Spirit AI Becomes China's Newest AI Unicorn

The funding round for Spirit AI was a strategic assembly of capital, combining top-tier financial institutions like Yunfeng Capital and HongShan with state-backed funds from Chongqing and Hangzhou. The participation of industrial giants such as TCL Capital, Huawei, Xiaomi, and CATL provides Spirit AI with not just capital, but also critical real-world deployment scenarios and data streams for its embodied AI systems. Spirit AI was founded in January 2024 by former Rokae Robotics CTO Han Fengtao and Tsinghua University assistant professor Gao Yang. The company's core technology is its proprietary Vision-Language-Action (VLA) model, Spirit v1.5, an end-to-end system integrating perception, reasoning, and action that was open-sourced in January. This model powers the Moz1 humanoid, a full-body force-controlled robot with 26 degrees of freedom. At CATL's Zhongzhou facility, the Moz robot is already deployed on battery production lines, handling end-of-line quality control tests. The robot performs tasks like inserting high-voltage plugs for resistance testing, a critical step that carries spark risks for human workers. CATL reports Moz has achieved a connection success rate of over 99% while matching the throughput of skilled human operators. This investment is part of a larger frenzy in China's embodied intelligence sector, which saw cumulative funding surpass 230 billion RMB since early 2023. In February 2026 alone, at least six Chinese startups in the field raised over $100 million each. This rapid expansion has prompted warnings from China's National Development and Reform Commission about the risk of "duplicate projects" and overheating in the sector. Orchestrating the complex behaviors of such robots relies on multi-agent frameworks. Open-source projects like LangGraph and CrewAI are becoming central to building stateful, multi-agent systems. LangGraph, part of the LangChain ecosystem, uses graph-based architecture to define control flows between different agents and tools, enabling more reliable and complex workflows. However, scaling humanoid robots from pilots to mass deployment faces severe technical barriers. Gartner predicts fewer than 100 companies will move humanoids beyond the proof-of-concept stage in the near term due to limitations in current technology. The most critical bottleneck is battery life, with most humanoids operating for only 90 minutes to 4 hours, far short of the 8-20 hours required for an industrial shift. For a CTO at a rapidly scaling startup, the primary challenge shifts from technical execution to leadership and team architecture. Growing an engineering team beyond 20-30 people necessitates introducing new leadership layers, such as dedicated engineering managers and tech leads, to avoid the CTO becoming a bottleneck. The focus transitions from writing code to building scalable processes for hiring, delegation, and aligning the tech strategy with business goals. While industrial use cases are advancing, consumer-facing embodied AI is also surging in China. E-commerce platform JD.com reported that sales revenue of embodied AI robots, including educational and companion bots, grew over 700% in late 2025 compared to the previous year, signaling growing consumer acceptance of AI agents in daily life.