Japan Launches Domestic Physical AI 'FRONTia' with NVIDIA's Latest GPUs, ¥1 Trillion Investment Over 5 Years

Japan's Ministry of Economy, Trade and Industry officially launched the "FRONTia Project" on July 16, a national project to develop a domestic multimodal foundation model for "physical AI"—AI that operates in the real world, such as in robots and manufacturing sites. The development will be primarily handled by Noetra, an AI development company funded by a total of 44 firms with SoftBank Corp (9434.T), Sony Group (6758.T), NEC (6701.T), and Honda Motor (7267.T) as core investors, along with the National Institute of Advanced Industrial Science and Technology (AIST). U.S.-based NVIDIA will provide a large-scale computing infrastructure, including approximately 27,500 units of its next-generation "Rubin" GPU. This public-private partnership aims to achieve "real-world native AI" by fiscal year 2030. The Japanese government expects to provide support totaling approximately ¥1 trillion (approximately $6.2 billion) over the five years starting from fiscal year 2026.

The project's kickoff event featured appearances by NVIDIA CEO Jensen Huang, Minister of Economy, Trade and Industry Ryosei Akazawa, Prime Minister Sanae Takaichi (via video message), University of Tokyo Professor Yutaka Matsuo, and Noetra President Hirotoki Tamba. Minister Akazawa appeared on stage wearing a black leather jacket, the trademark attire of CEO Huang, and expressed his determination to "win with technology and win in business." Prime Minister Takaichi positioned the project by stating, "We envision ¥80 trillion (approximately $492.5 billion) in public and private investment, including semiconductors, by 2040, and the FRONTia Project is the first step."

The Inevitability of "Japan AI"

In his keynote speech, CEO Huang emphasized that "Japan must build 'Japan AI.'" He stated, "The concepts that 'Made in Japan' signifies—the highest quality, the highest precision, craftsmanship, kaizen, kanban, and genba—are Japan's industrial knowledge itself. We must not lose the knowledge of the people who operate them. A nation cannot outsource its intelligence," arguing for the necessity of Japan owning, improving, protecting, and deploying its own AI models.

Behind this "Japan AI" concept is the global trend of "sovereign AI," where nations protect their own data and industrial base. NVIDIA is already participating in national AI projects in over 20 countries, including the United Kingdom, Germany, France, India, and the UAE, with sovereign AI-related revenue for fiscal year 2026 reaching approximately $30 billion (approximately ¥4.9 trillion), more than triple the previous year. At an event the previous day on July 15, CEO Huang also stated, "Every country and company should use the best possible AI, but a nation's AI must be nurtured, strengthened, and developed domestically. It cannot be exported or outsourced."

A Three-Stage Development Roadmap

The FRONTia Project's development plan consists of three stages. In fiscal year 2026, work will begin on building a language foundation model with advanced Japanese comprehension and logical reasoning. By fiscal year 2028, this will be expanded into a multimodal foundation model that integrates images, video, and audio. The final stage, in fiscal year 2030, aims to realize "real-world native AI" that understands spatial recognition and physical laws and can operate autonomously in the real world.

The core computing infrastructure for development will be built as the world's first national-scale "AI factory," adopting NVIDIA's next-generation "Vera Rubin" architecture. It will be equipped with approximately 27,500 Rubin GPUs and over 13,750 Vera CPUs, with a power capacity of 140 megawatts—a scale capable of training large models with one trillion parameters. Construction is scheduled to begin in April 2027, with operations expected to start in June 2028. Until then, development will proceed using existing computing infrastructure held by domestic operators.

Japan's "Genba Data" as a Source of Competitiveness

Minister of Economy, Trade and Industry Akazawa stated, "Japan's greatest path to victory with big data and AI lies in being a super-aging society and a disaster-prone nation. That is where the big data is. What looks like a crisis is actually an opportunity," expressing the view that Japan's social challenges will become strengths in physical AI development. He specifically cited the world's best manufacturing field data, healthcare data for the elderly, disaster response knowledge, and extreme-environment work data accumulated from the decommissioning of the Fukushima Daiichi Nuclear Power Plant, asserting that "the key is genba (field) data."

Prime Minister Takaichi also positioned physical AI in her video message as "a technological field where the high-quality data accumulated in Japan's industrial sites will become a source of competitiveness and a winning path for Japan," stating the intention to "develop safe, secure, and trustworthy domestically-founded AI and launch it onto the world stage."

University of Tokyo Professor Yutaka Matsuo pointed out that while "the development of foundation models is a battle we cannot afford to lose," he noted that "physical AI is not something that can be used immediately like large language models (LLMs). That is precisely why there is meaning in doing this at the national level now," and called for the participation of more companies and talent.

An "All-Japan" Framework Uniting 44 Companies

Hirotoki Tamba, who has led large-scale language model development under SoftBank Corp, has assumed the role of President and Representative Director of Noetra. In addition to the four core companies, the investing firms span a wide range of industries from manufacturing to finance and pharmaceuticals, including Hitachi, Nippon Steel (5401.T), JFE Steel, Kawasaki Heavy Industries, Mitsubishi Electric, Fanuc, Yaskawa Electric (6506.T), DMG Mori Seiki, Omron, Fujitsu, Toshiba, Tokyo Electron, KDDI Research, Rakuten Group, Mizuho Bank, Sumitomo Mitsui Banking Corporation, MUFG Bank, Nippon Life Insurance, and Daiichi Sankyo. Preferred Networks (PFN), which has a proven track record in AI development, will also participate through both investment and the dispatch of engineers.

The trained weights of the developed models are planned to be sequentially released to domestic companies and developers according to the progress of research, development, and social implementation, and are expected to be provided alongside NVIDIA's open model portfolio.

NVIDIA's Comprehensive Commitment

Beyond providing GPUs, NVIDIA has also announced the provision of its "NemoTron" and "Cosmos" foundation models for AI agents and robotics, as well as semiconductor technology for embedding in robots. Prior to the July 16 announcement, CEO Huang also revealed strengthened collaboration with Fanuc, Yaskawa Electric, and Kawasaki Heavy Industries, and it was also disclosed that Fujitsu plans to supply an operating system for executing physical AI.

Furthermore, NVIDIA plans to collaborate with Toyota Motor, the world's largest automaker, on developing robots for automobile manufacturing, and to strengthen cooperation on AI and robotics technology development with Mitsubishi Heavy Industries, Hitachi, and others. CEO Huang stated, "Collaboration with Japan's manufacturing sites is essential to develop robotics technology that can be freely utilized at diverse, small-to-medium-sized manufacturing sites around the world," expressing strong expectations for the fusion of Japan's "monozukuri" (manufacturing) culture and physical AI.

Massive Public-Private Investment

The government subsidy for the FRONTia Project's first year, fiscal year 2026, is approximately ¥387.3 billion (approximately $2.4 billion). Adopted as a project under the public solicitation of NEDO (New Energy and Industrial Technology Development Organization), support totaling approximately ¥1 trillion (approximately $6.2 billion) is expected over five years. The "Japan Growth Strategy" championed by Prime Minister Takaichi envisions ¥80 trillion (approximately $492.5 billion) in public and private investment, including semiconductors, by fiscal year 2040, with FRONTia positioned as a preceding project.

Securing data sovereignty is also a key objective. By establishing a system to safely process and utilize the highly confidential physical data held by Japan's manufacturing sites domestically, the project aims to break free from dependence on overseas clouds and accelerate the social implementation of autonomous robots to address labor shortages.