20 Firms Collapse, $20B Vanishes! China’s Robot Bubble Bursts, R&D Lags Behind Pig Farming

A Chinese robot suffered another public setback on October 22nd. At the 2026 World Physical AI Conference and Unitree Partner Ecosystem Summit in Hangzhou, Zhejiang province, the host announced, it's time to show what this thing can really do. A white humanoid robot took a short running start, leaped into the air and delivered a spinning kick. But one leg broke off at the knee and flew away. The robot then crashed onto the black carpet. The robot was the H2X Kung Fu Legend Guardian, which Unitree had unveiled just one day earlier. A promotional video had shown the H2X easily lifting the front of a car, denting a locked car door with a flying knee kick, and jumping from a platform about 13 feet high before completing a flip. The promotional video also showed the H2X blocking a knife attack and helping with a traffic accident rescue. The company warned users against dangerous modifications or operations. It also told users to keep a safe distance of at least 7 to 10 feet during use. Commenters wrote, Great takeoff. Falls apart on landing. You guys keep going. The lower leg had somewhere else to be. Another wrote, They should have used a Unitree robot to press the Maextro pedal. Then people would argue about whether the leg or the pedal broke off first. Huawei might have gotten off the hook. This was in reference to a recent controversy about Huawei Maextro's brake failures. Another commenter wrote, Forget the robot's quality. If I can still see this news, Unitree's legal team clearly isn't doing its job. Someone else wrote, Tell a story. Put on a programmed show and make a fortune off the A-share market first. The legal team may be too busy with a stock offering to deal with this. Another commenter wrote, A failure like this just one day after the launch is a bad sign. Unitree is headed for serious trouble. Bad news could be coming soon. A few days later, Unitree released its first earnings report since going public. Net profit for the third quarter of 2026 fell sharply from a year earlier and was cut roughly in half from the previous quarter. Chinese media reported on October 28th that Unitree released its third quarter 2026 report that evening. revenue for the first three quarters reached 1.8 billion yuan about 254 million us dollars up 55.7 percent from a year earlier net profit attributable to shareholders was 387 million yuan after excluding one-time gains and losses that profit was 349 million yuan down 18.8 percent In the third quarter alone, revenue reached 665 million yuan, up nearly 70 percent from a year earlier. Profit attributable to shareholders was 113 million yuan, down about 17.3 percent. After excluding one-time gains and losses, that profit was 105 million yuan, down 17.4 percent. Figures in the company's earlier prospectus put first half net profit attributable to shareholders at 274 million yuan. That included about 50 million yuan in the first quarter and about 224 million yuan in the second. Based on those figures, third quarter net profit fell 49.5% from the second quarter The report showed third-quarter research and development spending of 127 million yuan. That was up nearly 250% from a year earlier and accounted for 19% of quarterly revenue. Spending for the first three quarters totaled 263 million yuan, up 191% and equal to about 14.5% of revenue. Unitree's relatively low research spending has already drawn attention. Economist Fu Pang previously told the media that as the leader in humanoid robotics, Unitree should use the money raised from its stock offering for R&D, but its financial reports showed research spending below even what Muyuan Foods spent on research for its pig farming business. The report also showed a sharp rise in selling expenses. Unitree spent 230 million yuan in the first three quarters. That was up 202% from 76 million yuan, about $11 million a year earlier. Net cash from operating activities for the first quarter was 194 million yuan, down 54.7% from a year earlier. By the end of September, inventory had risen to 1.02 billion yuan, up 177% from the end of last year. Accounts receivable reached 143 million yuan, up 144%. Unitree's report showed revenue growth without profit growth, alongside worsening financial and operating conditions. Its stock had rebounded sharply before the results came out. The Southern Metropolis Daily reported that shares rose more than 65% over seven consecutive trading days. They closed at 695.9 yuan on October 28, valuing the company at about 281.5 billion yuan. Industry insiders describe this as a move by major market players to push up the price and sell before bad news, raising suspicions. Even before the rebound, Unitree had lost 163.4 billion yuan in market value in just over two months. Its market value had peaked at about 444.9 billion yuan during trading on its August 19th debut. Fu Pang previously said Unitree's valuation before its listing was already well above the market's general expectations. The robotics industry is still at an early stage. Although the company has completed its initial public offering, he said it's still more like a private venture capital investment. It is not suited to investors seeking steady returns. Industry professionals said the timing of Unitree's failed stage performance and weak earnings was sensitive. Even before the accident drew attention, several scholars had pointed to a gap between Chinese robotics demonstrations and real-world uses. A wave of warning signs has recently emerged at robotics companies. The industry appears to be preparing for a robotics bubble to burst in China. The National Business Daily recently reported that a self- Self-identified employee of Nanjing Wujing Future