Even banks and hyperscalers are now sounding the alarm about the AI bubble

KETTLE From international banking worries to the market state of canary-in-the-coal-mine Oracle, the AI bubble is sure looking taut. The Bank for International Settlements, often referred to as "the central bank for central banks", said in a report at the end of June that it was worried the AI bubble was nigh on to popping and …

[Wang Cores]

Still won't pop.

[Stuart Castle]

It will when enough people realise that the only thing keeping their investments valuable is the fact they are investing more, and they start to panic and pull existing investments. It's a ponzi scheme. These usually collapse eventually when people realise the company they are investing in will never make any profit.

[Wang Cores]

You think Musk is propping up Reform because he's a legit white nationalist? fuck no. He just knows fascist loons are a cheap date and can be leveraged against the rest of us when the ponzi collapses.

[O'Reg Inalsin]

Recycling Bulimia Nervosa Emissions

It will pop so the debt can be written off and the assets sold back to the original perps for cents on the dollar with govt money fronted to "Rescue the Economy". The game of musical chairs will have some losers too - mostly small but even some big ones (remembering fearless Capt Larry "Lehman" Ellison - whose boat sunk so others could float).

[Anonymous Coward]

Go on

Pop already. So I can say "Told you so" to managers at work.

[SoSoTired]

Re: Go on

And so my kid’s school can stop trying to pRePaRe tHe kIdS fOr tHe FuTurE with chatbots. The sooner this pops the sooner tech idiots running schools won’t be fooled by these slop machines

[pip25]

Was the US government this involved in the Dot Com bubble?

That's the part that concerns me. Even if the bubble bursts, the government may fight to keep the whole thing alive... somehow. Chances are, that could make the fallout even worse.

[Steve Davies 3]

Re: Was the US government this involved in the Dot Com bubble?

Don't you mean...

Was the Trump family involved in this Dot Com bubble?

Then the answer is yes, a BIGLY Yes. All those non compete contracts.... Perfect for filling their Qatari bank accounts.

[BartyFartsLast]

Re: Was the US government this involved in the Dot Com bubble?

Nothing makes money for the super wealthy like a financial crash

[Cris E]

Re: Was the US government this involved in the Dot Com bubble?

@pipa

It should concern you, since "the government" is more or less equivalent to D Trump's latest whim, and that is almost always subject to the influence of last person who told him how he will personally profit from their advice.

Consider that Oracle is already deep into doing business with this administration and its approval of the massive Warner Bros purchase. As recently as 2024 David Ellison was dropping nearly $1m into Democratic hands and a year later he's Trump's best friend and a major donor. He believes nothing, of course, but he and Trump share an understanding of bribery so I suspect that most government intervention is going to go down the ORA sink hole to cushion the Ellisons' family transition from primarily software to a more diverse portfolio. Color me uninterested in the investment potential on that front.

[steelpillow]

Reality douche

It is important to distinguish between the AI bubble and the AI investment bubble. The latter depends on second-guessing the former. The wise investor splits their stake three ways between; AI and investment bubbles making it through, investment bubble crashing, and double-bubble-trouble post-burst profit centres. Proportional splits to be based on the entrails of the nearest futurologist.

We need also to remember that no snapshot means a lot; everything is massively dynamic, accelerating, and the landscape by Christmas as yet barely recognisable. Global world models, semantic associations, cognitive behaviours, all these things are beginning to appear in production autonomous and semi-autonomous vehicles and other bleeding-edge applications. By end-of-year they will have solidified into the next new normal.

So, who is going to ride the wave, who will fall off, and who will sit in the bar drinking it dry? And how will the financial fashionistas attempt to second-guess all that? Ladies, gentlemen and Rainbows, place your bets.

[Ian Johnston]

Re: Reality douche

Nah. It's still all just error prone autocomplete, however many silly buzzwords get thrown at it. Cognitive behaviours my arse.

[Bebu sa Ware]

Re: Reality douche

"Cognitive behaviours my arse."

Yes.

That part of your anatomy exhibit more cognition than the whole AI circus.

I am not saying that it is all shit… silly me… actually that is what I am saying.

