Spirit AI Raises Over $540M, Deploys Humanoids in Factories
Chinese humanoid robotics startup Spirit AI has reportedly closed two funding rounds totaling over $540 million, pushing its valuation past the $1 billion mark. The company secured a $290.5 million round and is said to have raised another $250 million to accelerate global expansion and embodied intelligence research. Spirit AI's Moz humanoid robots are already operating on production lines at battery manufacturer CATL, signaling a move from pilots to industrial-scale deployment in China.
The funding round for Spirit AI attracted a diverse group of heavyweight investors, including financial capital from Yunfeng Capital and HongShan, strategic industry players like TCL Capital, and state-backed funds from Chongqing and Hangzhou. Existing shareholders such as ShunWei Capital and Prosperity7 also increased their investment, signaling strong confidence in the company's trajectory. The company was co-founded by Han Fengtao, the former CTO of Rokae Robotics, who brings extensive experience in mass-producing tens of thousands of robots, and Gao Yang, a PhD from UC Berkeley and an assistant professor at Tsinghua University, who provides a deep academic background in artificial intelligence. This blend of practical hardware expertise and advanced AI research is a key driver of their rapid progress. At CATL's Zhongzhou facility, the "Moz" (also called "Xiaomo") humanoid robots are performing end-of-line (EOL) and direct current internal resistance (DCR) testing on battery packs. This involves plugging high-voltage connectors into the batteries, a task that poses high-voltage spark risks for human workers and requires high consistency. The Moz robot operates with a 99% success rate and its daily workload is reportedly triple that of a human worker because it can operate without breaks. The robot utilizes a Vision-Language-Action (VLA) model, allowing it to adapt to variations in component positioning and dynamically control the force needed to handle flexible wiring harnesses. This deployment is a significant step for "embodied AI," where intelligence is integrated into a physical system to learn and interact with the environment. Spirit AI's strategy focuses on training its models with vast amounts of "dirty data"—unstructured, real-world interaction data—to build more robust and generalizable AI. The company has already amassed over 200,000 hours of this data, aiming for 1 million by the end of 2026. The broader Chinese humanoid robotics sector is experiencing a surge in investment, with a focus shifting from pure hardware to the AI models and operating systems that serve as the robots' "brains." Competitors include well-established players like UBTECH and Unitree Robotics, alongside a new wave of startups such as AgiBot and LimX Dynamics, all vying for dominance in this emerging market.
