For stock investors, reading Gold Qilin analyst reports offers authority, professionalism, timeliness and comprehensiveness, helping uncover potential thematic opportunities! Source: Fukai Finance. Unitree Technology was not the biggest gainer this week (October 19 to 25), but it was undoubtedly one of the most closely watched stocks. From its opening price of 1,100 yuan per share on its debut day to a low of 404.03 yuan per share, Unitree Technology took only 36 trading days, with a maximum decline of more than 63% and a market value evaporation of over 280 billion yuan. On October 23, Unitree Technology recorded its first 20% daily limit-up since listing, closing at 624 yuan per share, with a weekly gain of 46.14%, pushing its market value above 250 billion yuan. Physical AI should be one of the drivers behind this rebound. On October 22, the 2026 World Physical AI Conference and Unitree Partner Conference was held in Hangzhou, marking the first high-profile industry conference held by Unitree Technology in its ten years of existence.
Multiple positive catalysts
In October this year, the first batch of six national standards for the humanoid robot sector was officially approved, covering core areas such as safety requirements and performance testing, clearing institutional obstacles for the standardized and large-scale development of the humanoid robot industry. At the same time, multiple government departments jointly issued documents explicitly promoting embodied intelligent robots entering commercial service scenarios, opening up new space for the industry to extend from the industrial end to the consumer and service ends. Tesla Motors (NASDAQ: TSLA) recently released its third-quarter report, which contained relatively positive statements about robots. Optimus has entered mass production at the Fremont production line, with the first batch of robots meeting mass production quality standards, and the manufacturing process officially confirmed as suitable for commercial products. Musk proposed a target of 20,000 units per week by the end of 2027, corresponding to an annualized rate of about 1 million units, doubling the previous guidance of 10,000 units per week. Huachuang Securities pointed out that Tesla's third-quarter earnings call released signals of Optimus mass production exceeding expectations, the product technology roadmap is gradually becoming clear, and mass production has taken a key step, with the robot sector expected to usher in a new round of market momentum. However, Musk also stated that it is still in the early ramp-up stage, and actual output depends on design iteration and reliability convergence. What made the market even more optimistic was the news that OpenAI CEO Sam Altman said the company would enter the robot-building space and produce a demonstration prototype by 2027, with GPT-6 Astra's benchmark performance on robots exceeding expectations. CICC noted that GPT-6 Astra's benchmark performance on robots exceeded expectations, marking that humanoid robots are beginning to have their own general foundation large models. Although there is still a gap compared with Transformer, this move is crucial for truly creating physical AI model applications. Tianfeng Securities pointed out that large model companies (OpenAI) entering the robot track will once again amplify data procurement demand, bringing value revaluation to data assets.
The industry's biggest opportunity is coming
Wang Xingxing, founder of Unitree Technology, said at the 2026 World Physical AI Conference that in the coming years, large model companies will pay increasing attention to machine intelligence and the robot industry. At the fastest, it could be one to two years, and at the slowest, two to three years, before everyone sees more progress. The company is developing its own models and cooperating with large model companies at home and abroad to jointly promote physical AI faster across the critical point. "Since last year, I have hoped that robots could cross such a moment: when completing about 80% of tasks, they would only need voice and language instructions. I hope that by the end of this year or next year, you can directly tell a robot to tidy up the room, and it will be able to tidy up the room. At present, the alignment problem between the input and output of AI models and real robots has not been well solved. Sensors will have errors, and actuators will also have errors when outputting actions," Wang Xingxing said. Wang Xingxing believes that physical AI needs to break through three major bottlenecks before it can open up a huge market. First, the large-scale manufacturing capacity for robot hardware. Current annual output is only tens of thousands of robots and hundreds of thousands of joint motors and modules, far from enough for large-scale popularization. In terms of production scale and cost control, Unitree Technology and global peers still have a lot of work to do. Second, independently develop or work with partners to advance the implementation of physical AI models. It is unrealistic for the company to directly develop exactly the same models as the world's leading multimodal large model companies, but it can do better in robots and industrial AI-related models. Third, larger-scale server clusters and computing power clusters, as well as corresponding computing power chip production capacity. Since last year, the company has adopted pretrained deep reinforcement learning methods to conduct large-scale training for each action, thereby improving the robot's control performance. At the conference, Unitree Technology announced the launch of the "World Physical AI Competition" and the "Yuchuang Plan." The World Physical AI Competition has an annual support amount of 10 million US dollars, rewarding teams or individuals who achieve outstanding technological breakthroughs in the year, with no regional or other thresholds, and categories including arbitrary movement, track and field, combat, dexterous manipulation, production and construction, sports events, full-modal human-robot interaction, and health care. The "Yuchuang Plan" targets teams that have already established companies or have clear entrepreneurial directions, recruiting entrepreneurs with core technology, product capabilities or industry resources who are committed to promoting the commercial implementation of embodied intelligence.
Comments