I acknowledge your point about alternate use, but we also need to look at a datacenter we may or may not need as a “power consumption plant.” These jackasses just keep loading and loading up the grid, looking to make a private dollar on public infrastructure. It’s wasteful and not necessarily a baseline good thing ^TM even if AI goes flop.
Comment on How does the private equity bubble compare to the AI bubble if at all?
litchralee@sh.itjust.works 3 days ago
I’m not going to come running to the defense of private equity (PE) firms, but compared to so-called AI companies, the PE firms are at least building tangible things that have an ostensible alternative use. A physical data center building – even one located far away from the typical metropolitan area that have better connectivity to the world’s fibre networks – will still be an asset with some utility, when/if the AI bubble pops.
In that scenario, the PE firm would certainly take a haircut on their investment, but they’d still get something because an already-built data center will sell for some non-zero price, with possible buyers being the conventional, non-AI companies that just happen to need some cheap rack space. Looking at the AI companies though, what assets do they have which carry some intrinsic value?
It is often said that during the California Gold Rush, the richest people were not those which staked our the best gold mines, but those who sold pickaxes to miners. So too would PE firms pivot to whatever comes next, selling their remaining interest from the prior hype cycle and moving to the next.
I’ve opined before that because no one knows when the bubble with burst, it is simultaneously financially dangerous to: 1) invest into that market segment, but also 2) to exit from that market segment. And so if a PE firm is already bet most of the farm, then they might just have to follow through with it and pray for the best.
scarabic@lemmy.world 2 days ago
litchralee@sh.itjust.works 2 days ago
Used for AI, I agree that a faraway, loud, energy-hungry data center comes with a huge host of negatives for the locals, to the point that I’m not sure why they keep getting building approval.
But my point is that in an eventual post-bubble puncture world where AI has its market correction, there will be at least some salvage value in a building that already has power and data connections. A loud, energy-hungry data center can be tamed to be quiet and energy-sipping based on what’s hardware it’s filled in. Remove the GPUs and add some plain servers and that’s a run-of-the-mill data center, the likes of which have been neighbors to urbanites for decades.
I suppose I’d rehash my opinion as such: building new data centers can be wasteful, but I think changing out the workload can do a lot to reduce the impacts (aka harm reduction), making it less like reopening a landfill, and more like rededicating a warehouse. If the building is already standing, there’s no point in tearing it down without cause. Worst case, it becomes climate-controlled paper document storage, which is the least impactful use-case I can imagine.
BlameThePeacock@lemmy.ca 3 days ago
It’s important to note that in some previous bubbles, the leftovers of the crash ended up spurring new beneficial growth after.
GPUlike computing power available at scape for essentially free after the ai crash could be used in all sorts of potential ways.
Maybe it makes rending movies with special effects super cheap, and available even to tiny indie studios. Maybe scientists grab it for running physics simulations or disease treatment computations.
gravitas_deficiency@sh.itjust.works 2 days ago
The problem is that the deprecation/obsolescence/lifetime cycles of GPUs is WAY more rapid than anyone in the “AI” circlejerk bubble is willing to admit. Aside from the generational upgrades that you tend to see in GPUs, which make older models far less valuable in terms of investment, server hardware simply cannot function at peak load indefinitely - and running GPUs at peak load constantly MASSIVELY shortens the MTBF.
TL;DR: the way GPUs are used in ML applications mean that they tend to cook themselves WAY quicker than the GPU you have in your gaming machine or console - as in, they often have a couple of years lifetime, max, and that failure rate is a bell curve.
BlameThePeacock@lemmy.ca 2 days ago
You’re pulling shit out of your ass at this point, there are some doom reports out of people suggesting that may be a problem, but there are also reports out of other companies(meta for example) with documentation saying the rate is much lower and the mean failure is 6+ years.
The other leftovers from the crash also won’t have that problem. It’s not just about GPUs. Datacenters and their infrastructure last a lot longer, and the electric generation/transportation networks will also potentially be useful for various alternative applications if the AI use case flops.
gravitas_deficiency@sh.itjust.works 2 days ago
MTBF is absolutely not six years if you’re running your H100 nodes at peak load and heat soaking the shit out of them. ML workloads are particularly hard on GPU RAM in particular, and sustained heat load on that particular component type on the board is known to degrade performance and integrity.
As to Meta’s (or MS, or OpenAI, or what have you) doc on MTBF: I don’t really trust them on that, because they’re a big player in the “AI” bubble, so of course they’d want to give the impression that the hardware they’re using in their data centers still have a bunch of useful life left. That’s a direct impact to their balance sheet. If they can misrepresent extremely expensive components that they have a shitload of as still being worth a lot, instead of being essentially being salvage/parts only, I would absolutely expect them to do that. Especially in the regulatory environment in which we now exist.
Munkisquisher@lemmy.nz 2 days ago
Cloud compute was attractive to 3d rendering for a while, as you could put your non urgent renders on the cloud at the lowest priority and take advantage of off peak pricing. Now model training demand has wiped out off peak pricing and forced the cloud rendering cost way higher than rendering locally.
fullsquare@awful.systems 2 days ago
gpus as used for genai aren’t really suitable for normal loads like aerodynamic simulations, genai uses low precision data like fp8, fp4, blackwells and such are optimized for it so hard that you can’t really do anything else on this thing
BlameThePeacock@lemmy.ca 2 days ago
They’re still somewhat functional for those workloads, and they can even use those low-precision components to emulate high-precision using libraries like cublas, though obviously not as fast as hardware that could do it natively.
It’s not like they can’t do it at all. It’s just that Hopper was better at FP64 than Blackwell is but if Blackwell chips become effectively free due to an AI crash then you could likely still use them in that capacity.
fullsquare@awful.systems 2 days ago
Or you could use different hardware, maybe from competition, because result isn’t worth electricity it used
litchralee@sh.itjust.works 3 days ago
Absolutely, yes. I didn’t want to elongate my comment further, but one odd benefit of the Dot Com bubble collapsing was all of the dark fibre optic cable laid in the ground. Those would later be lit up, to provide additional bandwidth or private circuits, and some even became fibre to the home, since some municipalities ended up owning the fibre network.
In a strange twist, the company that produced a lot of this fibre optic cable and went bankrupt during the bubble pop – Corning Glass – would later become instrumental in another boom, because their glass expertise meant they knew how to produce durable smartphone screens. They are the maker of Gorilla Glass.
dustyData@lemmy.world 2 days ago
Rack space is literally the only thing valuable that would be left. Those GPUs are useless for non LLM computation. The optimization of the chips and the massive amounts of soldered RAM. They are purpose made, and they were also manufactured very cheap without common longevity and endurance design features. They will degrade and start failing after less than 5 years or so. Most would be inoperable in a decade. Those data centers are massive piles of e-waste, an absolute misuse of sand.
litchralee@sh.itjust.works 2 days ago
Racks/cabinets, fiber optic cables, PDUs, CAT6 (OOBM network), top-of-rack switches, aggregation switches, core switches, core routers, external multi-homed ISP/transit connectivity, megawatt three-phase power feeds from the electric utility, internal power distribution and step-down transformers, physical security and alarm systems, badge access, high-strength raised floor, plenum spaces for hot/cold aisles, massive chiller units.
Feyd@programming.dev 3 days ago
Running those power hungry gpu based data centers is going to cost beaucoup bucks regardless of the usage.
BlameThePeacock@lemmy.ca 3 days ago
The power costs are nothing compared to the hardware costs for those things.
Getting the massive power required is difficult for datacenters but the usage per unit of compute is actually quite low.