Japan to buy 27,500 Nvidia Rubin chips for homegrown robot AI

Japan's Noetra initiative represents a coordinated national effort to consolidate domestic AI capabilities around a shared robotics-focused platform. Rather than allowing companies like SoftBank, Preferred Networks, and NEC to develop competing large language models in isolation, the government is channeling ¥387.3 billion ($2.4 billion（約3800億円）) into a unified venture that will leverage 27,500 Nvidia Rubin chips—a substantial order, though the body notes it is small compared with Microsoft's plans for hundreds of thousands of Vera Rubin systems. This consolidation serves two strategic purposes: lowering Japan's reliance on foreign technology infrastructure and positioning the nation to capture more than 30% of the estimated ¥60 trillion global robotics market by 2040.

The timing reflects both economic and demographic pressures. Japan faces a shortage of skilled craftspeople—welders, machinists, and other industrial workers—retiring without knowledge transfer mechanisms in place. The government and Nvidia co-founder Jensen Huang both frame AI as a tool to preserve and transmit these specialized skills to the next generation. Noetra's roadmap supports this vision: an initial model release by March of next year, followed by regular updates and a robotics-specific variant within a few years. The June 2028 data center launch provides the physical infrastructure foundation. By coordinating these efforts under a single entity led by Hironobu Tamba (who previously led SoftBank's language model development), Noetra aims to create an alternative to U.S. and Chinese AI systems rather than fragmenting investment across competing national champions.