Your average cost only drops over a relatively narrow range: from where N towers is enough to support your users to where N+1 towers are necessary. After N is big enough to cover your area, cost still goes up basically linearly with the number of users (just in a saw-tooth fashion rather than a straight line).
It's totally unlike the costs of any service with real economics of scale, such as software.
Your thinking in terms of a small coverage area. Networks needs to be nationwide and users are not limited to a single tower.
As density increases infrastructure can more closely map to usage.
Net effect if 50 million people take X towers then 100 million people take less than 2x towers.
Re:The rational exuberance of 5G
by TheCoelacanth on Friday July 01, 2016 @12:37PM
(#12017600)
For vast swaths of the country, usage would be low enough that it doesn't even fully use one tower. For those areas, the marginal cost of adding a user is effectively zero.
Put another way, to have good coverage and 1 customer would be ridiculously expensive. There are always going to be areas with few customers. You add towers in areas with lot's of customers.