Japan has formalised a plan to deploy ten million AI-powered robots across 18 industries by 2040, backed by up to one trillion yen over five years. The first task is a national “physical AI” foundation model able to interpret language, images, video and sensors together.
The ambition responds to an ageing population and persistent labour shortages, but its credibility will be determined by near-term stage gates rather than the headline robot count.
A consortium connects models with Japan’s hardware strengths
The Ministry of Economy, Trade and Industry and NEDO commissioned Noetra and national research institute AIST to lead development from fiscal 2026 to 2030. An initial model is due within the current fiscal year, followed by annual upgrades.
Noetra is majority-owned by SoftBank, NEC, Sony and Honda, with researchers from Preferred Networks and AIST. The group combines cloud, computing, imaging, robotics and industrial expertise.
Manufacturers will contribute operational data, which can help a model learn real environments rather than curated demonstrations. Contracts need to protect commercial information and ensure contributors understand how shared data benefits competitors.
The model should expose interfaces usable by different robot bodies. A national platform has limited value if every machine requires a separate integration and safety framework.
Labour shortages create opportunity and difficult expectations
Japan wants robots in restaurants, food manufacturing, medical care and other sectors where vacancies are persistent. The country already has experience in elder care, disaster response, factories and hazardous cleanup.
Robots can take repetitive, dangerous or physically demanding tasks, but they do not automatically replace an entire job. Deployment often creates new work in supervision, maintenance, workflow design and exception handling.
Healthcare and hospitality require social judgement as well as manipulation. A robot that technically completes a task may still reduce dignity or service quality if introduced around staffing cuts rather than user need.
Workers should help select tasks and evaluate ergonomics. Safety testing must include crowded spaces, unusual objects, children and equipment failure—not only controlled factory floors.
Training policy should anticipate changed roles. Technicians, care staff and restaurant workers may need confidence supervising and stopping machines, while vocational programmes develop maintenance and integration skills. The transition will be judged partly by whether workers share the benefit.
The national strategy could become exportable if it produces dependable components for industries facing similar demographics. International success will also require standards, language support and confidence in data governance.
Stage-gate funding makes early evidence decisive
The trillion-yen figure is a ceiling. Only the first two years are secured, with funding reviewed annually against milestones. That gives the government a route to reduce support if the model fails to deliver.
Useful milestones should measure generalisation, safe recovery and deployment economics rather than benchmark performance alone. A model that recognises a scene but cannot act reliably is not physical intelligence.
The first release will indicate whether a shared model can transfer among robot types and industries. Independent testing and publication of failure cases would help developers judge where it is ready.
South Korea’s parallel robotics push adds competitive pressure, while Japan’s consortium model spreads risk across industry and state. Competition should not encourage premature deployment in settings involving vulnerable people.
Ten million robots by 2040 is an aspirational number, not a labour policy by itself. Japan still needs training, workplace redesign and decisions about immigration and care.
The programme’s strongest contribution could be a common intelligence layer built from Japan’s practical robotics knowledge. If early stage gates reward safe, useful work rather than spectacle, physical AI may become a tool for sustaining services with fewer workers—without treating people as simply another shortage to automate away.
