Orient Securities has released a research report noting that the recent opening of the second World Humanoid Robot Games could drive continuous iteration and advancement in embodied intelligence models across the industry. The firm believes this competitive platform may accelerate the transition of embodied intelligence from laboratory settings into real-world application scenarios. Looking ahead, the report suggests that embodied intelligence companies actively positioning in vertical use cases, along with integrated manufacturers demonstrating stronger model generalization capabilities, are likely to secure greater competitive advantages.
According to the report, the second World Humanoid Robot Games officially opened on the evening of August 22, 2026, at the National Speed Skating Oval, also known as the "Ice Ribbon." The five-day event features 51 competition events across athletic contests and scenario-based challenges. New athletic events include tug-of-war, pitch-pot, and weightlifting, while scenario-based competitions have added tasks such as clothing organization in household settings, firefighting operations in emergency rescue scenarios, and cooked food preparation in retail environments. Participating robots are expected to demonstrate greater autonomy, dexterity, and practicality.
At the opening ceremony, the world's first fully autonomous tennis-playing robot was unveiled, engaging in the inaugural human-machine tennis exhibition match alongside renowned tennis player Zheng Jie. During three rounds of serve-and-return play, the robot consistently completed over 20 shots per exchange in each round.
The humanoid robot's fully autonomous tennis play highlights a breakthrough in embodied intelligence large-model applications. According to Galaxy General, the tennis demonstration utilized its embodied intelligence large model, the AstraBrain. The technical framework integrates task comprehension and decision-making with body movement control and neural coordination. The "brain" component handles assessing incoming balls and match situations, selecting shot types and target placements, while the "cerebellum" component translates these decisions into running, balance, and swing actions. Notably, the training data for this robot did not rely on expensive or difficult-to-obtain precision parameters from professional athletes. Instead, Galaxy General employed its proprietary "Galaxy Star Data" foundation, allowing the robot to continuously train and compete against agents of varying skill levels in virtual environments, completing tens of millions of learning and evolution cycles.
From the demonstrated results, Zheng Jie expressed being "deeply impressed" by the humanoid robot's performance. Orient Securities believes this achievement reflects continued breakthroughs in applying embodied intelligence models to vertical scenarios while also showcasing the immense potential of humanoid robots to enhance productivity. In certain domains, these robots have already achieved basic commercial viability, which could strengthen market investor confidence in the industry.
Beyond the tennis scenario, the report highlights that this year's World Humanoid Robot Games added numerous competition events, including tug-of-war, pitch-pot, weightlifting, clothing organization, and emergency rescue tasks. Previously, some market investors expressed concerns about the lack of unified benchmarks for comparing embodied intelligence models, making it difficult to assess capability differences among humanoid robots. Orient Securities argues that the competitive nature of the Humanoid Robot Games is likely to drive ongoing iterative improvements in industry embodied intelligence models and accelerate the movement of embodied intelligence from laboratories toward practical application scenarios.
Going forward, the firm maintains that embodied intelligence companies actively establishing footprints in vertical use cases, along with integrated manufacturers possessing stronger model generalization capabilities, are positioned to achieve greater competitive advantages.
Risk warnings include: manufacturer production falling short of expectations, scenario demand realization remaining unclear leading to underperformance, national policy changes slowing industry development, industry financing falling short of expectations, model development and data collection progressing slower than anticipated, order execution results underperforming expectations, and product price reduction risks.