Japan is making its biggest coordinated bet on artificial intelligence — and it’s not building another chatbot.
SoftBank Group, Sony, NEC, and Honda have launched a joint venture to build a trillion-parameter physical AI model capable of autonomously controlling robots, vehicles, and industrial systems by 2030. The project comes with roughly $6.28 billion in Japanese government funding and backing from some of the country’s largest industrial firms.
The venture, whose name translates to “Japan AI Foundation Model Development,” represents the most significant sovereign AI initiative outside the United States and China — and it’s aimed squarely at the physical world, not the digital one.
What Physical AI Actually Means
Physical AI refers to AI systems that operate in the real world — controlling robots that walk, cars that drive, and factories that run with minimal human oversight. Unlike language models that generate text or images, physical AI must understand physics, navigate uncertainty, and make decisions in real time where mistakes have physical consequences.
It’s the difference between telling you about a robot and being the robot.
The joint venture’s goal is a trillion-parameter foundation model optimized for physical control. SoftBank and NEC will lead the AI development. Honda will be the first to deploy the model in its autonomous vehicles. Sony’s role centers on robotics and gaming hardware. Preferred Networks, a Tokyo-based AI startup focused on deep learning and IoT applications, is also participating.
The Sovereign AI Angle
This project has a dimension that goes beyond technology. For years, Japanese companies have been dependent on U.S. cloud infrastructure providers, paying billions for access to AI capabilities and creating what Japanese policymakers call a “digital deficit” — capital flowing out of the country to American tech companies.
The new joint venture intends to train its AI on Japanese data and keep that data in Japan, out of the hands of companies like Google and OpenAI.
That’s a striking pivot for SoftBank, which has been funneling billions into U.S. AI firms. The company led OpenAI’s $40 billion funding round in 2025. Now it’s anchoring a domestic venture designed to be independent of the U.S. AI ecosystem entirely.
The sovereign AI logic is straightforward: if AI becomes as fundamental to economic and military power as semiconductors or energy, depending entirely on foreign providers for the most critical technology is a strategic vulnerability. Japan isn’t the only country reaching this conclusion — France, India, and the UAE have all made similar moves. But the scale and industrial coordination here is unprecedented.
The Industrial Coalition
The venture extends well beyond tech companies. Kobe Steel, Nippon Steel, Mizuho Bank, Sumitomo Mitsui Banking, and MUFG Bank are all listed as investors. The involvement of steel companies and major banks signals that this isn’t a research project — it’s an industrial mobilization.
Japan’s government is contributing through the New Energy and Industrial Technology Development Organization, which has earmarked roughly one trillion yen over the next five years. The venture plans to hire about 100 AI engineers, with a senior SoftBank executive serving as president.
The breadth of the coalition matters. Physical AI requires integration with physical systems — factories, vehicles, construction sites, supply chains. Having steel companies and automakers at the table from the start means the model will be built with deployment in mind, not just benchmarks.
Why 2030 Matters
The companies are targeting physical AI applications by 2030. That timeline aligns with a broader global consensus that the late 2020s will be the period when embodied AI moves from controlled environments to real-world deployment at scale.
Tesla is targeting one million Optimus humanoid robots annually. China’s AGIBOT has opened the world’s largest real-world training ground for embodied AI. Figure AI and Agility Robotics are expanding humanoid deployments across U.S. manufacturing. The race is already underway.
Japan’s entry is distinctive because of its industrial base. The country has world-class manufacturing, automotive, and robotics sectors. If physical AI is fundamentally about making machines that work reliably in the real world, Japan’s engineering culture and manufacturing infrastructure could be a decisive advantage — assuming the AI itself is good enough.
The Competitive Landscape
The physical AI race is shaping up as a three-way contest between the U.S., China, and now Japan.
The U.S. has the leading AI models and the most aggressive deployment timeline. China has the manufacturing scale, government coordination, and national standards infrastructure to push humanoids from demos into factories. Japan has the industrial precision, robotics heritage, and now the coordinated investment to compete.
What makes this different from the LLM race is that physical AI success depends on things that money alone can’t buy: manufacturing expertise, safety culture, regulatory frameworks that enable rather than block deployment, and physical infrastructure that can absorb autonomous systems.
Japan’s bet is that its strengths in those areas can compensate for whatever gap it faces in raw AI capability. If it’s right, the physical AI era won’t be dominated by whichever company builds the smartest model. It’ll be dominated by whichever country builds the most reliable machines.
Sources
- SiliconANGLE — “Japanese tech giants launch joint venture targeting physical AI for robots and machines”
- Nikkei Asia — SoftBank, Sony, NEC, Honda joint venture report
