Noetra: a consortium of four giants

The key player in the entire project is the private consortium Noetra, whose majority owners are four Japanese technology and industrial companies: SoftBank, NEC, Sony Group, and Honda. It was this group that won the NEDO (Japan's government agency for new energy and industrial technology) public tender on June 30 and received the mandate to lead the FRONTia project.

FRONTia is the abbreviated name for "Development of Multimodal Foundation Models with a View to AI Robotics and Physical AI" — a government initiative by Japan's Ministry of Economy, Trade and Industry (METI). The Japanese industrial research institute AIST is the second grant recipient. The project will run from fiscal year 2026 to 2030.

Breathtaking numbers: 27,500 GPUs and 140 megawatts

The core of the entire infrastructure will be an AI factory built on the latest hardware from NVIDIA:

• 13,750 NVIDIA Vera CPUs — processors specifically designed for AI workloads
• 27,500 NVIDIA Rubin GPUs — the new generation of graphics accelerators following the Blackwell architecture
• 140 megawatts of data capacity — comparable to the output of a medium-sized power plant
• Architecture: NVIDIA Vera Rubin NVL72 racks on the NVIDIA DSX platform
• Networking: NVIDIA Spectrum-X Ethernet and NVIDIA BlueField DPU

For comparison: the world-renowned supercomputer of Japan's RIKEN institute, RIKYU, has "only" 1,600 Blackwell GPUs. Noetra is thus planning infrastructure on a completely different order of magnitude.

The funding matches the ambition: in the first year, the project will receive 387.3 billion yen (approximately $2.4 billion). Over the five-year horizon, the total investment could reach up to 1 trillion yen, or roughly $6.2 billion.

Physical AI: what does it actually mean?

The term "physical AI" refers to a category of artificial intelligence that doesn't just work with text or images but directly controls the physical world — robots, machines, vehicles, industrial lines. While ChatGPT or Claude are language models, physical AI must understand space, movement, physical laws, and react to the environment in real time.

That's precisely why Japan is betting on this area. The country is a traditional powerhouse in industrial robotics — companies like FANUC, Yaskawa, and Kawasaki Heavy Industries are world leaders. Now they want to integrate cutting-edge AI into their robots and stop depending on foreign cloud platforms.

Ecosystem: from cars to pharmaceuticals

The project is attracting the elite of Japanese industry. According to the NVIDIA blog, a number of giants have joined the platform:

• Toyota is developing L2++ autonomous vehicles on the NVIDIA DRIVE AGX platform
• Fujitsu, FANUC, Yaskawa, Kawasaki Heavy Industries are deploying NVIDIA technologies in industrial robotics
• SoftBank is building AI-native mobile networks with GB200 infrastructure
• Mizuho Bank is planning the largest on-premise AI factory in Japanese banking on DGX B200 systems
• Pharmaceutical companies Eisai, Astellas, Daiichi Sankyo, and Ono Pharmaceuticals are using NVIDIA BioNeMo for drug development
• Canon and Fujifilm have released CT systems with NVIDIA acceleration

Models emerging from the FRONTia project will be available as open pre-trained weights through NVIDIA Nemotron, Cosmos, Isaac GR00T, and NeMo libraries. This is an important detail: unlike closed models such as GPT-4, these models will be accessible to researchers and companies worldwide.

Strategic goal: 30% of the global AI robotics market

In its AI Robotics Strategy from March of this year, Japan set a clear goal: by 2040, to control more than 30% of the global AI robotics market. The estimated value of this market? $133 billion.

This is an extraordinarily bold ambition. Today, the US and China are the most visible players in the AI battle, with China having the advantage of cheap labor and massive state support. Japan is betting on a unique combination: an industrial robotics tradition, top-tier NVIDIA hardware, and now its own AI models designed specifically for the physical world.

NVIDIA CEO Jensen Huang recently stated at an event in Japan: "Forty years ago, you could have a personal computer. Today, you can have your own personal AI." Words aimed directly at Japanese industry, which — as we can see — has embraced the offer with open arms.

What does this mean for Europe and the Czech Republic?

The Japanese project is a direct inspiration (and challenge) for Europe. The EU is only just building its AI infrastructure under the AI Factories initiative — and the Japanese model shows how quickly things can move with sufficient political will and industrial coordination.

For Czech companies, one key detail matters: NVIDIA Vera Rubin GPUs and the entire software layer (Cosmos, Isaac GR00T) are globally accessible. Czech firms involved in industrial automation, robotic systems, or AI model development can access the same tools — albeit on a significantly smaller scale. Platforms like NVIDIA Isaac GR00T for robot training or NVIDIA Cosmos for digital twins are available via the cloud without needing your own supercomputer.

Moreover, the EU is approving the first wave of AI factories in 2026 under the EuroHPC program — shared computing infrastructure for European research and industry. The Japanese model can serve as a template for how to connect government, industry, and technology companies into a functional whole.

The world is realigning around physical AI

Japan's move is part of a broader realignment of forces in the AI world. The US dominates language models (OpenAI, Anthropic, Google), while China is building its own AI stack separate from the West. Japan is now betting on a third path: specialization in physical AI and robotics, where it has a decades-long industrial advantage.

The Noetra project is currently the largest concentrated national program in the field of physical AI in the world — and NVIDIA has gained a key strategic partner in the Asian market. The results of this partnership will influence what factories, hospitals, and warehouses will look like in the second half of the 21st century.