Saturating even a 100 megabit connection to a cloud provider gets really expensive, really fast. That's 45 gigabytes/hr. Only a subset of power users are going to have daily or even weekly sustained full-speed transfers of that size, and 100 Mbps is usually the bottom tier plan, from a US telco.
Even if you are doing that to initially back up your network or install Call of Duty, after the initial usage you're only making differential updates, unless something goes very wrong, which hopefully doesn't happen even once a month.
For some users, it may be worth an extra $1000/yr to get the gigabit plan instead of 100 megabit, so that a Call of Duty install could theoretically happen in tens minutes instead of an hour or two, but really it's going to take an hour for your computer to decompress and write the data. Everyone else paying that $1000/yr extra is just wasting their money.
Also, Netflix only uses 16 Mbps at most, for a 4K video which is only a small portion of their catalog and only available to their top-tier subscribers. The extra bandwidth is a recommendation to allow for other users to still use the network and also to account for a special case of high latency that used to occur when saturating the connection on the fastest plans some ISPs offered. Here's some good research into actual streaming usage: https://www.wsj.com/graphics/faster-internet-not-worth-it/ (It's WSJ, but there's no paywall)
On top of that, the cheapest plans offered by most telecoms in the US are over 100 megabit, so you still get a guaranteed three 4K streams Netflix without issue, but realistically could pull off double that.
You're right -- saturating a 100Mbps connection is really hard. But I'm not talking about running servers from one's home. The goal isn't to saturate your connection for the month, it's to make it so that if your usage "wants" more bandwidth, it's there for you. I don't know about most people's usage, but I can say that Docker & NPM in particular benefit from a very fast connection.
> which hopefully doesn't happen even once a month.
This is the mindset I was referencing. Nobody thinks about how often their computer has to write to its local disk, and if you have a sufficiently fast Internet connection you can adopt the same carefree attitude towards Internet use.
> an extra $1000/yr
This is the disconnect. In my area, the price gap is much smaller, less than half that. The bigger consideration for us was going to a gig connection was also the only way to get a connection without a usage cap and therefore a predictable bill (Xfinity is the other provider).
Incidentally, it looks like the WSJ test used mobile devices like phones and tablets. AFAIK all the major streamers deliver lower-bandwidth content to those devices by default, so this is not the best test.
Even if you are doing that to initially back up your network or install Call of Duty, after the initial usage you're only making differential updates, unless something goes very wrong, which hopefully doesn't happen even once a month.
For some users, it may be worth an extra $1000/yr to get the gigabit plan instead of 100 megabit, so that a Call of Duty install could theoretically happen in tens minutes instead of an hour or two, but really it's going to take an hour for your computer to decompress and write the data. Everyone else paying that $1000/yr extra is just wasting their money.
Also, Netflix only uses 16 Mbps at most, for a 4K video which is only a small portion of their catalog and only available to their top-tier subscribers. The extra bandwidth is a recommendation to allow for other users to still use the network and also to account for a special case of high latency that used to occur when saturating the connection on the fastest plans some ISPs offered. Here's some good research into actual streaming usage: https://www.wsj.com/graphics/faster-internet-not-worth-it/ (It's WSJ, but there's no paywall)
On top of that, the cheapest plans offered by most telecoms in the US are over 100 megabit, so you still get a guaranteed three 4K streams Netflix without issue, but realistically could pull off double that.