Uber has struck a partnership with Daimler to host Daimler's (hypothetical) self driving cars on Uber's ridesharing service, and they could strike similar partnerships with other Autonomous vehicle providers. So should Uber fail to develop their own AVs, that's one alternative model. Alternately, they could license the technology from Intel (which recently acquired Mobileye at a whopping $15 billion). These options may be suboptimal to vertically integrating the hardware, software and services under one company, but it's better than death.

But Uber is in 400-something cities worldwide, it'll be a long time before Robotaxis reach market saturation everywhere, it'll require millions of autonomous fleet vehicles to do that. One advantage that Uber has over a traditional automaker entering the space is that Uber can, at the outset, use autonomous vehicles to augment human driven networks. If a company such as Ford, say, cannot provide enough robotaxis to meet demand in a given locale, then their network won't be reliable and it'll just piss off customers. Uber won't have that problem, their network of human drivers will always be able to fill in gaps in service.

A lot would have to go wrong for Uber to be displaced entirely. We can only guess as to what the demand for autonomous rideshare will ultimately be, but it could easily be 10x, or even 100x what the demand for conventional rideshare ever was.

Although even amongst rideshare companies, both Lyft and Didi are looking more interesting. Didi has secured $5 billion in new capital and will likely be making some high profile AV start-up acquisitions out of their Silicon Valley office soon, and will be partnering with some Chines automaker to bring robotaxis to Asia. Lyft has a secure partnership with GM, the most exciting developer of L4 out there at the moment, given that Cruis automation is making dramatic progress using the Chevy Bolt platform, the first long range mass-market electric vehicle to go into production.