Japan is making its biggest formal bet on what it’s calling “physical AI”, embedded intelligence in hardware, robotics, and manufacturing systems where Japan’s existing industrial base gives it a competitive starting position. The Growth Strategy Council announced its targets on April 19, according to two independent publications: Yomiuri Shimbun and Sedaily, a South Korean financial daily.
The headline figure is ¥40 trillion in annual semiconductor sales by 2040, equivalent to approximately $258 billion at current exchange rates. The council has set a target of 30% global market share in the physical AI sector. Both figures are government targets, aspirations with a 14-year horizon, not current achievements or near-term projections.
What is “physical AI”? Physical AI refers to artificial intelligence embedded in hardware systems: industrial robots, autonomous manufacturing equipment, smart sensors, and edge devices where computation happens at the point of use rather than in a remote data center. Japan’s semiconductor and precision manufacturing industries, long-established and globally competitive, are natural feeders for this category. The government’s bet is that Japan can own the physical AI layer the way Taiwan owns advanced logic fabrication: through a combination of industrial scale, technical expertise, and strategic investment.
Three distinct components make up the announced strategy. First: the semiconductor production target. Second: a new national LLM consortium. Third: STEM education reform.
The LLM consortium brings together SoftBank, NEC, Honda, and Sony to develop a large language model at the 1-trillion-parameter scale. Two editorially independent publications corroborate the consortium’s formation, though source page content is unavailable for direct confirmation. The compute requirements for a 1-trillion-parameter model are consistent with estimates from Epoch AI’s model compute tracking data, and the scale of investment that SoftBank, NEC, Honda, and Sony could collectively support aligns with those requirements. This is contextual plausibility, not independent verification of the consortium’s specific technical plans.
On the talent side, the council plans to increase STEM university enrollment from 35% to 50% to supply workers for 17 strategic sectors including AI. That’s a 15 percentage point increase in STEM enrollment share, a major educational policy shift that, if achieved, would take at minimum a decade to produce its intended workforce effect.
Japan’s AI equity markets on the same day: Monday’s Nikkei 225 session, covered in today’s separate brief, saw SoftBank gain roughly 5% as investors priced AI sector optimism. SoftBank is named in both stories, it’s a participant in the consortium and a top equity market gainer on the same day the strategy was announced. These aren’t coincidental data points. They’re the same underlying bet expressed in two different markets.
What to watch: whether the consortium produces a formal governance structure and funding announcement beyond the initial announcement. Whether the 40 trillion yen target gets embedded in Japan’s formal budget or remains an aspirational council document. Whether the STEM enrollment shift produces legislative action.
The TJS read: Japan’s physical AI strategy is coherent in ways that generic national AI strategies often aren’t. The semiconductor target, the LLM consortium, and the STEM pipeline reform are all oriented toward the same specific bet: that the next competitive frontier is embedded intelligence in physical systems, not cloud software. Whether the strategy holds together depends on execution across three policy domains simultaneously over 14 years. The announcement is ambitious. The gap between announcement and delivery is where Japan’s industrial AI story will actually be written.