Lang Xianpeng on Recreating 'Mass-Produced Intelligent Driving at Million Scale' Through Robotics Entrepreneurship

Lang Xianpeng, who led Li Auto's autonomous driving from inception to mass production in 1.5 million vehicles, left to co-found Kunlun Xing with Ren Geng from Huawei. The company, which focuses on embodied intelligence robots, raised billions in funding within three months, becoming a unicorn. In an interview, Lang stressed that the key to embodied intelligence is understanding, not imitation, and that the industry's ultimate test is mass production and self-sufficiency, comparing the current state to autonomous driving in 2015-2016. He detailed their technical path: a physical-causal world model and 'data compilation' to enrich training data with physical properties. They chose a full-stack approach, self-developing key components like joints and motors. Their business strategy starts with toB applications, aiming eventually for toC home use. Lang drew parallels to his experience at Li Auto, emphasizing data-driven approaches and paradigm shifts from HD maps to end-to-end and VLA. He also highlighted the importance of engineering alongside models, and described their 'one brain, multiple forms' architecture, which learns physical laws that are inherently generalizable.