Uber wants to turn its millions of drivers into a rolling sensor grid for self-driving companies. The company's CTO says the bottleneck in AV development is no longer technology. It is data.
Uber has spent years figuring out what its role is in a world where self-driving vehicles eventually replace the drivers it currently depends on. The answer it is landing on is not to build the autonomous vehicles itself, but to become the data infrastructure that makes everyone else's autonomous vehicles smarter. At TechCrunch's StrictlyVC event in San Francisco on May 1, Uber's chief technology officer Praveen Neppalli Naga laid out a plan that reframes the company's entire driver network as a potential Uber sensor grid for autonomous vehicle development.
The logic starts with a problem that every AV company is dealing with right now. The bottleneck in self-driving development is no longer the underlying technology. It is data, specifically the kind of dense, city-specific, scenario-rich real-world driving data that is extraordinarily expensive and time-consuming to collect at meaningful scale. An individual AV company operating its own fleet of test vehicles in a handful of cities can only capture so much. Uber, which has millions of drivers operating across hundreds of cities globally, can theoretically capture orders of magnitude more, across every type of road condition, traffic pattern, time of day, and edge case that autonomous systems need to learn from.
The program Naga described is an extension of something Uber announced in January called AV Labs. In its current form, AV Labs runs on a small dedicated fleet of sensor-equipped vehicles that Uber operates separately from its driver network. But the stated ambition is to eventually equip ordinary driver vehicles with sensor kits and turn the entire fleet into a distributed data collection platform. Naga was direct about the sequencing: the company first needs to understand how the sensor kits work and how to integrate them reliably before scaling to the broader driver network.
What is already operational is more significant than it might appear. Uber currently has partnerships with 25 AV companies, including London-based Wayve, and is building what Naga called an AV cloud, a library of labeled sensor data that partner companies can query and use to train their models. Partners can also run trained models in shadow mode against real Uber trips, simulating how an autonomous vehicle would have handled a given situation without ever putting one on the road. That capability alone is genuinely valuable for AV developers trying to validate model performance at scale before deployment.
For the MENA region, the implications are tangible. Uber operates actively across the UAE, Saudi Arabia, and Egypt, markets where ride-hailing density is high and urban driving conditions are distinct from anything an AV company could capture in North American or European test environments. Dubai's Roads and Transport Authority has been building toward autonomous mobility for years, with a stated ambition for 25% of all trips in the emirate to be driverless by 2030. Saudi Arabia's NEOM and broader smart city initiatives carry similar autonomous transport ambitions. If Uber's sensor grid expands into these markets, it would give AV developers access to Gulf-specific driving data that is currently almost impossible to obtain at any meaningful scale, creating a data asset with direct relevance to the region's own autonomous vehicle deployment timelines.
After divesting its own self-driving division years ago, Uber is now positioning itself as the essential data layer for the AV ecosystem rather than a competitor within it. The Uber sensor grid strategy is a clean reversal of a threat into an opportunity, one that could make Uber more central to the autonomous future than it would have been had it stayed in the AV race itself.
