Focus shifts to putting robots to work

A Chinese robotics company, Joyful Embodied, is leading a shift in the industry by prioritizing data over hardware, building a 3,000-square-meter training facility in Fujian province. The facility will use humanoid, quadruped, and wheeled robots performing real-world tasks—such as stacking cups and sorting objects—while human operators in VR gear guide their movements. Cameras and sensors will capture every detail, creating high-quality training data for embodied AI systems, which Joyful Embodied CEO Chen Yishi calls the 'fuel' for next-generation robots. The company, founded in September 2025, positions itself as a full-stack embodied AI developer, combining data collection with its own large AI models and a platform called Joyful Studio. This platform aims to enable customized robot applications in manufacturing, security, logistics, education, and service industries. Chen stated that each continuously operating robot in the facility will require three technical roles for data annotation, algorithm optimization, and maintenance. Fujian, a manufacturing hub, is aggressively expanding into robotics and AI infrastructure, supported by China’s 2026 Government Work Report, which designated embodied AI as a strategic industry. The Ministry of Industry and Information Technology recently approved the country’s first industry standard for embodied AI benchmarking, set to take effect, formalizing the sector’s development. While investors have focused on humanoid robot hardware, Joyful Embodied argues that high-quality training data—particularly teleoperation and motion-capture data—is the industry’s scarcest resource. Chen believes the economics of the sector will increasingly revolve around data, with high-quality datasets priced by the hour. The company also plans to leverage Fujian’s export-oriented economy and overseas Chinese networks to expand robot deployment globally. For now, the competition remains domestic, with China accelerating its push into embodied AI. Joyful Embodied’s facility reflects this shift, moving beyond flashy demonstrations to practical, data-driven robot training. The company’s CEO emphasized that the real challenge is no longer about robots walking or dancing but about which system can learn and adapt most effectively through real-world interaction.