Is China’s humanoid robot industry a bubble? Expert comment jon.wallace 2 September 2026 The market debut of Unitree, China’s leading humanoid robot company, has stoked fears that Beijing’s ambitions for robots exceed demand. Until humanoid robots reach their ‘ChatGPT moment’ those worries will remain. China’s robotics industry has been making headlines again. Videos of humanoid robots running, jumping and boxing at Beijing’s World Humanoid Games swamped social media feeds in late August – including a robot that broke Usain Bolt’s 100m record.Meanwhile financial media were gripped by wild gyrations in the stock market valuation of Unitree Robotics, China’s leading humanoid company. After soaring to a $66 billion valuation on its debut listing in Shanghai, it then plunged by around $30 billion. By 1 September, the share price had stabilized at levels roughly 45 per cent down from its post-IPO peak.Underlying these events lies a stubborn question. Does the enthusiasm surrounding China’s extraordinary emerging robot industry outstrip its financial fundamentals?The answer is complex. The industry’s future prospects are not only a matter of whether humanoid robots can earn enough to justify the market valuations of the companies making them. The technological trajectory of the industry is also crucial. One key issue is when, if ever, the industry will attain a watershed ‘ChatGPT moment’.There are also important issues relating to state support. Beijing classified robotics as a priority strategic emerging industry in its 15th Five-Year Plan (2026-2030). It has backed this intent with a venture capital fund focused on robotics, AI and innovation that is expected to raise around $138 billion. Filling out the picture, Beijing recently unveiled ambitions to lead efforts to set industry standards for humanoid robots.Indeed, China has already established a dominant lead in the global humanoid robotics industry: its makers accounted for more than 90 per cent of global shipments in the first half of this year. What Unitree’s debut means for the industryThere is little doubt that Unitree’s IPO was met with irrational exuberance. At one point, its share price rocketed so high that the company’s price-to-earnings ratio reached a giddy multiple of 1,300, compared with an average of 124 for the Shanghai STAR market. It is therefore little wonder that its share price retreated in the days after the debut. The question is: are humanoid robots capable enough to do a broad spectrum of jobs well enough to make money? And if not, how soon before they are? Of course, these events do not necessarily signify that China’s humanoid robot sector lacks technological merit and commercial prospects. When the dot com bubble burst between 2000 to 2002, trillions of US dollars were wiped off the valuations of internet companies. But it did not signify the death of the internet.What Unitree’s gyrating share price does highlight is the business model for humanoid robots. The question is: are humanoid robots capable enough to do a broad spectrum of jobs well enough to make money? And if not, how soon before they are?Making robots fit for jobsHumanoid robots still have important flaws. Battery life is often too short. Robots often struggle with navigation, avoiding obstacles and knowing when to stop: many ran into barriers at the World Humanoid Games. Manual dexterity is often clunky. And robots’ ‘eyes’ are a bottleneck in areas such as contextual comprehension, or for example guessing what is hidden behind a partial view.But while shortcomings remain, progress to address them is rapid. A recent breakthrough by AgiBot, a Chinese maker, claims a battery life of up to 10 hours on a single charge. On dexterity, industry leader Linkerbot has developed a hand that can position objects with a 0.2 millimetre margin of error and lift something weighing 30 kilogrammes. And industry leaders such as Hikvision are developing 3D machine vision that can capture an object’s depth, shape and spatial coordinates.One key goal for the industry is to reach the so-called ‘ChatGPT moment’. That is defined by those in the industry as when a humanoid robot can be placed in unfamiliar surroundings and complete 80 per cent of tasks after hearing text or voice commands. When such a moment is reached, humanoid robots will be able to complete a range of tasks without the kind of specific pre-programming currently required. That would give them the adaptability to undertake a range of tasks with similar efficiency as human workers, but at a cheaper cost – the technology’s key selling point. Wang Xingxing, CEO of Unitree, said last month that such a moment could arrive ‘in two to three years if things move fast, or five to ten years if they move slowly’. Others have more aggressive projections. Wang Xiaogang, chairman of ACE Robotics, said his company expected to reach the moment by the end of 