Waypoint
Back to news
Industry

Xiaomi loses its L3 driving chief to an embodied AI startup as auto talent drain widens

Wang Naiyan, a nine-year TuSimple veteran who joined Xiaomi in May 2024, is pitching investors a venture built around robot foundation models — following similar exits at XPeng and Li Auto, while Nio is funding its own driving chief's side venture.

2026-09-08

Xiaomi loses its L3 driving chief to an embodied AI startup as auto talent drain widens

Wikimedia Commons

Wang Naiyan, the executive leading Xiaomi Auto's L3 autonomous driving program, has left the company to found a physical AI venture, according to reports from 36Kr and 21jingji. The startup is focused on "robot brains" and is currently raising funds, with a team of more than 10 core researchers that includes several of Wang's former TuSimple colleagues. He holds a doctorate from the Hong Kong University of Science and Technology and spent nine years at TuSimple, serving as China CTO from 2019, before joining Xiaomi in May 2024.

At Xiaomi, Wang reported to Ye Hangjun, head of the autonomous driving and robotics division, and contributed to L3 development, including a self-driving demonstration staged at the Nürburgring Nordschleife in June. He is now the second of four core members of Xiaomi's autonomous driving team to leave and start a company, after Chen Long.

The exit fits a broader migration of Chinese carmakers' self-driving talent into robotics. Former XPeng driving chief Li Liyun and ex-vice president Chen Yonghai have joined EngineAI Robotics; former Li Auto chief AI scientist Chen Wei founded Xieyue Intelligent, targeting home-use products; and Li Auto's former AD head Lang Xianpeng co-founded Kunlunxing Robotics with ex-Huawei executive Ren Geng.

Nio has taken a different route: CEO William Li confirmed on the September 1 earnings call that driving chief Ren Shaoqing will launch a physical AI company with Nio as a strategic shareholder while keeping his existing role. Industry sources caution that the leap may not be easy — robots face tougher motion-control demands, and data collection and simulation training are harder than in automotive.

This is an original summary compiled and translated from the source below, not a direct translation of it.

🔗 Source: CnEVPost