The Zhitong Finance App learned that Orient Securities released a research report saying that the 2nd World Humanoid Robot Games recently officially opened. The bank believes that the robot competition in the humanoid robot games is expected to promote the continuous iterative progress of the industry's personalized intelligence models, which is expected to accelerate the transformation of embodied intelligence from laboratories to actual application scenarios. Looking ahead, the bank believes that intelligent manufacturers with active layouts in vertical scenarios and machine manufacturers with stronger model generalization capabilities are expected to gain a greater competitive advantage.
Orient Securities's main views are as follows:
Incident: On the evening of August 22, 2026, the 2nd World Humanoid Robot Games officially opened at the National Speed Skating Hall “Ice Ribbon”
This year's games will last for 5 days, including competition and scene matches, with a total of 51 events. Events such as tug-of-war, pot throwing, and weightlifting have been added to the competition, and clothing organization in the home scene, fire extinguishing operations in emergency rescue scenes, and cooked food preparation in retail scenes. The participating robots will be “more autonomous,” “more agile,” and “more practical.” At the opening scene, the world's first fully autonomous tennis robot was unveiled, and the first man-robot tennis competition was staged with famous tennis player Zheng Jie. In the three-round kickoff round, the robot hit the ball more than 20 times in each round.
The humanoid robot showcased fully autonomous tennis matches, showing a breakthrough in the application of a large intelligent model
According to Galaxy GM, this tennis display uses its large intelligent model, Galaxy Star Brain, AstraBrain. Its technical system combines task understanding with decision making, physical movement control, and neural control: among them, the “brain” part is responsible for judging incoming and game situations, and selecting batting methods and landing points; the “cerebellum” part is responsible for transforming relevant decisions into running, balancing, and swinging movements. During the training process, the data used by the robot did not rely on expensive and difficult to obtain accurate parameters from professional athletes. Instead, through the “Galaxy Star Number” data base developed by Galaxy GM, the robot continued to practice and play games with different levels of intelligence in the virtual world to complete tens of millions of learning and evolution times. Judging from the actual display results, Zheng Jie said that her performance on humanoid robots was “very shocking.” The bank believes that on the one hand, it reflects the continuous breakthrough of the application of physical intelligence models in vertical scenarios. On the other hand, it also shows the huge potential of humanoid robots in improving productivity. They already have basic commercial value in some fields, which is expected to enhance market investors' confidence in the industry.
Embodied intelligence accelerates from laboratories to actual application scenarios, and related machine manufacturers are expected to fully benefit
In addition to the tennis scene, the tour saw that many events such as tug-of-war, bowling, weightlifting, clothing organization, and emergency rescue were also added to this World Humanoid Robot Games. Previously, some investors in the market worried about the lack of a uniform comparison benchmark between embodied intelligent models, making it difficult to reflect the capability gap between humanoid robots. The bank believes that the robot competition at the humanoid robot competition is expected to promote the continuous iterative progress of the industry's embodied intelligence models, and is expected to accelerate the transformation of embodied intelligence from the laboratory to actual application scenarios. Looking ahead, the bank believes that intelligent manufacturers with active layouts in vertical scenarios and machine manufacturers with stronger model generalization capabilities are expected to gain a greater competitive advantage.
Risk Alerts
Manufacturers' production falls short of expectations, unclear implementation of scenario requirements results in lower than expectations, industry development slows down due to changes in national policies, industry financing falls short of expectations, model development and data collection are slower than expected, order execution results fall short of expectations, and product price reduction risks.