Humanoid robot sector sees wave of limit-up stocks; JPMorgan upgrades Unitree rating to Overweight

Over the weekend, physics AI research reports emerged in dense succession, JPMorgan sharply upgraded Unitree Robotics' rating, and Qiyuan Robotics' mass-production milestones landed one after another—with these multiple catalysts stacking up, the humanoid robot sector set off a wave of limit-up moves on the afternoon of October 26.

On Monday afternoon, humanoid robot concept stocks continued to climb! Shangwei New Materials surged to the limit-up, at 151.57 yuan per share, up 20%, with total market capitalization surpassing 61.1 billion yuan; Unitree Robotics rose more than 18% intraday, with its share price breaking through the 700 yuan mark.

In addition, Fenglong Co. and Dafeng Industrial sealed limit-up in the afternoon, while Molding Technology, Jack Technology and other stocks hit the limit-up. In Hong Kong stocks, Huancheng Technology rose more than 16%, Wanan Robotics rose nearly 13%, Yuejiang rose about 9%, and Hesai-W rose more than 8%, as the physics AI concept continued its strength.

The core driver of this rally lies in the market's expectation that the breakthrough time window for physics AI has clearly moved forward. Unitree Robotics founder Wang Xingxing publicly stated that physics AI "has shown signs of being broken through," and that the breakthrough window may fall in "the next 1–2 years or 2–3 years," a substantial advance from the previous judgment that it "may take up to 10 years." At the same time, Qiyuan Robotics' 5,000th personal robot officially rolled off the production line, with mass-production capacity accelerating, further strengthening market confidence in industrialization.

According to an article in Wallstreetcn, foreign institutions including JPMorgan and Nomura concentrated on publishing related research reports over the weekend, with domestic sell-side firms following suit, and market discussion of physics AI heating up markedly. Overall, the biggest change is that the market has begun to look at the breakthrough timing for physics AI earlier, while the industry is re-searching for cheaper ways to realize it.

JPMorgan's shift this time is particularly key. After previously visiting the China International Industry Fair, the bank remained cautious about the commercialization progress of domestic robot makers, believing that hardware leadership had not yet stably translated into repeatable commercial demand, with orders mainly coming from R&D, developer platforms and small-batch pilots. But after the 2026 World Physics AI Conference and Partner Ecosystem Conference held in Hangzhou, the bank adjusted its judgment.

In its research report, JPMorgan cited Wang Xingxing's statement that the physics AI breakthrough window may fall in "the next 1–2 years or 2–3 years," and therefore believes the industry's sales S-curve may move forward. Analysts pointed out that a shortened timetable would raise the strategic value of deployed robots, developer ecosystems and real-world data, and that the industry bottleneck is shifting from high-level reasoning capability to "real-world execution, data generation and deployment iteration."

Nomura, meanwhile, published a report over the weekend introducing the SuperMimic technical path—a method that first learns a model describing the robot's own motion patterns, then trains a controller on that basis. According to Nomura's account, after three hours of training on a single H100, its performance exceeded the previously released SONIC model checkpoint, whose training consumed about 21,000 GPU hours, a reduction of about 1,000 times in computation for training specific motion control tasks. Nomura also cautioned that this figure applies only to specific motion control tasks, cannot be directly extended to the entire robot's full computing demand, and has not yet solved object manipulation and complex industrial tasks.

JPMorgan upgrades Unitree's rating directly from "underweight" to "overweight," raises target price to 995 yuan

Building on the above shift in judgment, according to Chase Trading Desk, JPMorgan published a research report on October 25, upgrading Unitree Robotics' rating directly by two notches from "underweight" to "overweight," and sharply raising its December 2027 target price from 300 yuan to 995 yuan, about 59% above the October 23 closing price of 624 yuan.

