China's Humanoid Industry Explodes

China's government is a significant catalyst in the humanoid robotics surge, providing over $20 billion in subsidies through grants, loans, and tax credits in late 2024 and early 2025 to bolster the industry. The Ministry of Industry and Information Technology (MIIT) is targeting the establishment of a comprehensive innovation system for humanoid robots, aiming for two to three globally influential companies by 2025 and a secure industrial supply chain by 2027. To accelerate development, cities like Beijing have launched large-scale training bases to generate millions of data points annually, addressing the data scarcity that can hinder AI progress. A key strategic move is the recent release of a national standard system for the humanoid robot industry, a top-level design covering the entire industrial chain from components to applications and safety ethics. This framework, developed by the MIIT's technical committee with input from over 120 institutions and companies, aims to unify technical specifications and promote component modularization to reduce costs and enhance supply chain efficiency. The goal is to shift the industry from scattered technical exploration to a new phase of collaborative and standardized growth. At the heart of the technical advancements are embodied AI and large-scale models. Companies are moving beyond simple hardware and focusing on the "brain" and "cerebellum" of the robots. For instance, Xiaomi's CyberOne humanoid leverages a "Mi-Sense" depth vision module combined with AI interaction algorithms to perceive 3D space and recognize human gestures and emotions. Similarly, Agibot has developed a "One Robotic Body, Three Intelligences" architecture that integrates interaction, manipulation, and locomotion into a unified platform. Reinforcement learning is a critical technology being deployed for motion control, enabling more dynamic and human-like movements. Companies like Unitree and LimX Dynamics are actively using reinforcement learning to train their robots for complex tasks like running and maintaining balance. Unitree even offers training programs focused on reinforcement learning for its G1 humanoid, utilizing Isaac Lab for simulation and sim-to-real deployment. To fuel this rapid expansion, a strong talent pipeline is being built through close collaboration between universities and corporations. Recognizing a shortage of engineers with interdisciplinary skills, top institutions like Shanghai Jiao Tong University and Zhejiang University are launching the first undergraduate majors in "embodied intelligence." These programs are designed to produce "full stack" engineers with expertise in AI, mechanical engineering, and computer science to meet the high demand from tech giants and robotics firms. For software and embedded systems engineers, the most sought-after skills in China's robotics sector include proficiency in C++ and Python, along with experience in the Robot Operating System (ROS). There is a particularly high demand for algorithm engineers with expertise in machine learning, deep learning, and reinforcement learning. Companies are looking for individuals who can not only develop algorithms but also translate them into practical, real-world applications. The supply chain for humanoid robots in China is maturing rapidly, with a significant domestic capacity for core components. This vertical integration allows for cost advantages of 50% to 70% for major components compared to foreign counterparts, enabling more competitive pricing for the final products. This robust manufacturing ecosystem, originally developed for industries like smartphones and electric vehicles, now supports the large-scale production of actuators, sensors, and AI chips for humanoid robots. Leading the charge in commercialization are companies like Agibot, which shipped over 5,100 units in 2025, and Unitree Robotics. These companies are not just focusing on domestic sales but are actively expanding globally, with some deriving over 50% of their business from overseas operations. The application of these robots is also broadening, with deployments in industrial manufacturing, logistics, and even automotive assembly lines, as seen with UBTECH's Walker S2.