Japan is positioning its manufacturing expertise as a foundation for 'physical AI'—robot systems that learn to act flexibly in the real world—but faces a critical risk: the technology's real value is shifting away from hardware toward the AI models that power it.

What happened

Prime Minister Sanae Takaichi announced at her New Year's news conference that Japan would harness operational data from its manufacturing and service sectors to advance physical AI—systems that allow robots to autonomously support human workers with precision. The AI Robot Association was established in 2025 to develop foundation models for AI-powered robotics, while government research calls have been launched for AI robotics and physical AI development.

Why it matters

Physical AI represents a fundamental shift from traditional industrial robotics, which perform rigid, pre-programmed tasks in controlled settings. The new systems can function flexibly in unfamiliar environments and unexpected situations without explicit programming of every movement—a capability that demands multimodal AI models, diffusion models, and large language models as the 'brain' rather than hardware alone. Japan's decades of manufacturing expertise could theoretically help it compete in this space, but observers note that value creation is increasingly shifting away from hardware and toward the AI foundation models that serve as the 'intelligence' of robotic systems.

What to watch

The critical tension is whether Japan can build competitive AI foundation models despite lagging in the global AI race. While the country retains strength in industrial robotics and manufacturing precision, physical AI requires software-centric innovation—a domain where hardware excellence alone no longer determines competitiveness. The outcome will depend on whether Japan's AI research efforts can close the gap in the model-building layer.