By WaabiLearning Objectives
- Understand the end-to-end learned approach versus a modular self-driving stack
- See how generative simulation ("Waabi World") aims to replace millions of physical miles
- Evaluate the AI-native bet against its pre-commercial-scale reality
What Is Waabi Driver?
Waabi, founded by leading self-driving AI researcher Raquel Urtasun, builds the most AI-native autonomous-driving system in the sector. Where most companies assemble a stack of separately engineered modules — perception, prediction, planning — Waabi trains an end-to-end learned system and validates it largely inside a high-fidelity generative simulator called Waabi World.
That simulation-first approach is the whole thesis: the claim that a generalizable driver can be taught and tested far more efficiently in simulation than by driving millions of physical miles. In 2026 Waabi raised roughly one billion dollars (investors include NVIDIA, Volvo Group, Porsche, and Uber) and began expanding from autonomous trucking toward robotaxis on Uber's network.
💡Key Concept
Generative simulation for driving: Instead of only replaying recorded roads, a generative simulator can synthesize endless novel, physically-realistic scenarios — including rare, dangerous edge cases — to train and stress-test the driver. If it works, it compresses years of physical testing into compute. That "if" is the frontier Waabi is betting on.
Core Capabilities
- End-to-end learned driving — a single trained system rather than hand-engineered modules.
- Waabi World generative simulation — synthesizes realistic scenarios to train and validate the driver.
- Multi-domain — expanding from autonomous trucks to robotaxis.
- Compute-efficient validation — the bet that simulation replaces most physical-mile testing.
Company Details
| Detail | Info |
|---|---|
| Company | Waabi (private) |
| Founder | Raquel Urtasun (ex-Uber autonomy lead) |
| Founded | 2021 |
| Headquarters | Toronto, Canada |
| Funding | ~1 billion dollars (2026); NVIDIA, Volvo, Porsche, Uber |
| Approach | End-to-end learned stack + Waabi World generative simulation |
| Website | waabi.ai |
Best Use Cases
| Task | Why Waabi |
|---|---|
| Understanding AI-native autonomy | The clearest end-to-end + simulation bet |
| Simulation-driven AV development | Waabi World is the differentiator |
| Trucking-to-robotaxi generalization | Multi-domain from one learned stack |
When to choose alternatives: For proven driverless commercial operations today, Kodiak or Gatik are further along in the field. For OEM-backed production, Torc (Daimler).
Key Takeaways
- Waabi Driver is the most AI-native autonomous-driving bet — an end-to-end learned stack, not hand-engineered modules.
- Waabi World generative simulation aims to teach and validate the driver in compute rather than millions of physical miles.
- Roughly one billion dollars from NVIDIA, Volvo, Porsche, and Uber backs an expansion from trucks to robotaxis.
- It is a bet on a method as much as a product — genuinely differentiated and genuinely unproven at driverless commercial scale.
