Uber's drivers aren't just a labor cost, they are one of the company's main sources of capital since it is the drivers who must buy or lease the cars and make the payments. Here in Seattle there are around 9,000 drivers, presumably with roughly one car each costing about $25,000 apiece (avg. Prius), or over $200MM in capital equipment for just one city. Current estimates are that an Uber driver working full time nets <$40,000 per year not including the cost of the car, which is likely at least $6,000 in payments. There are over 160,000 Uber drivers in the US, collectively they're paying about $1BN a year to provide the cars for Uber. In a driverless model that cost is entirely shifted to Uber itself.
The car companies have another option for getting into the car sharing business with instant rentals. Daimler (car2go) already has about 750 cars here in Seattle and BMW (ReachNow) another 520, and they keep adding them. Or they can go the Lyft/GM's ExpressDrive route and provide a pool of rental cars to drivers. It will be far more profitable in the near term for other companies to jump into these markets than invest the billions it would take to build out a driverless fleet.
Yes but Uber would benefit from fleet pricing on purchasing and maintenance as well as the centralization of facilities (you could put the Uber fleet offices far away since the cars can drive themselves there for servicing).
Facilities = more capital. Just the yard to park and service 9000 vehicles is a lot to buy and then there's the buildings: if each car needs service just once a year you'd need at least half a dozen service bays running full time.