Power Technology. known as Wujing Power, said the company had not paid their wages since August. It stopped making housing fund contributions in September and social insurance contributions in October. Operations have now stopped. Total unpaid wages are nearly 8 million yuan. The company's controlling owner, Cui Xiaofei, also cannot be reached. Wujing Power's CEO and general manager, Yuan Haihui, told the outlet that he could not reach the controlling owner and was not in a a position to comment on the situation. Company information shows that Wujing Power develops embodied AI and four-legged robots in Jiangsu province. It was founded on February 11, 2026. Its core team includes people from leading universities such as Tsinghua University, Zhejiang University, and Shanghai Jiao Tong University, along with experts in AI, robotics, and chips. That situation differs sharply from earlier media coverage. In June, Nanjing's official WeChat account, Nanjing Press, said Wu Qing Power had signed framework orders worth nearly 500 million yuan, with several central government-owned companies just four months after its founding. Deliveries had begun. It also had potential local orders worth nearly 500 million yuan and secured 110 million yuan in funding. Industry insiders say Chinese robotics companies often announce large orders, but some are fake. Shaotian Lan, founder of MechMind Robotics, previously wrote on WeChat that some embodied AI companies mainly bring other players together. Based in Beijing and Shanghai, they have big names, high valuations, and appearances on the spring festival gala. He said they used big data collection centers and related party deals with local governments, investors, and suppliers to generate fake, unsustainable revenue. This can trap companies in a cycle of continued fraud and cash losses. These companies create value by bringing together funders, technology providers, manufacturers, and sales channels. They do not necessarily own core assets or have production capacity themselves. Humanoid robot data collection centers are also called robot driving schools. Operators use controllers or virtual reality equipment to remotely guide robots through tasks such as picking up objects, moving them, and assembling parts. The centers record visual, touch, and force data. Along with joint angles and other physical data, he explained how fake orders work. A local government or state-owned investment platform pays to build a data collection center. A robotics company sells equipment to the center and books the revenue. The center then sells data back to the company, sending the money back around. The fake order practices Shaotian Lan exposed are widespread in the industry, and his own company is among those involved. MechMind listed on the Hong Kong Stock Exchange on September 1st this year. It was billed as the first listed company providing the eyes, brain, and hands of embodied AI. But its shares fell below the offering price on the first day of trading. The Star Market Daily recently quoted Tang Jingcao, chairman of Shuimu Capital, as saying, Shaotian Lan's decision to speak openly may mean these problems have built up to a certain point. What is actually happening inside the industry is probably more widespread than outsiders can see. The co-founder and CEO of robotics maker Lingyu Intelligence was more direct. Almost all data collection centers use this approach, buy equipment from robotics companies, then sell data back to them. Market research and consulting firm Interact Analysis said that by the end of April 2026, China had at least 90 humanoid robot data collection and training centers, either in use or planned for construction. Beyond fake orders, the robotics industry faces a more serious problem. Very few orders lead to real commercial use. MIR, a Beijing-based research and consulting firm focused on China's business-to-business industrial market, reviewed humanoid robot procurement orders in July. Across 46 projects, universities and vocational schools dominated purchasing. Educational institutions accounted for more than 90% of the total value of awarded contracts. Data from U.S. market research firm International Data Corporation showed actual global humanoid robot shipments of about 18,000 units in 2025. More than 85% went into entertainment, education, research, tours, and retail assistance. These uses mainly involved testing technology, demonstrations, and human interaction. Fewer than 15% went into manufacturing. and logistics, and most of those remained pilot projects. Bank of America Global Research put global humanoid robot shipments at about 20,000 units in 2025, with Chinese manufacturers accounting for 95%. Analysts said a genuine commercial order should meet several conditions. The customer buys the robot for an actual production need. The robot has been delivered and accepted. And it can work reliably without constant intervention from engineers. Judging by where the robots have gone, at least 85% do not qualify as genuine commercial deployments. Wu Jingpower is not the only Chinese robotics company to run into trouble over the past two years. Public reports and court notices show that during the 21 months from February 2025 through October 2026, more than 20 robotics companies went bankrupt, dissolved, entered liquidation, or shut down business lines. They operated in at least 11 different fields. The Shenzhen Intermediate People's Court held the first creditors' meeting for Yinghe Robotics on October 19 this year. The agenda covered a bankruptcy liquidation plan and asset disposal. The company raised more than 600 million yuan over six years. Its shareholders included Meituan, ByteDance, and GSR Ventures. At its peak, it was valued at more than 500 million