Wholesale rates from quality providers are about $0.30 per Mbps per month at 95th percentile. For paid transit, that is, and Comcast does a lot of peering which brings the cost down for them.
So that's $30 per month for your 100M line, even if you run it full tilt all month and manage to hit only paid transit routes all the time. That's not too bad and most people use only a small fraction of their theoretical monthly throughput.
first, the costs cited above are for handoffs at major peering points, private or public. Pulling 10Gb/s across to your apartment complex is going to add a lot to the cost, if it can even be done (right of way issues).
second, the expertise needed to realize that you can do this and then provide support for it is neither common nor cheap.
If you had a tower on a nice highly-visible location (say, the giant Jesus statue in Rio) and could do line-of-sight laser links, that would be an extremely cost-effective solution. Sadly, that sort of opportunity is rare.
So it is disingenuous to complain that your top speed means you burn up your allocation. It's unlikely anyone is happy to pay per Mbps, so they gotta work out something which requires compromise.
Now, the limits are tiny, the pricing nuts, etc., but it's not fundamentally a terrible model. On cell data, I'm very happy to buy per GB so I can decide how much to spend.
The real downside is that capping users or offering low cap plans stifles innovation. It sounds nice to be able to "pay what you use" but that has a negative impact overall.