Cao Xudong

Cao Xudong is the founder and leader of Momenta. In 132. 对星海图创始人高继扬的3小时访谈：鲶鱼、曾国藩、Waymo与Momenta的两面、一只狼与许华哲的离开, Gao Jiyang describes him as technically strong, decisive, AI-oriented, and long-term in his commitment to autonomous driving.

Momenta IPO后再访曹旭东：就是想做没有尽头的AI adds Cao’s own account of his method. He traces his path from aerospace-related study and statistical physics into statistical learning, then through Microsoft Research Asia, Sun Jian’s experiment culture, SenseTime product delivery, and finally Momenta’s Autonomous Driving Data Flywheel. The interview presents him as a founder who combines long-horizon AI ambition with a practical insistence on customer value, mainline reuse, and Low-Cost Short-Cycle Validation.

Key Points

• Cao says Momenta’s mission is “Better AI, Better Life” and frames the company as a Physical AI company whose autonomous-driving work can extend toward robots.
• He argues that advanced driving has moved from a fun toy toward a usable product and that supplier concentration will reward companies with data, R&D scale, customer programs, and delivery systems.
• He treats World Models as the bridge between cars and robotics: pretraining physical common sense, mid-training domain common sense, and post-training behavior alignment can be reused across related physical tasks.
• He says organization change was necessary when Momenta moved from a loose research-institute style into a customer-centered startup, especially after overpromising and early mass-production delivery pain.
• His decision method emphasizes years of early thinking for major choices, first-hand information, discussion with strong operators, and short-cycle evidence rather than last-minute founder impulse.

Connections

• Momenta — company leadership context.
• Gao Jiyang — source guest who chose Momenta partly because of Cao.
• Waymo — contrast case for Gao’s career decision.
• Microsoft Research Asia, Sun Jian, and SenseTime — research and product-execution training path.
• Autonomous Driving Data Flywheel, World Models, and Physical AI — technical strategy.
• AI Organization Design and Low-Cost Short-Cycle Validation — organization and decision-method themes.