Why the Humanoid Race Has Left the Station

    It is a strange feeling, standing at Computex 2026 in Taipei, watching Jensen Huang unveil NVIDIA’s Isaac GR00T reference humanoid robot – an open platform combining Unitree’s hardware, Sharpa’s tactile hands and NVIDIA’s AI brain – and realising that Japan, once the undisputed leader in humanoid robotics, is nowhere to be seen. The irony is sharp: growing up watching Honda’s ASIMO mesmerise audiences at the 2005 Aichi Expo, dancing and pouring drinks with what then seemed like magic. Today, ASIMO is retired. Toyota’s partner robots never reached the market. That legacy has been replaced by silence.

    The humanoid industry has quietly entered a moment that feels genuinely historic. What was once the domain of choreographed PR stunts is now clocking in for actual shifts. Figure AI’s Helix-02 robots recently completed full eight-hour autonomous factory shifts at BMW’s Spartanburg plant, processing more than 90,000 parts without a single human babysitter. Boston Dynamics, the company built on viral YouTube acrobatics, is now deploying production-ready Atlas humanoids into Hyundai’s Georgia factory – enterprise-grade machines with 56 degrees of freedom, capable of lifting 30 kilograms and autonomously swapping their own batteries for 24/7 operation.

    Tesla, too, has moved beyond the awkward, tethered prototype that shuffled onstage at AI Day 2022. Musk now speaks of a million Optimus units per year, consumer-ready models by the end of 2027 – ambitions that would have been laughable three years ago. And then there is China. Unitree, the Hangzhou-based company that barely existed a few years ago, shipped over 5,500 humanoids last year – more than Tesla, Figure and Agility combined – and now targets 20,000 units in 2026. Their H1 robot recently ran 1.9 kilometres of winding racecourse in just over four minutes, a performance that, by proportional calculation, would break the human 1,500-metre world record.

    This is no longer a race of who can build the most gymnastic prototype. The goalposts have moved. The question now is who can manufacture at scale, who can commercialise, who can put a useful, affordable humanoid into a factory – or a home – first. And in that question, Japan is conspicuously absent.

    But perhaps not for much longer. In March 2026, Japan published its draft “AI Robot Strategy”, a national plan that aims to recapture more than 30 per cent of the global AI robotics market by 2040, positioning the country as a third pole alongside the US and China. Sixteen focus sectors – manufacturing, logistics, nursing care, disaster response – are earmarked for subsidised deployment, with a government‑academia‑industry foundational AI model expected for trial by mid‑2027.

    One wants to believe. Japan’s legendary precision engineering, sensor technology and industrial automation expertise could still be channelled back into the field it once defined. But belief is not a strategy. The window is closing faster than many in Tokyo seem willing to admit. The global humanoid robot market is projected to grow from US$1.16 billion in 2025 to nearly US$42 billion by 2032 – a compound annual growth rate exceeding 50 per cent. That is not a niche. That is the next industrial revolution.

    Japan pioneered bipedal walking. Japan created the cultural iconography of the helpful humanoid companion. Yet today, the headlines are all from California, from Shenzhen, from Munich. The question is no longer whether humanoid robots will reshape the global economy. The question is whether Japan will help build them – or merely watch from the sidelines.