It is interesting to consider the choice of many small vs a few large Supercharger sites. Larger sites have better efficiency statically (Erlang efficiency); there is a huge win going from 4 to 8 or 12 stalls (4->8 is 3.5x; 4->12 is 5x capacity), but the percentage gain starts falling off after that. Also, site acquisition is a bear, and big sites are harder. My prediction is that Tesla will continue with the 12 stalls or so size in busy areas, and gain from there with smart tools in the car. Tesla could easily monitor Supercharger usage and make recommendations to the driver on busy and lightly used sites ahead. One step further and Tesla could add a feature to the Nav function that offers alternatives for Supercharging based on how busy sites are, just like there are alternatives offered in navigation based on traffic; Supercharger usage is just another form of traffic. Offering such info to the driver combines the number of stalls in a region to gain better queueing (Erlang) efficiency.
I predict many 8-12 stall sites with real time usage info given to the user in some form to help users find open stalls.