US dollars. In early September, a new bankruptcy application was filed against Zhicheng Power. A creditor asked the Suzhou Industrial Park Court to place the company into bankruptcy restructuring. Zhicheng Power specialized in pool cleaning robots and had cumulative sales of nearly 100 million yuan, but problems including difficulty collecting payments caused its cash flow to break down. On June 17, the Dongguan Third People's Court accepted a creditor's application to liquidate Dongguan Lesheng Robotics through bankruptcy. The company specialized in robot vacuums. On May 7, the Shanghai Pudong New Area People's Court accepted a bankruptcy liquidation for Deep Blue Robotics Shanghai. Employee claims exceeded 23 million yuan. Its parent, Deep Blue Technology, once appeared on the Hurun Global Unicorn Index, with a valuation of 14 billion yuan. It was described as one of three industry giants, alongside SenseTime and iFlyTek. On March 4, the Hangzhou District People's Court in Lianyungang accepted a bankruptcy liquidation case for Chuangtze Intelligent Robotics Jiangsu. Chuangtze focused on commercial public service robots. Courts have listed it as subject to enforcement four times this year. On January 23, the Shanghai Pudong New Area People's Court accepted a bankruptcy liquidation case for Shanghai Longhui Medical, the company specialized in robots for orthopedic surgery. In October 2025, the Beijing Fangshan District People's Court accepted a bankruptcy liquidation case for CloudMine Robotics. Its parent, CloudMines, was once valued at more than 20 billion yuan. Taiwan's central news agency recently cited commentary, saying two layers of China's robotics bubble have been fully exposed. One is a technology bubble, built on products companies cannot put to practical use. Another is an investment bubble, built on deliberate fraud and misleading presentations. An October report on embodied AI from the China Academy of Information and Communications Technology also described a divided industry. Investor and policy interest keeps rising, but large-scale deployment and sustainable business models still face obstacles. There is a clear gap between technology demonstrations and actual deliveries. The report said the market would complete its first round of consolidation over the next two or three years, eliminating companies and business models that lack real value. Commenting on the wave of robotics bankruptcies and liquidations, Fu Yifu, a special researcher at Sushang Bank, recently told the media, This does mark a shift from unchecked expansion to consolidation and a shakeout. Companies once survived on concepts and fundraising. Now they have to survive on orders and gross profit. The central problem with building a viable business is that orders do not equal revenue. Chairman of EFORT Intelligent Robot Company, Ye Wei, said at a public event in June 2026 that cutthroat competition in China's industrial robotics sector was nearing its end. Leading companies were pulling ahead, while mid-tier and smaller companies were leaving the market faster. The industry was entering a stage where the survivors would dominate. Competition would shift from price to product quality and capabilities. Experts say China's robotics industry is declining before it has fully matured. They link that decline closely to falling export orders and a slowing economy, which has left large numbers of factory robots idle. Both investors and users have become pessimistic about the industry's prospects. At a recent auction at a shuttered factory, industrial robots that once cost hundreds of thousands of yuan each were offered for less than one-tenth of their original price, a factory representative told prospective bidders. And these robots are all from Sichuan. These robots are all included in the total lot of 400,000 yuan, about 56,000 US dollars. Don't wonder why the price is so low. That's just how things are now. Go ahead and bid. Reports of old robots sitting idle on a large scale have spread across Chinese social media in recent months. Many users say they often see posts describing robots taking a break at a factory, with nothing wrong except a lack of work. Several business owners said orders have fallen sharply in recent years, leaving production lines below capacity. Expensive automation equipment now sits unused. Like laid-off workers in the past, the machines wait quietly, with no certainty that they will run again. Analysts say views of China's economy are converging among economists. Alicia Garcia Guerrero, chief Asia-Pacific economist at French bank Natixis, recently wrote that Chinese companies financial health continues to deteriorate. Even the so-called three new industries are no exception. The new three industries refer to EVs, lithium batteries, and solar cells. The Chinese Communist Party's support for the new three industries follows a model of heavy investment in physical assets aimed at expanding overseas and capturing global markets. In the early stages, central and local governments directly lowered costs for businesses and consumers. They provided large consumer subsidies, favorable land and factory construction terms for producers, and subsidies for electricity generated from solar power. For robotics, authorities have gradually scaled back direct subsidies for prototype development. The goal is to stop companies from racing to build prototypes just to collect subsidies. In practice, the industry has entered a stage where companies must respond to real-world needs and survive on their own. Even the heavily supported New 3 Industries face these problems. For a robotics industry short of core technology and supporting hardware, with authorities unable to inject more funds, the bubble is already bursting.