China moved roughly 19,000 humanoid robots in the first half of 2026, about 77.9% of a global market that grew more than fourfold year over year, with IDC lifting its 2030 forecast past 750,000 units. The headline number matters less than what produces it.

That dominance is not an accident of demand. It is the downstream result of a coordinated stack: local actuator and harmonic-drive supply, subsidized manufacturing capacity, and a domestic buyer base of factories and municipalities willing to absorb early, imperfect hardware. The West and Japan have the reverse problem. Better core research and precision components in some areas, but no equivalent volume flywheel to drive unit cost down the learning curve. Humanoids follow the same physics as EVs and solar: whoever ships the most units first sets the cost floor, and the cost floor decides who survives. A 432% growth rate compounding against a rival that ships a fifth of your volume is how structural leads become permanent.

The strategic risk for global buyers is repeat dependency. If humanoids become embedded labor in logistics, elderly care, and manufacturing, sourcing them predominantly from one country recreates the exact concentration risk that made semiconductors and rare earths geopolitical flashpoints. Expect procurement scrutiny and eventual content rules in the US and EU.

For Japan the situation is uniquely sharp. Japan owns genuine strength in the parts that matter, precision reducers, servo motors, force sensors, and industrial robotics heritage from firms with decades of factory-floor credibility. But strength in components is not the same as owning the integrated platform, and China is now buying those competencies while building its own substitutes. Japan's demographic crisis, the labor shortage that should make it the natural first market for humanoids, is being answered by imported hardware rather than domestic platforms. That is the trap: demand without a supply flywheel.

For SIers and enterprise IT teams, the near-term opportunity is not building humanoids but owning the integration, safety certification, and orchestration layer around them. Fleet management, ROS-based control software, warehouse and care-facility workflow integration, and compliance with Japanese labor and safety standards are high-margin work that hardware origin does not dictate. RPA vendors should read this as the physical extension of their thesis: software automation of screens is converging with embodied automation of tasks. Teams that treat a humanoid as another endpoint to orchestrate, rather than a robotics moonshot, will capture value regardless of whose metal ends up on the floor.