I understand that breakthroughs such as multiplexing caused a huge glut in the amount of fiber laid. But what prevented demand from increasing to use it?

I watched multiple fiber companies implode and my friends laid off in that time period. It was not a lack of demand but the unrealized not properly studied cost of digging in cities and associated time delays and extra costs and extra extra unforeseen costs by companies new on the scene. Some were warned by telco and warnings were ignored, funding was obtained by greedy short sighted people. This was in medium to large cities where there was already a significant footprint of other infrastructure that prevented trenching in new fiber and cross many different boundaries.

I can't really answer your question about AI.

▲

AbstractH24 13 hours ago | parent [-]

> unrealized not properly studied cost of digging in cities and associated time delays and extra costs and extra extra unforeseen costs by companies new on the scene

Was that unrelated to the potential ROI imploding as existing supply grew more efficient? That even if prices stayed high it still wouldn’t have been worth what it cost to overcome obstacles?

If so, sounds like a reminder that if things had gone “according to plan” the oversupply would have been exponentially greater. Something i never considered.

▲

Bender 13 hours ago | parent [-]

That even if prices stayed high it still wouldn’t have been worth what it cost to overcome obstacles?

They would have required significantly more funding and this was even before they got through all the permitting costs. There never would have been an over-supply as most of them could not even get through permitting much less all the approvals to start trenching up the roads or sharing existing pipes with other companies. Eventually over a long period of time individual data-centers and telcos worked through this but the initial idea of making a profitable business of running fiber in 2000/2001 was a pipe dream pun intended.

There was plenty of demand and plenty of data-centers popping up all over the place. I helped manage one around that time. It wasn't my primary role but we went from about 700 people to 6 in 6 years. A lot of businesses at that time were started by people with big hopes of getting rich and very little business experience.

▲

AbstractH24 13 hours ago | parent [-]

Woof.

Maybe you have perspective on this — how much of way tech reshaped life during the 2010s only occurred because of the over-builds of the dot-com era?

Again, makes me wonder what the parallels are to today. Imagine timelines will compressed and/or the level of shock will be even greater, but regardless the real impact of AI on society will only be seen after a broad-world-wide socio-economic correction the likes of which we saw in the early to mid 2000s)

	▲	Bender 13 hours ago \| parent [-]
		The only over-builds I saw were primarily done for making a "dog and pony show" to investors and prospective customers. By over-build I mean a lot of data-center space was built out, many of the data-centers built out large highly redundant ATS/UPS with numerous generators and fuel tanks. Some data-centers had redundant commercial power coming in from 2 to 3 sub-stations when feasible and multiple egress to local telco and many redundant circuits that by todays standards would be very low bandwidth. Most of this was for show and caused many companies to file chapter 11 several times, change names several times, merge with several companies several times and so on. Over time circuits got faster, servers got faster and used less power. That's about it. Many of those build-outs are still around, just different butts in chairs and far fewer of them. I don't know what the parallels will be compared to today because I don't know the ultimate goals for government, military and civilian use of AI {AGI}?. You may enjoy listening to this podcast [1] for some idea of why people are scrambling and why it's probably premature to even guess where all of this will land. [1] - https://www.youtube.com/watch?v=NufB1LL_rCU [video][2 hours 7 mins]