China’s real robot revolution is not about humanoids

The recent humanoid games alongside Beijing’s robotics conference at times resembled an expensive comedy reel — an array of robots stumbling, misidentifying obstacles, and executing tasks with toddler-like grace. But the striking spectacle served a purpose — drawing the world’s eyeballs further to the progress being made in robots, adding to a building hype. That buzz helped Unitree, the Chinese producer of humanoids that went viral earlier in the year on social media with their backflipping kung-fu bots, go public last month with a valuation at the end of its first day of trading exceeding 1,200 times its trailing earnings.

Yet China’s real robot revolution has been happening with much less hoopla in industry. Rather than set sprinting records or breakdance, industrial robots do much more prosaic, repetitive stuff, but China is using its superpower of scaling and building supply chains to challenge the global market.

Chinese companies installed 295,000 new industrial robots in China in 2024, capturing 54 per cent of the global total, according to the International Federation of Robotics. Two years ago China already operated over 2mn industrial robots — about 4.5 times more than Japan, the second-ranked nation.

China’s robotics industry revenue surpassed Rmb300bn ($44bn) last year, with a five-year average annual growth exceeding 20 per cent, according to data from the country’s Ministry of Industry and Information Technology. Industrial robot exports grew 49 per cent in 2025 — the first year China became a net exporter of industrial robots. Chinese manufacturers have now captured the majority of its domestic industrial robot market, which was dominated by the global “Big Four” — Fanuc, ABB, Yaskawa and KUKA — prior to 2024.

In the past year, we have seen the mass rollout of autonomous mobile robots in Chinese warehouses, precision surgical arms in hospitals, and specialised cleaning bots in skyscrapers. These devices are purpose-built. They don’t have faces. They have functions.

The application evidence is everywhere. BYD’s Xi’an facility, its largest manufacturing hub, operates thousands of automation arms on EV assembly lines with close to full autonomy on key parts. Shanghai Yangshan Port built an automated terminal that is the world’s largest, operating almost entirely on automated guided vehicles and remote-controlled gantry cranes with minimal human intervention. In Guangdong province alone, regulators recently approved 48 “Robot+” deployment scenarios spanning manufacturing, logistics and service sectors.

None of this requires general-purpose intelligence. It requires solving specific problems with specialised machines.

So why are billions being invested into building humanoids? The obsession with the human form may be the main culprit. We are building robots in our image not because it is efficient, but because it is emotionally resonant. Yet for 99 per cent of industrial tasks, a humanoid shape is a structural liability. And to try to replicate bipedal balance, manufacturers are using algorithms and specialised motors known as servos at astronomical cost.

None of this is to dismiss the genuine progress being made. Advocates for humanoids raise two counter-arguments that merit consideration. The first advantage is environmental fit. Our world — factories, hospitals, homes — was built for humans. Retrofitting it for specialised robots is costly and slow. A humanoid can use existing spaces and tools as is. This unlocks markets in logistics, care and rescue — no wheeled robot can traverse a collapsed building’s rubble.

Recommended

The second is software. AI models that simulate how the physical world works could enable real-time reasoning, not just rote moves. Because humanoids share our form and mimic our mobility, they can learn from vast video data of human activity — unavailable to other designs. This hardware-AI synergy could eventually bring robotics its “ChatGPT moment”, allowing one platform to be updated for many tasks.

But while these are serious prospects, they are distant ones. Lab demos are far from factory-floor reliability. Environmental fit doesn’t justify humanoids’ cost and complexity when most tasks suit optimised machines. Their flexibility remains a solution looking for a problem better solved more simply, for now. Robots are, of course, the future. They are already reshaping our world. But the machines that are having most impact, at least in the near future, are specialised and decidedly non-humanoid.

[email protected]