A look at the different methods and technologies that Waymo and Tesla use to collect data from the billions of miles their self-driving vehicles have driven
Sean O'Kane / The Verge :
Context & Ripple Effects
By early 2018, the two most-watched autonomy programs had already committed to opposite data strategies: Tesla was harvesting telemetry from customer cars at scale — a practice later detailed in [[a:981616|IEEE Spectrum's breakdown of the breadcrumb GPS trails and gateway logs its 3M vehicles generate]] — while Waymo refined its stack on a dedicated fleet alongside Google Brain researchers applying deep learning and neural nets to that data.
The Verge's comparison matters because those collection choices set up everything that followed: Tesla's California testing was still tiny in early 2017 — four cars covering just 550 miles, while by 2025 Waymo's approach had scaled into commercial operation with Pichai citing 200K+ weekly paid rides.
First-order effects
- Tesla's customer-fleet model turns every owner's car into a data collector feeding Autopilot development, while Waymo's purpose-built sensor suites produce richer but costlier data from a controlled fleet — each company locked into the economics of its own choice right now.
Second-order effects
- Waymo's fleet-first path forces it to build operational infrastructure competitors don't need — training roughly 25,000 humans to support about 3,000 robotaxis and wiring towing through the Honk app — costs Tesla defers by running an estimated 30 Austin robotaxis with safety drivers aboard.
Third-order effects
- If both patterns hold, the industry splits structurally between fleet-learning consumer autonomy and supervised robotaxi operations, with data ownership — already contested in Tesla's case per IEEE Spectrum — becoming the regulatory battleground that decides which model scales.
The trend: Autonomy development is bifurcating between Tesla's mass-market fleet-telemetry model and Waymo's purpose-built robotaxi operations, with data collection strategy determining each company's cost structure and regulatory exposure.