Waabi Driver

Waabi was founded in Toronto in 2021 by Raquel Urtasun, previously chief scientist at Uber's self-driving unit and a professor at the University of Toronto. Its bet differs from the rest of the industry in two ways. First, the Waabi Driver is a single end-to-end AI model with what the company calls verifiable reasoning, rather than a pipeline of separate perception, prediction and planning modules. Second, most of its training and validation happens inside Waabi World, a simulator built to be accurate enough that fleet miles are the exception rather than the raw material. The approach produced a result in June 2026 that no competitor has matched publicly. The Waabi Driver, trained on a Peterbilt 579, was installed on a Volvo VNL Autonomous — a different vehicle with different sensors, controls and physical behaviour — and drove competently on highways and surface streets from the first mile, with no new real-world data, no simulation data and no fine-tuning. Zero-shot transfer between truck platforms had not been demonstrated before. If it holds up at scale, it attacks the cost that has kept autonomous trucking small: every new vehicle model has so far meant a fresh engineering programme. Commercially the company is behind its own timetable. Waabi said it would run its first fully driverless trucks, with nobody in the cab, by the end of 2025; as of August 2026 there is no announcement that this has happened, and the public material describes road testing of the Volvo platform rather than driverless commercial service. Money is not the constraint — Waabi raised USD 1 billion in January 2026 — and neither is industrial backing, with Volvo Autonomous Solutions committed to building the VNL Autonomous around the Waabi Driver and Lior Ron, co-founder of Otto and Uber Freight, joining as chief operating officer in August 2025. The honest summary is a technology lead paired with a deployment gap. Aurora and Kodiak are already being paid for driverless miles; Waabi is the company with the strongest argument that its approach will generalise once it starts.

▸Videos

No videos yet.

▸Reviews