China's Baidu launches AI model simulating road environment for robotaxis

What happened

Baidu launched Apollo World, an AI system that simulates real-world driving environments to train robotaxis and predict how they evolve. Baidu says it can generate up to 100,000 km of simulated driving daily, equivalent to about three years of real-world testing.

Confirmed

Global impact / market context

Testing self-driving cars on real roads is slow and costly. Simulating up to 100,000 km a day could let Baidu train robotaxis faster and cheaper, potentially speeding up development of autonomous driving and reducing the expense that weighs on robotaxi companies.

Analyst inference

Baidu's Apollo Go is China's largest robotaxi operator, competing with WeRide, Pony.ai and Momenta. XPeng plans robotaxi operations in 2027. Morgan Stanley estimates China's robotaxi fleet could reach 1 million vehicles by 2035, making cheaper development a competitive advantage.

Analyst inference

What to watch

1. Baidu said Apollo World is already being used to train the Apollo Go robotaxi fleet, which the company intends to operate globally, so watch for signs of wider rollout. Confirmed
2. Watch whether rivals WeRide, Pony.ai and Momenta respond with their own simulation tools, since cheaper training could shift who leads robotaxi development. Analyst inference
3. Watch XPeng's planned 2027 robotaxi launch, which could show whether simulated training lowers costs enough for new entrants to compete with Baidu's larger fleet. Proposed

Evidence

• Original source link unavailable