2027. Finding jobs fit for robotsAs humanoid abilities improve, the hunt is on to find them money-making jobs. Thus far, the market is still in its infancy. But the possibilities appear considerable: at the World Artificial Intelligence Conference in July, robots were filmed making coffee and pizza, doing laundry, dispensing medicines, doing production line work, cleaning toilets, giving massages and more. Related work China’s Five Year Plan commits to economic resilience – as the Iran war exposes the fragility of global supply The search for robot employment is part of a bigger push by Beijing to diffuse AI throughout its economy. The Five-Year Plan states that by 2027 AI devices, agents and applications should be used in 70 per cent of key industry sectors and by 2030 in 90 per cent. However, Unitree’s prospectus shows that the lion’s share of the company’s humanoid robot revenue comes from universities, academic bodies and research laboratories – who often are buying the robots with the assistance of state subsidies. These labs then commonly sell the data they derive from training the robots back to robot companies such as Unitree. The circular nature of this economy has raised some concern over its sustainability. Only about 9 per cent of Unitree’s revenue came from sales to industry, suggesting lacklustre commercial demand thus far.More technological dominance?The upshot of such market dynamics is that the humanoid robot industry in China looks similar to the early years of the Chinese EV industry. Considerable government funding is helping to bolster demand, buying time for companies to develop technologies.But there are many uncertainties ahead. Until robots reach their ChatGPT moment, fears of bubble dynamics will persist, creating risks for investors and for scores of Chinese robot start-ups daring to dream big. Beijing hopes to reach that moment swiftly and create another industry like EVs, in which Chinese manufacturers lead the world, and China has an opportunity to set global industry standards for a potentially revolutionary future technology. In that case the Chinese robotics industry could follow the course of the EV industry. That would see it max out domestic demand, before seeking to establish export success through a similar mix of high quality and unbeatable cost. But the EV industry’s export success may itself hinder a similar effort in robotics. Chinese EVs have become a source of significant security, economic and resilience concerns in markets like Europe, as they threaten to overwhelm European manufacturers and create dependencies on Chinese supply.As a result, these markets could be quicker to raise barriers to Chinese robot imports. The US already banned new foreign-made humanoid robot imports in July. And Washington may pressure other countries to follow suit. On the other hand, it’s not clear what the alternative to a future robot industry dominated by Chinese models will be – when other countries’ robot industries lag so far behind. The sooner the ChatGPT moment arrives for China’s robots, the sooner European policymakers will be confronted by another set of familiar issues: how far should they go to defend their domestic industries? And how deeply should they integrate their technology with China’s?
Chatham House
The role of non-bank financial intermediaries in the financial system has increased markedly over recent decades. This column documents how the growth of private markets, one of the fastest growing segments of the sector, has created multiple channels of interconnection between banks, firms, institutional investors, private equity, and private credit funds. Banks’ exposures extend from lending to common borrowers, to direct lending to funds, financing vehicles and linkages with asset managers. While available measures suggest that euro area banks’ exposures remain limited, data gaps and hidden leverage make it difficult for intermediaries and supervisors to assess risks across the full intermediation chain.
Center for Economic Policy Research
Tom Griffiths and Arthur Spirling are co-directors of the DaIS initiative, which builds on the deep links between AI and data science at Princeton while creating space for continued rapid progress in each of the fields.
Princeton University
Pension funds have traditionally invested in global bond markets which offer predictable income and support stable portfolios. This column documents a recent shift in pension investing away from fixed-income securities and towards riskier and less liquid assets. A search for yield is an important reason for this change, along with the transition from defined benefit to defined contribution pension systems. This transformation presents opportunities for higher returns and more diversification, but also exposes bond markets and pension funds to more risk during volatile periods.
Center for Economic Policy Research
