I was thinking along these lines on a similar thread recently. The particular thread implied that the cost would be much lower, because there was no driver involved. I agree that it would be lower, but not orders of magnitude lower, because the demand shape will be exactly the same - or even more pronounced.

My hypothesis is that the driver is probably about 10-20% of the cost of a fare, the rest is the capital cost of the vehicle + licensing fees + insurance, and the marginal cost of maintenance and fuel. Because inevitably cars sit around most of the time, then the price of 5-6 busy hours of the day has to make up for the rest of the time.

Further, with a disruptive business idea like this, I could easily see an auction-style interface for the vehicle booking, which would give a much better revenue curve (we are talking about Google). In that case, the peak-demand period would probably exceed the current (regulated) taxi fares. But the plus side of that is that a midnight ride would be very cheap due to lack of demand and simultaneous lack of a need to pay drivers more money to work nightshifts.