Network capacity, at least in most regards, currently dwarfs the expansion that happened back in the late 90s (unless you're talking about some later "dot-com bubble"). I'm not sure how you can argue that we're still trying to fill up bandwidth built in 1999. Sure, there was some build out during the bubble...but, I don't see any reason to think network infrastructure is going to shrink or that costs will increase remarkably.
Makes sense that my specifics would be out of date. I'm not arguing that network infrastructure will shrink; I'm observing that the scaling factors for network infrastructure investment are not nearly so favorable as the scaling factors involved in CPU or storage capacity. We can already build hardware with capabilities greatly in excess of our actual needs. When network capacity becomes a bottleneck, it will be much cheaper to take advantage of all that latent capacity at the edge than to invest many billions increasing network capacity, and we'll see yet another pendulum swing away from thin clients toward fat independent hardware devices. Then the market will change and lots of money will flood into network infrastructure, and for a few years the pendulum will swing back again. It's been going back and forth like this for decades.
Network capacity is a continual bottleneck, and many billions are spent to upgrade the transport network yearly.
Transport signals have gone from 2.5G to 10G to 40G to 100G to 200G per wave, and from 8 waves to 16 waves to 40 waves to 96 waves to 100+ waves per fiber.
The bottleneck has existed for a decade at least, and is invested in to keep pushing it out at the same time all this other stuff is happening.
Hmmm. Interesting that speeds at the edge don't appear to have changed much at all, in years. I didn't think there was much ongoing coverage expansion going on. What's driving the capacity investment? Greater utilization of existing connections? What's using up this capacity?
