Podcast Episode
The system represents a major collaboration between the two Alphabet subsidiaries, combining Waymo's autonomous driving expertise with DeepMind's world model capabilities. Unlike traditional simulation systems trained solely on driving data, the Waymo World Model leverages Genie 3's broad world knowledge, gained from pre-training on an extremely large and diverse set of videos, to generate scenarios the fleet has never directly encountered.
Waymo completed fifteen million rides in 2025, tripling its annual volume and surpassing twenty million lifetime rides. The company currently handles more than four hundred thousand weekly paid trips across six major US metropolitan areas. With over twenty billion simulated miles already driven, the World Model represents the next evolution of a simulation-first development strategy that has been central to Waymo's approach since its inception.
Waymo Unveils AI World Simulator to Train Robotaxis in Tornadoes and Plane Crashes
February 6, 2026
Audio archived. Episodes older than 60 days are removed to save server storage. Story details remain below.
Waymo has launched its World Model, a generative AI system built on Google DeepMind's Genie 3 technology, capable of creating hyper-realistic virtual environments to test self-driving cars in extreme scenarios like tornadoes and planes landing on motorways. The tool arrives as Waymo pursues aggressive global expansion following a sixteen billion dollar funding round.
Waymo World Model Brings the Impossible to Life
Waymo has introduced its most advanced simulation tool yet, the Waymo World Model, a frontier generative AI system built on Google DeepMind's Genie 3 technology. The platform creates hyper-realistic virtual environments that allow engineers to test autonomous vehicles in scenarios ranging from tornadoes and vehicle fires to planes landing on motorways, situations that would be impossibly dangerous or impractical to recreate on real roads.The system represents a major collaboration between the two Alphabet subsidiaries, combining Waymo's autonomous driving expertise with DeepMind's world model capabilities. Unlike traditional simulation systems trained solely on driving data, the Waymo World Model leverages Genie 3's broad world knowledge, gained from pre-training on an extremely large and diverse set of videos, to generate scenarios the fleet has never directly encountered.
Three Pillars of Control
The model's architecture offers three key control capabilities. Driving action control allows engineers to explore "what if" scenarios by changing vehicle behaviour. Scene layout control enables adjustment of road configurations and environments. Language control lets operators modify weather, lighting, and conditions through simple text prompts. Crucially, the system generates high-fidelity, multi-sensor outputs including both camera and lidar data, essential for validating Waymo's sensor fusion approach.Expansion Fuelled by Record Investment
The announcement arrives as Waymo pursues aggressive global expansion following a sixteen billion dollar funding round at a one hundred and twenty-six billion dollar valuation, the largest investment ever in an autonomous vehicle company. The robotaxi leader plans to expand into over twenty additional cities in 2026, including its first international markets in Tokyo and London.Waymo completed fifteen million rides in 2025, tripling its annual volume and surpassing twenty million lifetime rides. The company currently handles more than four hundred thousand weekly paid trips across six major US metropolitan areas. With over twenty billion simulated miles already driven, the World Model represents the next evolution of a simulation-first development strategy that has been central to Waymo's approach since its inception.
Published February 6, 2026 at 7:51pm
