Unitree Shares Surge as Investors Bet Chatbot AI Can Give Robots a ‘Brain’
Key Takeaways
- •Unitree's stock has climbed about two-thirds since Oct. 15, closing at 672 yuan on Monday and valuing the company at roughly $40.8 billion.
- •Founder Wang Xingxing said large AI models could achieve a breakthrough in physical AI within one to three years, and claimed OpenAI's GPT-6 Astra can already direct robots with zero training.
- •JPMorgan upgraded Unitree to overweight from underweight and set a 995 yuan price target, about 48% above Monday's close, after revising its assessment of the company's software strength.
- •Unitree listed on Shanghai's STAR Market on Aug. 19 and sells its G1 humanoid for roughly $16,000, making it cheaper than competing humanoids.
- •Researchers have found security flaws in Unitree robots, including audio, video, and spatial data being streamed to external servers without user notification.

Shares of Chinese robot maker Unitree Robotics have risen about two-thirds since Oct. 15, closing at 672 yuan ($100.75) on Monday and lifting the company's market value to 271.8 billion yuan, or roughly $40.8 billion.
The rally reflects investor expectations that artificial intelligence systems developed for chatbots could soon operate machines in the physical world, a field known as physical AI. If that technology develops as expected, robots capable of tidying rooms or folding laundry on command could move beyond laboratory demonstrations, potentially benefiting the companies that manufacture their bodies, motors, and joints. The pace of the move shows how quickly markets can reprice hardware makers on software expectations alone, before the underlying technology is proven outside the lab.
Unitree founder Wang Xingxing helped drive the latest gains at the World Physical AI Conference in Hangzhou on Oct. 22. The event was the company's first partner conference in its decade of operation. Wang said large models—chatbot-style AI systems trained on vast amounts of data—could achieve a breakthrough in physical AI within the next few years.
“If everything goes well, it will take one to two years, otherwise it will take two to three years,” Wang said.
Wang also said that OpenAI's GPT-6 Astra can already direct robots with zero training. His broader forecast at the conference was that large models would break through in physical AI within one to three years. That timeline gives investors a concrete benchmark for judging whether the physical AI thesis holds.
Unitree shares reached the 20% daily limit on Friday, the maximum amount by which a stock can rise in a single session, before closing at 624 yuan ($93.55). The stock gained another 7.69% on Monday to finish at 672 yuan.
JPMorgan upgraded Unitree to overweight from underweight on Monday and raised its price target to 995 yuan ($149.17). In analyst terminology, an overweight rating indicates that the bank recommends holding more of a stock than its market benchmark. JPMorgan had previously argued that Unitree relied too heavily on hardware and was comparatively weak in software. The bank changed its view after hearing from Wang and other executives and reviewing the company's humanoid lineup in Hangzhou. The new target sits roughly 48% above Monday's closing price of 672 yuan.
Unitree listed on the Shanghai Stock Exchange's STAR Market, a technology-focused board, on Aug. 19. Its G1 humanoid robot costs roughly $16,000, making it cheaper than rival humanoids. Researchers have also identified security flaws, including the streaming of audio, video, and spatial data to outside servers without users being notified. Even so, the company's approach has attracted enough interest to drive its valuation higher. Security questions like these are one of the practical gaps between today's laboratory demonstrations and the command-driven household vision investors are pricing in.
Unitree already sells robots. Tesla has only recently started producing Optimus robots at its factory in Fremont, California, and expects outside deliveries around the end of 2027. Figure AI partnered with OpenAI on its humanoid robots in 2024, using OpenAI's models to power the machines. Those staggered commercial timelines—Unitree shipping hardware today, Tesla targeting outside deliveries in late 2027—give the emerging industry a set of concrete checkpoints, alongside Wang's one-to-three-year forecast, for gauging how quickly robots acquire a working brain.