Xpeng's robotics division closed a $900 million financing round at a post-money valuation exceeding $6.3 billion, with IDG Capital leading and Tencent and Alibaba participating. Xpeng founder He Xiaopeng and co-president Brian Gu personally invested approximately $100 million, underscoring the unit's strategic importance.
The round reflects a deliberate capital allocation decision: He Xiaopeng has cited razor-thin profit margins in automotive manufacturing as the rationale for pivoting into robotics, where gross margins and pricing power are different. The company is banking on humanoid robots—specifically its Iron model—as a more defensible business than selling cars.
Michael Dunne, CEO of advisory firm Dunne Insights, sees Xpeng as the Chinese automaker most rigorously studying Tesla's playbook. "They have all the hardware to get the job done," Dunne said. The critical unknown: whether Chinese automakers can match Tesla's artificial intelligence capabilities, which would determine whether robotics becomes a genuine profit engine or another crowded hardware play.
Xpeng is not alone. Chery's robotics unit, AiMOGA, is preparing for an initial public offering this month. BYD unveiled its Xiao Di humanoid robot. Changan, GAC, Li Auto, SAIC and Seres are all developing humanoid robots for commercial deployment. The collective move signals Chinese automakers are treating robotics as a viable escape route from the automotive margin squeeze.
Globally, Agility Robotics, Apptronik and Figure are racing to deploy humanoid robots at commercial scale. Hyundai-owned Boston Dynamics is further along: the company plans to introduce its Atlas humanoid to its Georgia factory this year, with deployment for parts sequencing targeted by 2028. Hyundai has partnered with Google's DeepMind to accelerate development and is opening a U.S. Robot Metaplant Application Center to train Atlas on complex industrial movements.
The investment wave rests on two technical bets: that recent advances in robot hardware are mature enough for industrial work, and that large language model techniques can enable humanoid robots to learn arbitrary tasks without task-specific programming. If both hold, humanoid robots could unlock substantial new revenue streams for hardware-heavy manufacturers. If either falters, capital will have flowed into the wrong boxes.
