AI datacenter boom could end badly, Goldman Sachs warns

Goldman Sachs warns that datacenter investments may fail to pay off if the industry is unable to monetize AI models, but hedges its bets by saying that demand could also overwhelm available capacity by 2030. Investors have been pouring cash into datacenters, buoyed by the AI-driven demand for compute resources that has led to …

[m4r35n357]

It is!

. . . and yet it is not! Ker-ching!

[bigphil9009]

Re: It is!

What an incredible business to be in!

[Snowy]

Re: It is!

Goldman Sachs, which reported record revenue for Q3 2025 of $15.2 billion, up 20 percent year-on-year, cautions that its report is for educational purposes only, and does not constitute an investment recommendation.

Kind of makes the whole report meaningless.

[b0llchit]

Re: It is!

The report is simply an ad with a disclaimer attached. The advertised scenarios are projected to optimize Goldman Sachs' profit margins and they profit regardless which scenario will unfold.

[David 132]

Re: It is!

They only call themselves "analysts" because "haruspex" is harder to pronounce.

[Paul Crawford]

Re: It is!

I had to look that up, and indeed it seems as good a guess as any...

[David 132]

Re: It is!

"The More You Know!" :)

From back in the days when a prediction was just a gut feeling...

[grumpy-old-person]

Re: It is!

Anal-ysts?

[JimBz]

Re: It is!

"You can't sue us if we're wrong, even if it destroys your business." It's a disclaimer. AI changes every week. It's currently in a phase of competition for capability, competition for efficiency will come.

[Anonymous Coward]

Re: It is!

I look forward to their view of quantum computing...

[Anonymous Coward]

Re: It is!

Yeah, that and microchipped quantum cats, contained in a blue carrier, that seemingly both exist, and don't, simultaneously, for hours on end ... ;)

[Tron]

Tulips, railway mania etc.

GAFA are betting on an exponential rise in income from AI. Businesses don't have that cash.

As for users, well. I wouldn't use copilot if MS paid me to.

Investors don't seem to realise that the very pricey chips that they are investing in depreciate rapidly from before they are turned on and will be obsolete real soon, requiring replacement. GAFA bilk customers with subscriptions. It will be novel for investors to be treated the same way.

Goldman Sachs are amusingly offering a pick-your-own future for investors to choose from.

They might want to consider how lucrative a product might be if it has to be forced on people who don't want it and wont pay for it.

[vtcodger]

Re: Tulips, railway mania etc.

"GAFA are betting on an exponential rise in income from AI. Businesses don't have that cash."

And the few that do have the cash -- Disney, MGM, and a handful of others -- will very likely build their own data centers.

[Evil Scot]

Re: Tulips, railway mania etc.

Don't MGM have their own Data Centres already. From Amazon.

[Anonymous Coward]

Re: Tulips, railway mania etc.

Goldman Sachs will have packaged them up in ETF’s, derivatives of ETF’s and exited long before this.

The ‘10%+ easy return/no effort suckers’ suckers on eToro, Revolut, Monzo and other buy, sell, make a return doing nothing leeches …: will eat the shit when it bursts.

[MyffyW]

Re: Tulips, railway mania etc.

Goldman Sachs are amusingly offering a pick-your-own future for investors to choose from.

Which, interestingly, is one of the few tasks that I would trust Generative AI with.

[elDog]

It would be a shame to have all that land, water, energy, technology, human effort wasted.

(Looking for the sarcasm thingee.)

I see dystopian pictures of tumbleweeds rolling by huge rusting/decaying building in the middle of nowhere. I also see a lot of governments (small and large) being left with huge holes in their budgets and nobody can find the original investors or their attorneys.

[Mimsey Borogove]

Re: It would be a shame to have all that land, water, energy, technology, human effort wasted.

I see dystopian pictures of tumbleweeds rolling by huge rusting/decaying building in the middle of nowhere. I also see a lot of governments (small and large) being left with huge holes in their budgets and nobody can find the original investors or their attorneys.

I am almost 100% sure that that's what will happen. And there will be lots of holes in lots of budgets. I talked to my "finance guy" the other week and told him I don't want my money going to anything involving crypto or AI. He said that avoiding AI is difficult, since almost all companies are doing something with it. With only 5% of users actually paying for it (per Ed at www.wheresyoured.at), where are the AI companies expecting all this growth to come from? Will they be able to get everyone so addicted to it that they'll be willing to pay the stupendous amounts it actually costs to run?

All the municipalities that are turning against data centers in their back yards will be so glad they did, to avoid your dystopian future.

[Filippo]

>Goldman Sachs warns that datacenter investments may fail to pay off if the industry is unable to monetize AI models, but hedges its bets by saying that demand could also overwhelm available capacity by 2030.

"It might go very bad, or it might go very well. Or it might go just bad, or just well. Unless it goes so-so."

And someone got paid, probably a lot of money, to write this?

[Kurgan]

No, it was made by AI. I'm quite sure about it.

[Roland6]

>Goldman Sachs warns that datacenter investments may fail to pay off if the industry is unable to monetize AI models

Sounds rational, given the increasing lack of evidence of service improvements.

> but hedges its bets by saying that demand could also overwhelm available capacity by 2030.

Also sounds reasonable; the hype is driving up “demand”: you mustn’t miss this bandwagon…

As we know there is a lengthening lead time on both building datacentres (construction lead-times) and filling them with “AI” ready servers (chip production lead times and shortages).

So both statements look like being true.

Bottom line 1: if you haven’t already got your AI datacentre up and running, you’ve largely missed the bandwagon and being able to cash-in on it.

Bottom line 2: unless you have a clearly identified opportunity for AI, best place a watching brief on it and just focus on improving your customer focus and experience, which will probably deliver greater margin improvement and retention than upsetting everything in the rush to implement AI everywhere.

[Throg]

Vampire Squid hedging their bets? You don’t say!

[trevorde]

Better investment

Tulips!

[Rich 11]

Re: Better investment

Threelips!

Oops, sorry, I thought we were doing generative AI.

[Paul Crawford]

Re: Better investment

I prefer degenerate AI

[Anonymous Coward]

Who needs Goldman Sachs?

Any fule no it's:

1) get AI

2) profit

[Sorry that handle is already taken.]

Gee...

...you think?

[xyz123]

Many LLMs are actually BETTER on a local machine. They're slightly slower than when the 'free' AI isn't under peak load, but faster than when it is under load from millions of queries.

Also Local LLMS are as censored as you want them to be......no US company deciding that today pictures of cucumbers are too racy to be generated.

[druck]

Build up

They need to be building taller bit barns for the investors to throw themselves off when the bubble bursts, you aren't going to get a decent splat from the current designs.

[Roland6]

Re: Build up

Well given, the sorts of technology being discussed such as satellites beaming sunlight to earth. Could explore using the lower temperatures of the atmosphere (mesosphere and below) as a source of cold air for cooling…

This would require some very tall structures which could have lifts attached, I think these might provide a suitable jumping point…

[Norm DePlume]

Analysts advise that tossed coin might land on heads.

But there is the possibility of it also landing on tails.

[LBJsPNS]

The hell you say. No wonder G-S is so revered for something or another. Their prognostication is stunning, don't you think?