If they are truly just taxis, then their penetration will be approximately what taxis are today plus some increment to account for lower pricing. Maybe add rental cars.
So where is the math that lets everyone suddenly afford to and want to take taxis everywhere?
In New York City, a taxi driver takes home maybe 50% of the money collected in fares after costs (whether depreciation on their own car, renting a taxi by the day or week, gas, etc.) That may be high but it's a nice round number.
If we take this as a general rule of thumb (and NYC cabbies make more than the average), that means you could but cab fares by roughly 50% by eliminating the driver. Maybe that forces a behavioral change on the part of some urbanites. But I can pretty much guarantee you that cutting cab prices by 50% will minimally affect behaviors in most places.
And that's not even counting the many places where having the most rudimentary cab services is barely a cost-effective business today.
Many analysts have crunched the numbers, and come up with similar results, but Brad Templeton has been thinking and writing about Robotaxis longer than anyone else, so I'll link to his numbers:
Those numbers seem quite consistent with what I wrote. Costs per mile of a "normal" car today of about 50 cents/mile. (Which is approximately the IRS reimbursement rate so that seems very reasonable.) So that would be a good ballpark. Which is 1/2 to 1/3 of Uber today.
Sure, you can speculate about optimized taxi designs (which aren't especially relevant to non-core urban areas with high car ownership today) but small electric vehicles could exist without self-driving too.
IMO this is an academic exercise for at least a couple of decades in any case but it's an interesting thought experiment.
