I would think dense wireless networks in urban areas could work. I'm not an expert, and would be interested in hearing from one, but I imagine that wireless networks could benefit from algorithmic advances in a way that might be more difficult in traditional copper/fiber networks.
If you gave me (someone with very little wireless experience but used to be a cable guy) carte blanche, here's what I would investigate:
Several routers in an urban area which require minimal installation and can be mass-produced and immediately interchanged - think roof access, and could be turned on and off remotely, maybe battery operated, flowing to relatively few fiber hubs.
Infrastructure advantages: Network algorithms have come a long way, and the ability to dynamically add or scale nodes as needed could have great performance/cost balances that aren't possible with fixed nodes. And upgrades do not require new lines, just more and better routers. If a router dies, replace, and send the old one to a shop somewhere inexpensive. No midnight service restoration pay.
Customer advantages: I connect to wifi the same way I do right now. Zero home installation (google Comcast horror stories if you need imagery).
If I were to start, I would do it in an urban area with a landowner that had a monopoly. Hyde Park, Chicago is like that. There are basically two landowners: University of Chicago and Mac Apartments. You have to make one community love you to pieces before branching out, and that would require a lot of cooperation. Of course a danger here is if Mac suddenly decides that they hate you and removes access privileges. But if that worked, and you got more funding to expand, imagine the word of mouth: "Here's how installation works: it's a wifi network and you connect to it".
A la "it's a folder, and it syncs" dropbox. Presumably something like this has been tried before, but that's true of many things that end up working (like search engines or social networks).