I'm not following your first point. Autonomous Ubers would be housed in discrete locations near metro areas in off-hours and then roam around during the day or

I'm not following your first point. Autonomous Ubers would be housed in discrete locations near metro areas in off-hours and then roam around during the day or use negotiated low-rate parking spaces, so the location cost would not be much higher. Insurance rates are based on actuarial estimates of likelihood to cause accidents. Autonomous vehicles will not be on the roads in a meaningful way until they are significantly better drivers than humans, thus insurance rates will drop low enough that fleet owners just self-insure to cover incidental damage from no-fault causes. I'm also not sure what you're getting at in your last point. Uber has already built out a highly capable dispatching service to manage large fleets roaming around urban areas and they have a massive dataset to be able to predict ideal fleet allocations. The only real added cost will be the fuel overhead to support cars driving around looking for fares (currently that cost is eaten by human Uber drivers).