←back to thread

The AI Investment Boom

(www.apricitas.io)
271 points m-hodges | 3 comments | 20 Oct 24 14:56 UTC | HN request time: 0s | source
Show context
m101 ◴[20 Oct 24 21:56 UTC] No.41898811[source]▶
There is a comment on this thread about this being like the railroads, but this is nothing like the railroads except insofar as it costs a lot of money.

The railroads have lasted decades and will remain relevant for many more decades. They slowly wear out, and they are the most efficient form of land transport.

These hardware investments will all be written off in 6 years time and won't be worth running given the power costs and relative output. They will be junked.

There's also the extra risk that for some reason future AI systems just don't run efficiently on current gen hardware.

replies(5): >>41899046 #>>41900169 #>>41900225 #>>41903674 #>>41905422 #
1. segmondy ◴[21 Oct 24 02:38 UTC] No.41900225[source]▶
Maybe, maybe not. Nvidia P40's GPUs were released 8 years ago, they are hot. Prices have doubled in the last year. Nothing will be written off in 6 years. Folks will be using A100 15 years from now. 7 yrs old V100 32gb GPU cards are still going for $1500 on ebay. We are more likely to invent a more efficient software architecture than invent a new type of better hardware and replace everything that exists.
replies(1): >>41904817 #
2. causal ◴[21 Oct 24 14:51 UTC] No.41904817[source]▶
And more importantly, the GPU is the most replaceable part of the datacenter infrastructure. Power, buildings, cooling, security, cabling, internet backbone connectivity, etc. are the harder to swap pieces that require time to build out.
replies(1): >>41905553 #
3. klabb3 ◴[21 Oct 24 16:03 UTC] No.41905553[source]▶
Right but that’s just digital infrastructure, and of course that will have positive side effects, no matter what role AI has in the future. Rail otoh is the infrastructure.