Internet service providers have to connect to each other somehow and this is done at IXP (internet exchange points: https://en.wikipedia.org/wiki/Internet_exchange_point )-- basically a huge building where separate providers can come together and exchange traffic.
I imagine that a peering connection in such a facility requires a heroic amount of maneuvering around 800 lb gorilla companies, but it certainly is possible if one has enough traffic and money.
Edit: grammar
There's a threshold beyond which peering make more sense than transit... (http://packetpushers.net/transit-vs-peering-makes-sense/).
I'm sure it'd be much easier if people just moved into their closest DC. That last mile really is such a drag.
Their operating margins are like 97% profit, and depending on the maturity of the network, the costs on debt service are something like 25% to 90% of costs. It's a business that generates consistent long term profit and cash flow.
Cable companies make bank btw.
Whoa, who told you that? Lets think about this in simplified terms, if you could run a coax cable to houses around you without caring about it being buried or hidden and start earning $50-$100 a month would you do it? I would be doing nothing else until I collapsed. I would go up and down the street with cable spools of cable putting posts in the ground to split it to people's houses. In one long day you could be earning what you would get from a minimum wage job. In a week you would be middle class. In a month you would live like a lawyer, in two you would live like a doctor, and you would never have to work again to make the same amount of money. Gold mines don't even do this well.
---
Building networks is, to no one's surprise, expensive. Financial analysts last year estimated that Google had to spend $84 million to build a fiber network that passed 149,000 homes in Kansas City, with the cost per home at $500 to $674. That figure did not include additional costs for actually connecting each home that requests service. A national Google Fiber build out passing 15 percent of US homes would cost $11 billion a year for five years, Wall Street analysts have estimated.
Montgomery discussed the challenge of finding what some people call "patient capital," or investors who are willing to put money into companies that might take five or 10 years to make a big profit.
"Costs are fixed. You're going pay your staff and maintenance people whether you have one customer or 10,000. All your costs are basically going to be the same," he said. "For a fiber network, you need to reach about 30 percent of the local market within a couple of years, or you're going to go out of business."
---
"One big reason we lack Internet competition: Starting an ISP is really hard
Creating an ISP? You'll need millions of dollars, patience, and lots of lawyers."
http://arstechnica.com/business/2014/04/one-big-reason-we-la...
The reason AT&T was able to gobble up the entire country in the 20th century was in part their ability to sell convertible securities that delivered premium returns with low/moderate risk.
http://www.dslreports.com/shownews/ATT-Takes-Billions-in-FCC...
If you're a cable company who already built all of that, it's marginally very profitable to layer on internet service.
If you think that you can go run a coax cable to someone's house and start collecting a monthly check from them, and collect routinely, without a) billing issues, b) collections issues, c) service outages, etc. etc. -- I mean, maybe for 3 neighbors it would work. But not at scale.
Btw 'at scale' all the overhead you mentioned decreases. I doubt it costs $70 a month per customer. You see issues constantly, but how big is the customer base? 100k? That's 7 million a month. With bandwidth costs of around $1 per person I'm going to guess there is a lot of room for profit.