If the car drives itself and you rent a trip rather than buy and own a specific vehicle, why would most people care to pay a premium for a Tesla trip rather than just getting a Nissan Leaf With Google or Hyundai e10 With Uber or a self-car2go or...?
My model assumes Tesla is the innovator, and has self-driving cars first (possibly in a partnership with Google). Uber is playing catchup, and incumbent auto manufacturers won't want to lose 90%+ of their demand to by-the-mile rentals.
A self-driving taxi service is not easy to scale because in addition to good maps of anywhere you want to service, you need to own and maintain physical cars (starting with nightly cleaning on weekends). This isn't an app, and being popular in California won't help you operate vehicles in Barcelona.
On the other hand, once someone makes a good-enough software platform, they can just license it to any local Addison Lee-type taxi/minicab company that already owns the cars and has experience maintaining them. If that someone is Google, they might just give it away like the rest of their products.
Uber got relatively large adoption by largely avoiding trying to scale too much horizontally, by using local drivers with their own cars - routing is the only thing they've done (and it's not as good as local knowledge). They claim to not be a transportation company at all, they just connect contractors. That sort of thing won't be possible with self-driving Tesla taxis.
Plus, we don't yet know if Tesla is any good at making non-premium cars.
Regarding incumbents trying to avoid cannibalisation - are you familiar with carshare systems operated by BMW and Daimler?
