Multi-theme resonance: a breakdown of the upward logic of the humanoid robot sector

Main body of the article EM_StockImg_Start

EM_StockImg_End

This week (October 19–October 23), humanoid robots became one of the most closely watched sectors in the market's hard-tech arena. Wind data shows that this week, the humanoid robot index rose 11.48%, the Unitree Robotics index rose 13.65%, and the embodied intelligence index rose 10.65%.

On October 23, humanoid robot concept stocks led the gains. Among them, Unitree Technology hit the "20CM" daily limit, with its share price at 624 yuan per share and total market value exceeding 250 billion yuan, with the stock up 46.14% this week; Zhaowei Electromechanical hit the limit up three times in five days, Daye Co. recorded six consecutive limit-up days, and Orbbec, Everwin Precision, DeepGlint, Leaderdrive and other humanoid robot component supply chain companies posted significant gains.

From the perspective of the industrial chain, the humanoid robot sector advanced on multiple fronts this week, covering every link from the robot's "body" to its "brain." The stocks involved include not only complete humanoid robot manufacturers, but also physical AI companies focused on developing robot "brains," as well as core component companies supporting the hardware body, spanning joint modules, reducers, vision systems and other sub-sectors.

The drivers of this round of gains in the humanoid robot sector and whether the rally can continue have become a market focus. A China Securities Journal reporter reviewed brokerage research reports and public industry information to break down the core logic of this rally from multiple dimensions.

Hardware side — mass production accelerates and empowers the industrial chain

Catalyzed by Tesla's mass production news

This week, news of mass production of Tesla's Optimus humanoid robot brought good news for the industry's hardware ecosystem. On October 22, a document Tesla filed with the U.S. Securities and Exchange Commission showed that the third-generation Optimus humanoid robot is ready for large-scale mass production, and the first production lines are under construction.

Domestic humanoid robot mass production speeds up

XPeng Robotics announced that it will achieve mass production of humanoid robots by the end of 2026 and begin deliveries in the first half of 2027. In September this year, AgiBot announced that its cumulative humanoid robot output had reached 20,000 units. In August, Unitree Technology announced that cumulative production of its bionic bipedal humanoid robots had reached about 18,000 units. Data from market research firm IDC shows that in the first half of 2026, global humanoid robot shipments reached nearly 25,000 units, up 432.1% year on year, of which shipments in the Chinese market exceeded 19,000 units.

Institutions believe that China's humanoid robot component supply chain is concentrated and its industrial base is solid, with mature supply systems in high-growth tracks such as motors, reducers and lead screws, and that it is expected to benefit from this round of humanoid robot development.

On the performance side, listed companies in the humanoid robot industrial chain have seen their business scale grow. Leadshine recently said in its investor relations activity record that its robot business has shown a rapid growth trend, and "the order scale for robot-related business for the full year is expected to be between 200 million and 300 million yuan." The company's frameless torque motors have more than 1 million units in orders on hand for the full year, with annual deliveries expected at 800,000 to 900,000 units; current monthly capacity is about 150,000 units, and it is expanding automated production lines toward an annual output of 3 million units. Financial reports show that in the first three quarters, Leadshine's net profit attributable to shareholders rose 86.80% year on year. In addition, Han's Laser's net profit attributable to shareholders in the first three quarters rose 152.19% year on year; Raycus Laser's net profit attributable to shareholders in the first three quarters rose 109.05% year on year.

The development of humanoid robots' "brain" capabilities was once a bottleneck for robot deployment and affected commercialization expectations. Recently, whether it is large-model vendors entering the humanoid robot space, leading complete-machine manufacturers releasing model progress, statements by prominent industry figures, or the explosive growth of the embodied data industry, all reflect predictable progress in the physical AI field and have strengthened the certainty of industry development.

Large-model vendors enter the field

General-purpose model vendors are joining the humanoid robot track. On October 17, OpenAI CEO Sam Altman said the company is entering the field to build humanoid robots, with a prototype demonstration expected in 2027, and its website publicly posted robot-related job openings. In addition, OpenAI released its new-generation flagship model GPT-6 Astra in September, which has been used by multiple humanoid robot companies for real-machine testing; they found that the model can perform task understanding and planning well.

Recently, another general-model giant, Google DeepMind, officially released the general physical reasoning model Gemini 4 Argon, which can perform functions such as robot control, simulation environment construction and training data quality inspection.

Practitioners believe that general-purpose models are expected to drive progress in embodied intelligence models and play a role in robot planning and decision-making, simulation training and other areas.

Wang Xingxing speaks again; complete-machine manufacturers release model progress

On October 22, Unitree Technology held its first 2026 World Physical AI Conference and Unitree Partner Ecosystem Conference. Soochow Securities believes that this conference marks leading body manufacturers' comprehensive transformation from single hardware manufacturing to "physical AI open source and brain ecosystem."

