor Shortage and Aging Population Drive Japan’s Automation Imperative
Japan’s robotics strategy is not purely about competing for international market share. The country faces a structural crisis in labor availability. Japan’s population is in documented decline, and the working-age population is shrinking faster than any other major economy.
Nvidia’s Jensen Huang explicitly linked the automation opportunity to Japan’s demographic bind, arguing that deployment of AI, robotics, and advanced automation could reverse economic stagnation driven by labor shortages. For Japan, successful robotics deployment is not a luxury but a near-term economic necessity.
This demographic reality explains why Japan is willing to commit such substantial resources despite China’s head start. Japanese manufacturers cannot simply hire additional workers to offset productivity losses or maintain factory output in the face of labor scarcity. Robotics is not a competitive edge in this context, it is a survival mechanism for the industrial base.
The physical AI layer that Noetra is developing will be most immediately valuable to Japanese manufacturers operating under labor constraints, giving domestic companies a first-mover advantage in deploying these systems within Japan before export competition intensifies.
Honda, Mitsubishi Heavy Industries, and other consortium members can pilot robotic systems in their own factories starting as soon as 2028, potentially generating competitive advantage in automotive, industrial equipment, and precision manufacturing by the early 2030s.
The technology and operational expertise developed during this pilot phase can then be packaged and exported to other aging industrial economies facing similar labor shortages.
April 2027 Construction Start and June 2028 Operations Launch Set Timeline for Physical AI Competition
The Noetra project operates on a defined timeline that will determine whether Japan can execute its strategy before competitive dynamics shift. Construction of the 140-megawatt data center in Sakai begins in April 2027, with full operations expected by June 2028.
This 14-month construction window is aggressive for a facility of that scale and power requirement, suggesting that detailed planning and land acquisition are already substantially complete.
The June 2028 launch represents a critical waypoint. By that date, the foundation models, operating system, and chip infrastructure must function sufficiently well that early consortium members can begin deploying robotic systems in production environments.
The 44-company structure means that success or failure will be immediately visible: if Fanuc, Yaskawa, and Kawasaki can demonstrate working robotic systems powered by Noetra infrastructure within 12-18 months of launch, Japan will have proven that the physical AI approach is viable. If deployment stalls, the entire rationale for the ¥1 trillion commitment comes into question.
The concrete test will arrive by late 2029 or early 2030, when the consortium companies
