Unitree Founder Says Humanoid Robots May Take a Decade To Jump To The Mainstream

Humanoid robots may be getting better at walking, dancing and even performing martial arts, but the breakthrough that makes them genuinely useful in everyday life could still be as much as a decade away, according to Unitree Robotics founder Wang Xingxing.

Speaking at the World Robot Conference in Beijing on Thursday, Wang said the robotics industry's equivalent of a "ChatGPT moment" would arrive when a robot could be placed inside an unfamiliar home and successfully perform about 80% of requested tasks using only voice or text instructions. He estimated that milestone could come within two to three years if development progresses quickly, or five to 10 years if progress is slower, CNBC reported.

His latest comments highlight the contrast between the limitations of today's humanoid robots and the enormous investor enthusiasm surrounding companies developing them.

Unitree made a blockbuster Shanghai stock market debut on Wednesday, with its shares closing 460% above the IPO price of 150.80 yuan. The Chinese robotics company raised 6.1 billion yuan, or about $900 million, after its offering was more than 8,000 times oversubscribed, Reuters reported.

That excitement has been fueled partly by Unitree's increasingly sophisticated machines. Founded in 2016, the Hangzhou-based company has attracted global attention with videos showing its humanoid robots dancing, running and performing kung fu. It delivered more than 5,500 humanoid robots in 2025 and reported revenue of 1.7 billion yuan, with net profit reaching 278 million yuan.

But Wang said robots still face fundamental problems that prevent them from operating as flexibly as humans. Current machines often need to be retrained from scratch when given a new task, making their inability to generalize skills across different jobs and environments one of the industry's biggest obstacles.

Precision remains another major challenge. Robots are increasingly capable of planning broad movements, Wang said, but can still struggle with the final few centimeters or millimeters needed to execute a task accurately. Unitree is trying to address the problem through a "self-evolving development loop," in which AI models write and test robot control code before the results are scored and fed back into subsequent versions.

The technical hurdles have not stopped China from building a commanding early position in the humanoid robotics market. Around 19,000 humanoid robots were shipped worldwide during the first half of 2026, a 272% increase from a year earlier, with Chinese manufacturers accounting for 97% of shipments, according to Morgan Stanley figures cited by CNBC. The bank expects shipments from China to reach 50,000 units this year, up from about 12,000 in 2025.

China has made humanoid robotics a strategic priority as Beijing seeks leadership in emerging technologies and ways to offset an aging population and shrinking workforce. The government has rolled out policy support for embodied AI and robotics, while Chinese companies have raced to lower costs and expand production.

Investor enthusiasm has followed. Nomura initiated coverage of Unitree with a buy rating this week, pointing to the company's relatively low hardware costs and rapid product development. The brokerage estimated that outsourced components represented less than one-fifth of Unitree's costs, helping its gross margin rise to 60% in 2025 from 44% in 2022.

Unitree's explosive IPO has also increased scrutiny over whether impressive demonstrations can eventually translate into profitable large-scale deployments. Ying Zhang of the Economist Intelligence Unit told CNBC that the next test for embodied AI would be whether technological advances generate productivity improvements and economic returns significant enough to justify widespread adoption.

The company's shares showed how quickly sentiment can shift after their extraordinary debut. Unitree stock fell nearly 19% on Thursday to close at 687 yuan, according to LSEG data, although that still left the shares more than four times above their IPO price.