There is a nice private solution to this, and this sort of thing is not talked about enough.
Imagen we have two companies, both are ISP with their own infrastructure. Then because of other reason a new expensive infrastructure project is needed. These two could ban together and form a new organisation (non-provit or a club) that economises on the infrastructure but provides access and price garanties to both companies and maybe even sell access capacity to third parties.
The differnce to a normal company providing infrastructure is that the costumers of that company always need to be afraid of price raises. Once you have this club situation you essentially have one infrastructure, but you avoid monopoly pricing because competitive game has moved to a higher abstraction. This would provide very low cost and very low price to consumers.
If you apply a regulatory system as soon as you think there is a natural monopoly the industry will almost certantly stagnate.
Would you say this is because the benefits to decentralization don't often bear themselves out as visibly as the costs over a shorter time horizon?
In other words, is a "natural monopoly" simply externalizing the costs of centralization to the future?
As for whether there really are natural monopolies, the classic examples are things like electrical distribution systems. If one company has already wired up a city, it's going to be hard for someone else to come in and wire up the same city in competition. (And may not be a great idea from the perspective of public policy.)
However, over longer timeframes technology change and market forces overturn many monopolies. IBM, Windows, Ma Bell--these were all considered to have monopolies at one time.
It's also fairly clear that at least over long enough time horizons there aren't many natural monopolies.