$500B is a lot of debt, it's comparable to the three largest car companies' debt.
It's still not super huge compared to the amount of debt in other industries, but I guess the thought is it's riskier?
It's still not super huge compared to the amount of debt in other industries, but I guess the thought is it's riskier?
We spent this much (without adjusting for inflation), in 5-10 years on telecom build out in the lead up to the dot com crash. I'm fairly sure that there is still leftover capacity in the ground (dark fiber) today that we can leverage.
Smell like a bubble yet?
Looking back to that pre 2000's era, SUN was running on 50% margins, Cisco at 68%
Nvidia, 70% (and MS openly admits that they have GPU's on shelves not making money: https://www.datacenterdynamics.com/en/news/microsoft-has-ai-... ) Micron (memory) 70%, SK Hynix (SSD's) 70%.
For as much fun as the dot com bubble was, for as hard as the pop was, what came after was MUCH better. This burst is going to be brutal, and the sooner it happens the sooner we can move on to actual (sane) innovation, that leverages this build out.
Unlike dark fiber which retained value, GPUs and compute hardware rapidly becomes obsolete. It won't make sense to use anything more than 4 years old, as performance and especially energy efficiency will have improved in that time.
I wouldn't count on getting significant value out of the debris of the coming data center implosion.
Copper thieves are likely the only ones who might, but there will likely be a glut and the risk likely won't be worth it for even them.
Historically this would be (somewhat) accurate for CPU's. However there are already companies that "unrack" at the 10 year mark for CPU based compute.
The problem is that performance per watt isn't going to go UP, and that matters. More power means more cooling, and harder to maintain (liquid over air).
I suspect that much of the hardware installed today, is going to be running almost unchanged in a decade.