This is a raw idea in progress.
In the early days, audio was always uncompressed. During the 90's audio compression started to get reasonable, and mp3 came to be a popular format. This happened around the time people started to get broadband, and the two created an explosive mixture of bandwidth growth.
But that bandwidth growth self-capped, and if u look @ downloaded audio today, its largely the same. The fidelity gap between 128kbps VBR mp3 and uncompressed audio is small enough that it just didn't matter. Thus the bandwidth growth due to fidelity capped itself. The second thing that happend is that the number of songs people could listen to (the amount of information they could process) capped out, and thus the overall amount of bandwidth due to song-swappers became driven only by number of users, no other variable.
Video is going the same direction as audio. It turns out that 1080p H.264 is sufficiently close to 'uncompressed live analog' that there's not as much point in going further. Living rooms aren't getting larger, so we don't need more resolution. The gains due to compression are slowing, and bandwidth growth is again being driven by number of users only.
This could mean that the overall internet bandwidth will self-limit to the rate of information processing a human brain can do live. There is probably some number on that somewhere of so many bits-per-second.
This could mean that the internet will eventually be 'done' when we each have a non-oversubscribed '1xhuman information rate' link to the back of our skull.
This might be approaching sooner than we think.
And lest the naysayers say "what about the collector, storing to disk only", well, the size of storage is also going to asymptotically approach molecular densities and stop growing too.