Autonomous vehicles · unit economics
Why Waymo just left Uber's app in Phoenix
A dozen robotaxis, one pilot, quietly ended. Underneath the press-release politeness is the most interesting fight in transportation: who owns the rider when the driver is software? Play with the economics below.
Drag to rotate. The lidar spins; raise utilization and watch the taxi speed up while cost per mile falls.
Fleet vs aggregator: the real story of the split
Own the app (Waymo One)
- Keeps 100% of fare and, crucially, the customer relationship and ride data.
- Waymo already runs 100k+ paid rides weekly across Phoenix, SF, LA and Austin on its own app.
- Cost: must build demand city-by-city — marketing, support, payments.
Ride the aggregator (Uber)
- Instant demand: Uber's ~150M users mean full cars from day one.
- But Uber typically takes a ~25–30% cut on rides — and owns the rider.
- Works while AVs are scarce; stings once your fleet can fill itself. Phoenix (a dozen cars) taught both sides the same lesson.
Worked example: a $20 fare at 30% commission leaves $14. If the robotaxi's fully-loaded cost for that trip is $12 (depreciation + remote ops + cleaning + insurance + energy), the aggregator route nets $2 while the own-app route nets $8 — a 4× difference on identical miles. That arithmetic, not any feud, is why AV companies treat aggregators as training wheels. Note: in Phoenix Waymo kept its own app running the whole time; Uber still partners with Waymo in Austin and Atlanta, where Uber manages fleet operations — a different revenue split.
Key concept — utilization: a personal car sits parked ~95% of the time. A robotaxi spreads its huge fixed cost (vehicle + sensors) over every earning hour. At 20% utilization the hardware dominates cost per mile; at 60% it nearly disappears. This is why AV firms obsess over dense cities and airport queues, and why a 12-car pilot can never show real economics.