Waymo / Google — Autonomous Driving AI Training Cluster

Waymo, Alphabet’s autonomous vehicle subsidiary, operates large-scale AI training infrastructure for simulation-based learning — training its driving models on billions of miles of simulated driving data generated via its proprietary Waymo Simulation system. The compute cluster, co-located at or near Google’s Mountain View and Bay Area data center infrastructure, runs continuous training cycles where real-world sensor data and simulation are combined to train the neural networks powering Waymo’s self-driving vehicles. Autonomous driving AI is among the most compute-intensive applications: Waymo processes petabytes of lidar, camera, and radar data per day from its commercial robotaxi fleets in Phoenix, San Francisco, and Los Angeles. The cluster supports Waymo’s 5th-generation Driver system, which required a complete retraining of perception, prediction, and planning models.