ISPs operate on a similar business model to gyms. They couldn't afford to provide the service if everyone maxed out what they are paying for. Allowing on demand increases of bandwidth would break this business model.
Wholesale bandwidth is 5 cents per Mbps or less at scale. Not that the ISP pays for all bandwidth, peering and CDN traffic is free.
In other words, even if you max out a 1 Gbps line 24/7, you are only costing the ISP $50 per month.
Add to that that most people are unable to consume even 10 Mbps on average and bandwidth costs become less and less of an issue. An average subscriber will use 2-4 Mbps of bandwidth during primetime, and primetime is all you care about as an ISP.
Nothing in this means that offering bandwidth on demand is impossible or a threat to the ISP business model.
It’s just that the overhead of offering an on-demand product to consumers isn’t worth the time and effort, as you’d be making peanuts while having to deal with a lot of grief.
Bandwidth on demand is a thing for businesses and operators. A good example of this is Megaport, who offers virtual cross connects and transport. Price start from $100 per month plus a varying fee per Mbps, from a few cents to a few dollars per month.
4k Netflix is around 15 Mbps alone, even 1080p is >5 Mbps.
And with games being now > 100 GB downloads these days, I would expect that to contribute meaningfully too to the bandwidth usage even if average person probably doesn't quite download new games daily.
Netflix is only about 30% of all Internet bits, and there's a limit to how much Netflix and chill anybody can do.
3 GB / hour to Mbps = 6.667 Mb/s https://www.wolframalpha.com/input/?i=3+GB+%2F+hour+to+Mbps
7 GB / hour to Mbps = 15.56 Mb/s
In bandwidth sure, but the world isn't wired that way. There's a bottleneck somewhere between the ISP and the customer, that can't support all customers on that link running maxed out. In the gym analogy, even if the gym has 1 treadmill per customer, that doesn't help the shortage of weight machines. An ISP may choose to (or choose not to) invest in adding capacity to remove bottlenecks, but that's additional capital outlay that may not be prioritized or budgeted for.
You can support 5x the bandwidth at 50% of the cost and 20% of the labor than you could in 2005-2010.
And, as I wrote before, in no real life scenario will all 100% of subscriber be running their line at full tilt 24/7.
Engineering a non-blocking 1 Gbps network isn't particularly hard or expensive using FTTH.
My ISP is a "small" provider that focuses on multi-unit urban residential buildings in the midwest. Their business model is to make an agreement with the building management and provide a baseline service to every unit, while offering upgrades and extra services to individual units that want them. They have an incentive to provide far more capacity than the baseline so they have room to offer upgrades.
I pay $25 per month for 250mb symmetrical and am definitely getting those speeds 24/7. I have the option to upgrade to 1G symmetrical for something like $500 per year, but haven't seen the need.
Unless an ISP is serving a very low density customer base, the cost of building out full capacity just isn't prohibitive anymore. My ISP only has around 20,000 customers and they have direct peering with Netflix, Amazon, Apple and others - clearly an ISP doesn't even have to get very large before those cost benefits arrive.
For example, Brazil literally has thousands of tiny ISPs, each with, at most, a few hundred customers. They operate on the narrowest of margins, squeaking out close to no profits. They cut corners everywhere. They definitely do not have spare capacity, and over sell by insane percentages.
Another example, Mongolia, where there's basically just one ISP, the government owned Unitel. They have zero incentive to do anything extravagant, or even provide good service. Its a monopoly.
The cost of building out full capacity is very much still a limiting factor in many parts of the world.