Steve Xie

Founder & CEO Lightwheel

Last Reviewed: September 23, 2026

Executive Summary

Steve Xie is the founder and chief executive of Lightwheel, a company building what it calls the data engine for physical AI. Robots are expensive to train in the real world, so Lightwheel supplies what they learn from and are tested against. It offers physically accurate simulated assets and scenes, recorded first-person human demonstrations, and benchmarks for evaluating robot foundation models. Xie came to it from autonomous driving, where simulation is already standard practice: he previously worked on driving simulation at NVIDIA, Cruise and NIO. In 2026 the company was reported to have raised money in a round led by Ant Group and in a strategic financing of about RMB 1 billion.

Career Highlights

• Founder and CEO of Lightwheel, per his UN Open Source Week 2026 speaker profile and RuntimeWire’s August 2026 coverage.
• Built a three-layer product line: SimReady (simulation assets and scenes), EgoSuite (egocentric human demonstration data) and RoboFinals (industrial simulation evaluation), plus an enterprise platform that combines them.
• Released EgoSuite-Open100K in August 2026, opening 10,000 hours of annotated first-person footage for robotics research with 90,000 more hours planned.
• Raised capital in a round led by Ant Group, reported in May 2026, and about RMB 1 billion in strategic financing, reported for June 2026.
• Previously worked on autonomous-driving simulation at NVIDIA, Cruise and NIO.

Professional Journey

Xie studied physics at Peking University and completed a Ph.D. in quantitative finance at Columbia University. He then went into autonomous driving, working on simulation at NVIDIA, the self-driving company Cruise and the Chinese electric-vehicle maker NIO. His speaker biography gives no dates or titles for these roles.

Simulation in autonomous driving exists because real-world testing cannot cover enough rare situations safely. Lightwheel applies the same reasoning to robots. Its public materials describe a stack in three layers. The world layer, SimReady, is a library of physics-accurate 3D assets and scenes for simulators such as NVIDIA Isaac Sim. The behavior layer, EgoSuite, collects multi-modal first-person recordings of people doing tasks, for robots to learn from. The evaluation layer, RoboFinals, is a benchmark platform for testing vision-language-action and world models on household, hospital and factory tasks. The company has published a series of technical collaborations with NVIDIA’s Isaac Lab and Newton physics engine.

In 2026 Lightwheel raised capital in at least one and possibly two rounds. Gasgoo reported in May that it had closed a round led by Ant Group with industrial investors. RuntimeWire later described about $140 million (RMB 1 billion) of strategic financing in June from investors including the Zhongguancun Science City Fund and the Sichuan Development Sci-Tech Innovation Fund. In August the company released EgoSuite-Open100K, which it describes as the largest fully annotated open egocentric human dataset.

Education

Peking University

B.S., Physics

Columbia University

Ph.D., Quantitative Finance