Unitree Founder Says Robots Face a Millimetre-Scale Challenge

At a sprawling trade fair in the eastern Chinese city of Hangzhou this week, one of the country’s best-known robotics entrepreneurs offered a surprisingly sober take on how far humanoid machines still have to go.

Wang Xingxing, founder of Unitree Robotics, took the stage at the fifth Global Digital Trade Expo on September 24 with a talk titled “From Mechanics to Intelligence: An Evolutionary Theory of Embodied Futures.” His central argument was blunt: the biggest bottleneck for robots today is not software or ambition, but whether they can consistently work within a tolerance of a few millimetres.

“The core bottleneck is whether we can solve a work error of just a few millimetres,” Wang said, according to a report by Jiemian News. That tiny margin, he suggested, is what separates a robot that performs a scripted demo from one that can reliably assemble, pack, sort or repair things in the messy real world.

Wang’s framing looked back at how quickly public expectations can shift. When ChatGPT arrived, he noted, it transformed how ordinary people thought about artificial intelligence almost overnight, opening the industry’s imagination in ways that previous generations of software never managed. He believes the world of physical, embodied intelligence is heading toward an equivalent turning point.

His benchmark for that moment is specific. When a robot can enter 80 percent of unfamiliar environments and complete 80 percent of tasks using voice commands and physical capability, Wang said, the industry will have reached its own “ChatGPT moment.” Until then, he implied, most demonstrations remain carefully staged.

Unitree has become one of the most visible names in China’s fast-growing robotics sector, known for four-legged machines and humanoid prototypes that have appeared at trade shows, research labs and viral video clips. The company’s public profile has risen alongside a broader wave of Chinese investment in automation, from warehouse logistics to consumer gadgets.

Yet Wang’s comments reflect a tension running through the entire field. Videos of robots walking, jumping and dancing generate enormous attention, but the commercial value of such machines depends on doing repetitive, precise work at a cost that undercuts human labour. Millimetres matter enormously in factories, where a fraction of a millimetre can mean a scratched component or a failed assembly.

The Hangzhou expo, a government-backed showcase for digital trade and technology, drew companies from across China’s consumer and industrial tech landscape. Wang’s appearance there placed him alongside a broader push to commercialise artificial intelligence and robotics, an area where Chinese firms have moved quickly from research to product launches.

For global observers, the message was a reminder that the humanoid robot race is not only about spectacle. The companies that win may be the ones that quietly solve unglamorous engineering problems, from grip force to vibration control to the calibration that keeps a robotic arm accurate after thousands of hours of use.

Wang did not offer a timeline for his predicted breakthrough, and he was careful not to promise that it is imminent. But his framing suggests that the next phase of robotics will be judged less by how human a machine looks and more by how precisely it can work, day after day, in places it has never been before.