Uber Deploys 500 Ioniq 5s to Supply AV Training Data
Key insights
- Uber's AV Labs will deploy 500 sensor-loaded Ioniq 5 vehicles targeting 2 million miles per month of autonomous driving training data.
- Roush Performance handles retrofits; each vehicle carries 14 cameras, eight solid-state lidars, nine radars, and Nvidia Dual Drive Thor compute.
- More than 30 AV technology partners, including Waymo, Avride, and WeRide, will receive geographically diverse datasets from the fleet.
Why this matters
Uber is leveraging its existing ride-hail network to become the primary supplier of geographically diverse, high-fidelity training data at a scale individual AV developers cannot easily replicate. The 30-plus partner client list spanning competitors like Waymo and WeRide reveals that even well-funded AV programs are willing to outsource ground-truth data collection rather than build proprietary fleets. Paired with the February launch of Uber Autonomous Solutions for robotaxi and self-driving truck operations, this signals Uber is building full-stack AV infrastructure rather than remaining merely a distribution channel.
Summary
Uber is assembling 500 modified Hyundai Ioniq 5 vehicles to collect driving data for autonomous vehicle partners, the company's first vehicle build since selling its AV unit to Aurora in 2020.
Each carries 14 cameras, eight solid-state lidar sensors, nine radars, and Nvidia's Dual Drive Thor compute. Roush Performance handles retrofits. Target: 2 million miles per month of high-fidelity data for 30-plus AV partners including Waymo, Avride, and WeRide.
Essentially: (Uber AV Labs, Roush Performance) are building the data supply chain AV developers cannot replicate at geographic scale on their own.
- 500 vehicles planned globally in 2026; 50 expected on roads by summer.
- AV Labs already draws from thousands of camera-equipped vehicles across dozens of cities and hundreds of Lucid Air vehicles in the U.S. and Europe.
Uber isn't becoming an AV company; it's becoming the data layer every AV company runs on.
Potential risks and opportunities
Risks
- AV partners like Waymo or WeRide could internalize data collection as their own fleets scale, reducing Uber's leverage within two to three years.
- Roush Performance as the sole retrofitter creates a single production bottleneck; any capacity constraints could push the 500-vehicle deployment past 2026.
- Collecting driving data across dozens of cities concentrates privacy and regulatory exposure, and a single adverse ruling in a major market could ground fleet operations.
Opportunities
- Sensor vendors supplying the 14-camera, 8-lidar, 9-radar stack face a 500-unit contract with expansion potential if AV Labs scales beyond its 2026 target.
- AV startups without capital for proprietary mapping fleets can now accelerate training timelines by sourcing from Uber AV Labs, compressing time-to-deployment.
- Nvidia's Dual Drive Thor gains a high-visibility reference deployment across 500 vehicles, strengthening its positioning against competing AV compute platforms ahead of coming commercialization cycles.
What we don't know yet
- Pricing model for the data: whether Uber charges per mile, per dataset, or takes equity stakes in partner companies is not disclosed.
- Which geographies the initial 50 vehicles will cover by summer, with city selection criteria and international deployment sequencing left unspecified.
- Whether training data delivered to Waymo is also available to competing partners like WeRide, and how Uber structures exclusivity across its 30-plus client base.
Originally reported by techcrunch.com
Read the original article →Original headline: Uber Deploying 500 Data-Collection EVs for Robotaxi Training This Year — 14 Cameras, 8 Solid-State Lidars, Nvidia Dual Drive Thor, Targeting 2M Miles Per Month for 30+ AV Partners