Not oversubscribing is a cost multiplier at every level. 1 million 1 Gbit customers in a city is going to need 1000 100Gbit connections out of that city and the same for transit, and that will have no impact on pricing? And everything is on average used at 1% of capacity.
If my ISP can only afford to supply me with 1TB of transfer at 1Gbit, that's fine. They can put it in the adverts, the contracts, and the pricing. For customers who want 10TB of transfer, they can offer a higher cost option.
And if they choose to gamble, advertising and entering into contracts promising "unlimited data", which they think will be more profitable across their entire customer base? Then they've got to do supply what they promised in the adverts. They chose to gamble that way, and if they lose money gambling that's their business.
You have that on mobile subscriptions usually, heavy users pay more and low usage users are not subsidizing them.
I take you are fine with paying 10x or even more for your no oversubscription Internet connection then?
Oversubscription is not gambling. The way it works after your last mile connection is that ISPs look at link usage in their network, city level distribution, city to city, transit, peering, etc, once it reaches 60-80% utilization at peak you start looking at adding more capacity. Bad ISPs (most US ISPs) will let this go too far though.
That's not the same thing. A more adequate comparison would be to say that you promised 30 people they can have a burger, but can only produce 5 burgers per minute. If everyone show up at exactly the same time, you won't be able to satisfy them all (they'll have to wait). But overall you can consider that the probability of such thing happening is small enough to take the "gamble".
It might be if traffic had sudden jumps of like 30%, but it doesn't and there is headroom available. Traffic increases slowly over time and you have plenty of time to upgrade your network.
Not oversubscribing is a cost multiplier at every level. 1 million 1 Gbit customers in a city is going to need 1000 100Gbit connections out of that city and the same for transit, and that will have no impact on pricing? And everything is on average used at 1% of capacity.