On sales forecasts, JPMorgan did not raise its assumption for Unitree's 2026 humanoid robot sales, which remains at about 14,200 units; but it sharply raised its 2028 forecast from about 41,600 units to about 111,600 units. Correspondingly, its global market share forecast is raised from 8% to 21% for 2028, and from 4% to 12% for 2030. The bank expects Unitree's revenue from 2026 to 2028 to be 2.819 billion yuan, 4.888 billion yuan and 13.383 billion yuan respectively, with 2028 revenue growth forecast at 173.8% year on year.

On valuation logic, JPMorgan also incorporated "scarcity" into its framework. The bank believes that as reviews of robot companies' IPOs and spin-off listings pay greater attention to commercialization quality and business model durability, new listing processes for embodied intelligence companies may be slower than previously expected. The report accordingly holds that Unitree is "one of the few direct embodied AI investment mappings" in the A-share market, and that its already-listed status, hardware platform and installed base form the basis for a "scarcity premium." On valuation methodology, JPMorgan uses a 30x forecast 2028 price-to-sales ratio, providing the basis for its 995 yuan target price.

Qiyuan Robotics' 5,000th unit rolls off the line, mass-production capacity accelerating

Milestone events on the industrial side also provided support for the sector. According to a Shangguan News report, Qiyuan Robotics' 5,000th personal robot officially rolled off the production line; from the September 20 launch of the Qiyuan Q1 and Qiyuan T1 products to the 5,000th unit rolling off the line one month later, mass-production delivery capacity is accelerating.

Gao Quanzhong, co-president of Qiyuan Robotics and president of its central R&D department, said at the mass-production rollout site that "from 0 to 5,000 units in stable mass production took only one month," and that the production line now rolls off one robot every 2.5 minutes, with daily capacity raised from more than 10 units previously to more than 200 units.

On user data, Gao Quanzhong disclosed that some users who have received their products use them for more than 1 hour per day on average, with heavy users averaging more than 3 hours per day and more than 150 effective interactions per day on average.

In addition, on October 24, Qiyuan Robotics launched the "Physical AI Qizai Program," opening its robot development platform, supporting tools and computing power support to developers, soliciting embodied intelligence skills, interactive personas and scenario solutions, and the developer community official website has also gone live simultaneously.

Soochow Securities: Development focus shifting from hardware to the model side, software-hardware integration expected to lead the world

In an industry commentary report released on October 24, Soochow Securities systematically reviewed the industrial logic behind this round of humanoid robot market momentum.

Soochow Securities noted that the development focus of the humanoid robot industry is shifting from upgrades and iterations on the hardware side to accelerated evolution on the model side. On October 22, 2026, Unitree Robotics held the 2026 World Physical AI Conference, marking a comprehensive transformation by leading body manufacturers from single hardware manufacturing to "physical AI open source and brain ecosystem"; OpenAI has entered the fray in a big way, building a robot hardware team and a real-machine data collection system, aiming to overcome the generalization bottleneck of Sim-to-Real.

On the model route, Soochow Securities noted that the humanoid robot model-side development route has shifted from VLA (vision-language-action model) to WLA (world-language-action model). The accelerated evolution of physical AI is unblocking the data flywheel of "hardware deployment—real physical interaction feedback—model training," and high-value physical feedback data will drive robot foundation models across the Scaling Law threshold, pushing the industrialization of humanoid robots to the 1-10 development inflection point.

At the supply chain level, Soochow Securities emphasized that China's humanoid robot model-side layout is relatively early, with significant first-mover advantages in data accumulation. Combined with a mature and stable component supply system, it is expected to fully leverage the advantages of software-hardware integration in this round of the humanoid robot development cycle and lead the development and progress of the global humanoid robot industry. The institution specifically cited Unitree Robotics' self-developed high-explosion generator, the reducer mass-production capabilities of Leaderdrive and Zhongda Leader, and the supply systems of Hengli Hydraulic and Zhejiang Rongtai in the lead screw field. Risk Disclosure and Disclaimer

Markets carry risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial conditions or needs of individual users. Users should consider whether any opinions, views or conclusions in this article are suitable for their particular circumstances. Investment based on this is at the user's own risk.