Unitree Technology founder Wang Xingxing reiterated at the conference the view he expressed at the 26th China International Industry Fair, saying that "faith in AI is not yet enough," and that with the release of GPT-6, physical AI has shown signs of being broken through, and the "ChatGPT moment" of embodied intelligence may arrive in the next one or two years, or two or three years. Two months later, the timeline Wang Xingxing gave has moved earlier; in August this year, at the 2026 World Robot Conference, his forecast was, "at the fastest 2 to 3 years, at the slowest 5 to 10 years."

In addition, Nvidia CEO Jensen Huang also said at this week's GTC conference that physical AI is the next wave of artificial intelligence.

English text is being prepared.

The embodied intelligence data industry is exploding

The embodied intelligence data industry is ushering in explosive growth, providing "nourishment" for the development of robot "brains." A reporter's research found that in the embodied intelligence data industry, leading robot manufacturers are working with local governments to establish embodied intelligence training grounds for real-machine data collection; many suppliers are also using bodyless collection devices for crowdsourced collection; and model vendors are providing simulation data. Multiple technical routes are proceeding in parallel, collecting data at multiple tiers. Practitioners say industry demand for embodied intelligence data is evolving from millions of hours to tens of millions of hours.

Policy continues to catalyze

At a State Council Information Office press conference held on October 19, Wei Wei, director of the Science and Technology Department of the Ministry of Industry and Information Technology, said that in the first half of this year, China's humanoid robot complete-machine products exceeded 400 models, with output exceeding 40,000 units, and full-year output is expected to exceed 100,000 units. Embodied intelligence products represented by humanoid robots are accelerating their exit from laboratories and gradually entering real-world scenarios such as commercial districts and stores. The output data boosts industry confidence.

On the policy front, the "Shanghai Action to Accelerate the Development of a New Generation of Intelligent Manufacturing" was officially launched in October, specifying that by 2028 it will promote 8,000 humanoid robots into factories, with industrial robot density reaching 600 units per 10,000 people.

Guoyuan Securities believes that the clear factory-entry target is expected to provide large-scale application traction for complete-machine enterprises, promote their connection with manufacturing customers, and accelerate the transition of products from prototype validation to batch deployment; at the same time, the industrial environment can continuously test robots' operational capabilities in tasks such as handling, loading and unloading, and inspection, accumulate real interaction data, and drive a feedback loop between model training and product iteration.

In addition, the State Administration for Market Regulation recently approved and released the first batch of six national standards in the humanoid robot field, covering environmental perception, decision planning, motion control, task operation, and simulation test platform specifications, among others. They will take effect on April 1, 2027, helping the industry develop in a standardized way.

Can the sector's heat last?

Whether the heat in the humanoid robot sector can last depends on real progress in industrial deployment. Brokerages and investors offer judgments from different angles on the development of the humanoid robot industry.

Soochow Securities believes that, based on the performance of many robot companies at the China International Industry Fair, humanoid robots already have preliminary commercial application capabilities. With the continuous accumulation of practical application data and physical AI opening a new paradigm for robot training, the data flywheel is expected to accelerate, and improvements in humanoid robot intelligence are expected to achieve nonlinear growth.

Guoyuan Securities believes that the evaluation focus for humanoid robots entering factories will gradually shift from single-machine function demonstrations to continuous operation stability, task success rate, deployment efficiency, and unit operating cost. As demonstration applications expand, companies able to form standardized solutions and achieve cross-scenario replication are expected to benefit first.

A Wanlian Securities research report said that the humanoid robot industry is currently at the dawn moment of moving from technological breakthroughs to large-scale commercialization. With policy and capital working together, and AI large models continuously injecting a soul into robots, humanoid robots are expected to form an emerging industry. And 2026 will be a key window for mass-production validation and scenario deployment.

Pacific Securities, meanwhile, said that the large-scale deployment of humanoid robots is still in its early stages. On the progress of the "brain," embodied large models worldwide are in an exploration stage of rapid technological development, and the industry has not yet formed a unified, mature technical consensus or competitive landscape. Overall, the humanoid robot industry's technical routes have not yet converged, and its key core technologies still need further breakthroughs.

Zhang Jiakang, an investor in the embodied intelligence track and founder of Gengxin Investment (Shenzhen) Co., Ltd., told the reporter, "This week's heat looks more like an emotional repair after an oversold decline. Before this, the humanoid robot sector had already pulled back sharply from its July high, and valuations have returned to a relatively low level over the past two years. The heat in humanoid robot-related sectors is expected to last for a long time, with short-term fluctuations."

Zhang Jiakang believes that the subsequent trend of the humanoid robot sector will diverge, "The market will recognize concepts that can be verified." He said that the key point for the next stage of the humanoid robot industry is for products to achieve 1,000 hours of failure-free operation, with supporting spare parts supply and localized services. Companies that can achieve a profitable logic and a commercial closed loop are expected to embark on an independent market trend; targets that rely only on conceptual narratives will gradually be cleared out.

(Source: China Securities Journal) Middle of the main text Embedded advertisement