Chinese electric vehicle manufacturer XPeng has announced that its robotics affiliate raised over $900 million in its first major institutional financing round. The investment values the robotics business at more than $6.3 billion post-money, representing one of the largest single private capital raises in the embodied AI sector to date.
The round was led by IDG Capital and Gaorong Ventures, with participation from strategic tech conglomerates Tencent and Alibaba alongside parent firm XPeng Inc.
Capital Allocation for Physical AI
According to statements from XPeng, the newly secured proceeds will be allocated across five core operational priorities:
- Training and scaling physical AI foundation models and multimodal vision-language-action (VLA) policies.
- Collecting high-fidelity real-world and synthetic manipulation data across industrial and commercial environments.
- Developing custom electromechanical actuators, sensor suites, and embedded compute controllers.
- Constructing dedicated mass-production facilities with end-to-end quality validation pipelines.
- Establishing international commercialization and distribution infrastructure.
XPeng CEO He Xiaopeng, who took direct leadership of the robotics division in June, noted that the convergence of automotive manufacturing expertise and physical AI architecture provides key supply-chain advantages for hardware scaling.
Roadmap for Humanoid Mass Production
The capital infusion accelerates deployment timelines for the company's flagship bipedal humanoid robot, XPeng IRON. XPeng plans to initiate trial assembly and pilot deployments by the end of 2026, placing units in retail showrooms and automotive assembly plants to perform repetitive logistics and material handling tasks.
Commercial availability and customer deliveries across enterprise and international markets are targeted for 2027.
The participation of Tencent and Alibaba underscores expanding strategic interest among Chinese technology conglomerates in embodied AI platforms, following recent capital flows into robotic hardware and real-world intelligence foundation models.
