The highest capital costs are the station equipment/installation and ongoing electricity. They have already made enough stations that there isn't much economy of scale to be had on the stations and electricity costs does not decrease per car. I don't see there being a significant reduction in costs per vehicle (esp. the 2-4x suggested).
Hmmm... I just realized you are looking at this from the completely different end of the telescope as I am.
The economy of scale I refer to is the cost of the DC fast charging hardware in the cars themselves. Not the cost of the hardware that is installed in the field for Superchargers. Imagine that if Tesla Motors reaches a point where they are selling 100,000 of the Model S, 100,000 of the Model X, and 500,000 of the Model ☰ per year. And all of them have the DC charging hardware built in that allows them to use the Supercharger network, or alternatively, the CHAdeMO chargers. 700,000 units per year is vastly more than the 20,000 units of combined output of Model S and Model X that Tesla Motors hoped for in 2013.
The only reason there used to be a fee to add Supercharger access to the Model S 60 was to pay for the DC fast charging hardware that was added to the car if requested. It was not for the operation of, or installation of, the Supercharger network at all. Once Tesla Motors realized it was easier to simply include the DC fast charging hardware in every car during the manufacturing process, it was decided to make the fee for activating a software switch to use that hardware.
Now, since we know that the Model S 85 has included Supercharger access from the very beginning... And, that now the Model X and every iteration of the current Model S has Supercharger access included... We may speculate that the cost is built in to the margin of the vehicles, which we generally express as being 25% or so. With that in mind...
The Model S 60 is now $66,000 and includes Supercharging. You might presume that if it achieves a 25% margin, it cost around $49,500 to build. If the DC fast charging hardware were still an option, and if one assumes there was no reduction to internal cost due to economies of scale since 2012... Then the $2,000 fee to add it would be roughly equivalent to 12.1% of the $16,500 profit margin.
[FYI: The number is only 11.4% of $17,500 with a $70,000 Model S 70 instead.]
Proportionately speaking, if there is only a 12% margin on the $35,000 Model ☰... The car would cost $30,800 to build... Making for a $4,200 raw profit, and 12.1% of that would be... ~$509.
[ASIDE: And at 11.4% this would be $480, so $500 works as a good compromise.]
Just as it is presumed that the embedded $2,000 per Generation II vehicle is put toward the Supercharger network... I presume that a $500 amount per Generation III vehicle will join the Supercharger funding pot
(yes, even if someone doesn't order the option). Therefore, even if
'only' 100,000 Generation II vehicles are sold per year, that yields $200,000,000 for the Supercharger network per year. And if
'only' 300,000 Generation II vehicles are sold per year, that adds another $150,000,000 to the pot.
OK, so with $350,000,000 to work with, even if you only used half of it to build Superchargers... And it cost around $400,000 to build an 8-stall location... You would have enough dough to build 437 of them from one year's earnings. And, with $175,000,000 left over for energizing them, even if you paid $0.25 per kWh, you'd have enough to buy 700,000,000 kWh of energy. If 49 kWh were expended during each charging session, you could do so 14,285,714 times. yeah. 14 million times. From the sale of 350,000 vehicles. Or, about 40 times each.
In a world where most people rarely use Superchargers at all, that's pretty good. Especially since the pay-per-use advocates seem to think they wouldn't use Superchargers 40 times in 20 years... We'll see.