Japan is making a bold move into the future of artificial intelligence, distinct from the conversational models dominating the global stage. Major Japanese corporations, including SoftBank, NEC, Honda, and Sony Group, have formed a new company focused on developing a trillion-parameter AI model. This ambitious project is not aimed at creating the next chatbot, but rather at advancing "Physical AI." This next frontier of AI is designed to power real-world applications such as robotics, autonomous vehicles, and industrial machinery.
The initiative leverages Japan's strong industrial base and extensive robotics heritage. SoftBank and NEC are expected to spearhead the core AI development, while Honda will integrate the technology into autonomous driving systems. Sony is contributing its expertise in robotics and gaming hardware. Preferred Networks, a recognized AI developer, is also a key partner. The new entity, effectively named "Japan AI Foundation Model Development," plans to recruit approximately 100 AI engineers.
This venture has garnered significant support from Japan's financial and industrial sectors, with major banks like MUFG Bank, Sumitomo Mitsui Banking, and Mizuho Bank, along with steelmakers Nippon Steel and Kobe Steel, joining as investors. The Japanese government, through its NEDO agency, has committed approximately ¥1 trillion (around $6.7 billion) in AI support over five years, starting fiscal year 2026. A critical aspect of this national strategy is to keep AI development and data domestic, avoiding reliance on foreign cloud infrastructure.
This domestic focus stands in contrast to SoftBank's global investments, including a significant role in OpenAI's funding. The "Japan AI Foundation Model Development" aims to cultivate an AI ecosystem independent of the large American tech firms, ensuring Japanese data remains within the country and does not feed foreign AI pipelines. This strategic pivot aims to reclaim technological sovereignty and capitalize on Japan's inherent strengths in physical industries. Practical applications for this advanced Physical AI are anticipated by 2030.