China’s humanoid robots aren’t smart enough to take your job – yet

Teaching humanoids to handle labor-intensive work could one day meet an acute need. China’s population is aging and shrinking, eroding a critical demographic advantage for its industrial base: a vast pool of workers. And China’s military researchers have explored the potential uses of robotics and artificial intelligence in future weapons systems. The Liuzhou center, which Reuters visited in April, aims to sell robot-training data to factories. But three staffers acknowledged the subsidy-dependent project currently has no clear path to profitability, citing high operating costs and low prices for its data in the nascent market. The facility captures a central tension in China’s humanoid push: Government funding and manufacturing investment are racing ahead of the robots’ ability to operate outside of choreographed demonstrations – like the kung fu routines, dances and boxing matches that have made the machines internet sensations. Though China’s robot makers produce world-beating hardware, the humanoids lack the intelligence needed for general-purpose work, limiting their practical uses, a half-dozen industry executives and researchers told Reuters. Critically, they struggle with situations requiring intuition or deviation from programmed routines. This Reuters examination of China’s humanoid boom is based on hundreds of corporate filings and government procurement records; visits to factories, robot-training centers and humanoid showrooms; and interviews with 40 people including executives, investors and researchers. It reveals a widening gap between government ambitions and robot capabilities – resulting in a market where manufacturing capacity outpaces commercial demand, key players rely on subsidies and prices are falling. The humanoid-robot industry shipped about 20,000 units globally last year, with Chinese manufacturers accounting for 95%, according to BofA Global Research. Gan Xiaobin, an official at China’s Ministry of Industry and Information Technology, said in July the country expects to build more than 100,000 humanoids this year. BofA forecasts 1.2 million annual shipments globally by 2030. The national, provincial and local governments spent at least $230 million buying humanoids and related items in the first half – up from $62 million a year earlier and $6 million in the corresponding 2024 period, public procurement data show. Reuters reviewed nearly 1,000 fulfilled tenders covering state-entity purchases of robots, training systems, equipment and demonstration projects. Competition is driving down prices. Morgan Stanley expects the average price of a humanoid to decline 15% in 2026, before recovering modestly thereafter. The average price of Kuavo, a full-size humanoid made by Leju Robotics, fell 26% year-on-year in 2025 to about $46,000. Kevin Li, chief marketing officer at Chinese robot-rental platform Sharebot, said dancing robots – which companies lease for promotional events – now rent for $440 to $600 a day, down from roughly $1,500 last year. Some analysts question whether a general-purpose machine built in humans’ image makes economic sense when robotic arms – purpose-built for specific manufacturing tasks – are cheaper, more reliable and already in wide use. Zhao sees potential in niches such as homes or care facilities, where a human-like machine could use stairs, doors and tools with minimal modification. But in the near term, the market for humanoids “will remain concentrated in dangerous, repetitive and structured work.” Early commercial cases for humanoids are emerging. Humanoid maker Galbot is deploying robots in pharmacies across more than two dozen Chinese cities to address staffing constraints and meet medicine-delivery demand. The AI brains of humanoids differ from those in chatbots. A factory humanoid needs to process vast quantities of visual data and convert what it sees into physical action: Gripping a piece of fabric; calibrating force on a screw. Companies call these vision-language-action models. They are much harder to train than LLMs, and their development is at an earlier stage, because they require thousands of hours of exposure to real-world scenarios. China’s humanoid campaign relies on a simple equation: More robots in training equals more data. The sector has roughly 500,000 hours of high-quality training data, but needs 100 million hours for robots to achieve true physical intelligence, said analyst Poe Zhao, citing industry estimates. “That is an orders-of-magnitude deficit that neither capital nor volume can close quickly,” he said. Inside a training center in southern China, more than 100 humanoid robots stand in neat rows. A dozen human trainers wearing headsets guide some of them through simple tasks – sorting crates, packaging noodles, making coffee. The humanoids are slow and awkward. A novice trainer might get them to produce one usable movement in 300 attempts; an experienced trainer, one in 50. Robot maker UBTech 9880.HK won an $18 million tender from the Guangxi regional government last October to supply the training facility with humanoids and related hardware. The mission: to produce training data for so-called embodied AI – robots able to perceive, decide and act in the physical world. The trainers use headsets to see from the robots’ perspective and handheld controllers fitted with sensors to translate their own movements into the robots’ actions.

China’s humanoid robots aren’t smart enough to take your job – yet

Government support is driving the sector

A boom built on state money and falling prices

Practical use cases remain narrow, but the data race continues

China’s humanoid push is racing ahead of real-world capability

Related News

Hooker Furnishings schedules fiscal Q2 earnings call

Nvidia to present at Goldman Sachs Communacopia + Technology Conference

U.S. FDA Approves Gilead’s Bixlenvo™, a New Once-Daily Single Tablet Option For Virologically Suppressed Adults With HIV, Including Those On Complex Regimens

Abbott CEO Robert B. Ford disposes of $ US 46.17 million in common shares

Satellogic - Appoints Dustin Greer As Interim CFO Effective August 21, 2026 - SEC Filing