They already have the cheapest solution for having a driver and a car go around town, so sending drivers with their “autonomous” cars in good conditions is an impressive step, but insufficient for full autonomy and not enough to fix their business problems — they need to be able to send the cars in real life situations with no driver.
So 5 years of technology and 5 years of regulation before widespread adoption is a long time for Uber to have to be spending investor money before they spend investor money on autonomous car fleets in every major city. (Or are they going to lease cars from other people? — it may be then that the cost of leasing vehicles from others never comes low enough.)
The better hope for their investors seems to be the rumor they’re profitable in large, established cities and their costs are expansions and turf wars.
IMO, the stuff which has already been demonstrated is a bit too creepily good for comfort, even if we're talking about limited Tesla autopilot.