Uber to put 500 data-collection vehicles on the road this year

Source: Transportation & Auto News | TechCrunch ↗

Uber unveiled a sensor-equipped Hyundai Ioniq 5 prototype and said it will deploy 500 of the modified vehicles globally this year to collect training data for its 30-plus AV partners, including Waymo, Avride, and WeRide. The fleet, built with Roush Performance and running on Nvidia's Drive Thor compute, carries 14 cameras, eight solid-state lidars, and nine radars, and is targeted to capture 2 million miles per month of high-fidelity data, with 50 vehicles on the road by summer.

Why it matters

Uber is rebuilding an AV business without owning a stack. Five years after selling its self-driving group to Aurora, it is back assembling vehicles, this time as a neutral data utility for a roster of 30 partners that competes with itself. The pitch to Waymo, WeRide, and Avride is geographic diversity: Uber operates in cities none of them can cover alone, and a 500-vehicle, 2-million-mile-per-month feed of time-synchronized camera, lidar, and radar is hard to replicate without an Uber-scale ground game.

The dependency stack is the quiet part. Roush does the retrofits, Hyundai supplies the platform, and every byte routes through Nvidia Drive Thor, which makes this another design win locking Thor in as the default AV reference compute. For operators, the question is pricing. If Uber sells this data on a per-mile or per-city basis to the same partners running robotaxis on its network, it has found a second margin layer on top of the dispatch take rate, and the AV partners have a new variable cost line that scales with how badly they need coverage outside their home geofences.

Background reading

Related signals

← Read other signals from this week

Friday issue, by email. Free. No spam.

Background reading

Related signals

Subscribe