XPeng Bolsters AI Team with StepFun Co-Founder to Accelerate Autonomous Driv...

摘要： XPeng增强AI团队，聘请StepFun联合创始人以加速自动驾驶研发，此举旨在强化公司在人工智能领域的实力，推动自动驾驶技术的快速发展，新成员的加入将带来专业知识和创新思维，促进...

XPeng增强AI团队，聘请StepFun联合创始人以加速自动驾驶研发，此举旨在强化公司在人工智能领域的实力，推动自动驾驶技术的快速发展，新成员的加入将带来专业知识和创新思维，促进XPeng在自动驾驶领域的领先地位。

Dr. Zhou Shuchang

XPeng Motors has secured a major addition to its artificial intelligence (AI) roster with the hiring of Dr. Zhou Shuchang, a founding team member of StepFun and a recognized AI expert.

Zhou will serve as Senior Director of Autonomous Driving Algorithms, reporting directly to XPeng’s head of autonomous driving, Liu Xianming, according to sources close to the company.

Industry insiders say Zhou will lead the deep training of XPeng’s autonomous driving foundation models, a move signaling the automaker’s growing investment in its “Physical AI” strategy, which aims to integrate AI capabilities into both vehicles and robotics.

Dr. Zhou, a graduate of the Chinese Academy of Sciences with degrees from Tsinghua University and the National University of Singapore, brings extensive experience from Google, Megvii, and StepFun. He is regarded as one of China’s earliest AI scientists to combine multi-modal large models, reinforcement learning, and intelligent agent research with real-world industrial deployment.

At StepFun, Zhou co-led major projects including Step1, Step2, and StepR, covering the full spectrum from algorithm development to engineering deployment. In his social signature, Zhou described his approach as “research × engineering × product,” reflecting a philosophy that aligns closely with XPeng’s talent strategy: cultivating AI engineers who can span the full cycle from algorithms to product deployment.

His direct report, Liu Xianming, previously held senior positions at Meta and Cruise and has been instrumental in deploying foundation models for autonomous driving at XPeng. Zhou’s addition is viewed by experts as a key step in strengthening XPeng’s large-model capabilities and expanding its autonomous driving algorithm stack.

Over the past year, XPeng has aggressively recruited AI talent from leading tech firms including Meta, ByteDance, and Alibaba, as well as PhDs and researchers from top universities such as Stanford and Carnegie Mellon. Core areas of focus include multi-modal foundation models, reinforcement learning, AI infrastructure, and hardware-software co-optimization. Observers note that the pace of XPeng’s team building rivals that of major internet AI labs.

XPeng founder He Xiaopeng has repeatedly emphasized the company’s shift toward a “Physical AI” strategy. Speaking on the blog show Luo Yonghao’s Crossroads, He outlined plans for long-term AI investment, with 30 billion yuan of a total 50 billion yuan budget dedicated to AI research and development, while the remainder covers hardware and supporting software.

XPeng has built China’s first AI compute cluster at scale in the automotive sector, and expects to invest 4.5 billion yuan in AI and autonomous driving in 2025 alone. In April, the company unveiled its “XPeng World Foundation Model,” a 72-billion-parameter multi-modal autonomous driving model integrating vision, language, and action (VLA).

This summer, XPeng rolled out the G7 Ultra featuring three Turing chips delivering 2,250 TOPS of computing power, along with a VLA-powered “parking-to-parking” functionality, demonstrating AI capabilities increasingly visible to users. An internal source indicated that XPeng plans to showcase key breakthroughs in its “Physical AI” initiative during a tech day on November 5, with a “World Model” capable of simulating the physical world—a cornerstone for Level 4 autonomous driving.

XPeng’s aggressive talent strategy is clear: attract top AI scientists with competitive compensation while building foundational capabilities for an AI-driven car company. The company’s headcount has grown from 6,000 to 8,000 this year, with AI roles representing more than half. Campus recruits for 2026 may see starting salaries up to 1.6 million yuan, with no upper limit for exceptional candidates, according to He Xiaopeng.