You are not factoring in the economies of scale that will kick in if Waymo really figures out autonomous driving. It will change the world forever, and the costs will go down.

But they're only really trying to figure it out in fenced areas like Phoenix, no? It's not like they can get Phoenix sorted out and then start building systems that are intended to operate outside Phoenix -- I don't see any sort of a reasonable path to scaling. They're not building learning systems -- only training systems.

> It's not like they can get Phoenix sorted out and then start building systems that are intended to operate outside Phoenix

Um... Yes it's absolutely like that. They're using phoenix because it very little rain, so they can get cars that work ignoring bad weather

Afterwards they can scale out to other good weather cities and focus harder on expanding the weather they can drive in

By giving more than 80,000 rides in each car.

For anyone curious, if each ride takes 10 minutes, and the car can spend 8 hours a day doing rides, it would take 5.7 years for a car to give 100,000 rides. (I added a little to accommodate the price of charging the vehicles and maybe some maintenance)

I think this is pretty reasonable considering that for $5/ride I would expect there to be quite a lot of demand, and there's opportunities for things like food delivery as well.