I doubt very much customers will have much brand loyalty to Uber. There is very little differentiation of the actual service.
Uber must be prepared for a world with self driving cars. But if their entire model requires self-driving cars in order to work, if they are simply burning cash every year until then, then I think they will burn-out long before self driving is a reality.
I doubt this is true, at least if you're talking about all-weather driving. Not even the optimistic researchers I've chatted with are that optimistic; this timeframe would require all-weather driving to be basically solid at the research level by 2018 or so, which is not totally impossible but unlikely. Maybe if it's limited to Uber Phoenix, though.
At least, Uber ATC is in Pittsburgh, where it can become more grounded in reality by potholes, haphazard construction, thin windy two-way roads without markings or lanes, and unexpected acute precipitation.
Self driving cars that work 90% of the day means Uber gets 90% of potential profits, and continues to use human drivers for the other 10% "surge" time period.
For fans of a technology, that's a familiar optimism. I was young once, 20 years ago, I felt that way about fusion myself, hah!
What protects Uber is the network effect. Even if I had a million self driving cars right now, nobody has my app installed, and I don't have an optimized pricing and logistics set up yet.
And if you have your own fleet of self driving cars, why would you build the whole setup yourself? Why not just rent them to Uber?