[Anonymous Coward]

Re: Reality douche

The wise investor has already gotten out months ago, and put funds into recession proof stocks..

[c1ue]

Re: Reality douche

No, sorry, you have totally missed the plot.

The LLM bubble - because LLMs are not actually AI in any sense except OpenAI marketing - will implode because of crappy unit economics. Frontier model LLMs are simply huge, boil the ocean, inefficient and therefore expensive boondoggles.

Equally, the "investment bubble" isn't infrastructure telecom builders overinvesting in fiber like Y2K; it is hyperscalers initially burning profit to maintain P/E multiples but now taking on massive debt and even selling equity to do the same. While Microsoft, Google and Amazon aren't going to go tits up because of this idiocy - their fate will certainly be much worse than Cisco's, a decade after Y2K because they won't even be making more money as Cisco did. This is Apres Moi, Le Deluge on a multinational scale.

[spireite]

don't want to sound like Donald...

... but I cannot wait for the 'big beautiful implosion'

[Gavsky]

Humans still creating economic 'bubbles' even when history shows us the danger. Greed, lack of understanding or foresight, a herd mentality - all these & more, over & again. Bit coin scam, tulips or dot com, everyone shakes their heads afterwards & wonders just how intelligent we really are. Some get very rich, plenty of others negatively so...

[FIA]

Humans still creating economic 'bubbles' even when history shows us the danger.

It's not a co-ordinated effort. They're not 'created'.

It's the functional outcome of human behaviour. It will always happen.

everyone shakes their heads afterwards & wonders just how intelligent we really are.

Yet we all continue to do the same things. Thing is when we personally do it we justify it to ourselves, as 'It's not the same'. ;)

Many of us expect other people's behaviour to change, but not our own.

[Tron]

A bubble is a reasonably reliable market.

Get in at the start, get out as it plateaus. It's more reliable than normal stocks and easier to get rich off, which is why everyone tries to create them, spot them, get in early and keep them going. Those of us with no cash just get to sit and watch and be ignored as we point out how fake they are.

[Will Godfrey]

Can't come soon enough

It's already cost the world far too much... in just about every respect.

[Luiz Abdala]

I got a hammer, find me nails.

AI is still a solution looking for a problem in most cases, it takes massive resources to work, very few projects showed an honest ROI (folding proteins, maybe? Big Pharma? The protein folding is speculative, so AI is perfect to generate models that can be checked later by conventional means) and I don't understand how the bubble didn't POP yet.

Yeah, for those that can lever AI, great. For the large majority of the rest of the mankind, not so much. The strain on memory chips and what not barely justifies investment.

People are still in the phase "this new tech is a hammer, find me some nails" enthusiastic adoption philosophy, which is very bad.

[cyberdemon]

Re: I got a hammer, find me nails.

Except "folding proteins" is NOT "AI", it's computational chemistry which has been around for decades.

So-called AI is about generating "content" (ie images, video, audio, text) based solely on a statistical analysis of what has gone before. It is self-referential navel-gazing which can fundamentally never solve any problem that has not been solved before. All it can do is apply a past solution to a slightly different problem into a new context, which is useful in a finite (and vanishing) number of cases as AI fills in all of the "unsolved in this context but solved in that one" holes.

It is fundamentally incapable of logic, reasoning, or intelligence. But it does a good impression of it, by parroting patterns of data that have existed in the past, on the Internet and in literature. It can't actually "innovate", never mind "think".

The only business case for AI is in doing things that no human would agree to, e.g. on moral or legal grounds. Thus, the explosion of AI-driven scams etc.

Yet it is already having a psychological effect as people give up on the idea of studying or building a career.

It's useless and it's poisonous and it needs to burst before it does even more damage to society.

[Cris E]

Re: I got a hammer, find me nails.

@CyberDemon:

Two answers -

1. I could not agree more, this is ruining the development of skills and habits in our young people who may never understand that Work (eg writing an article, researching a topic, anything semi-substantial) often takes longer than 15 minutes. As it stands they can toss whatever into the hopper and in a matter of minutes out comes something answer-shaped, at low cost and of better quality than they could produce by lunch. They may never find out if it's right, or if they could have created something better on their own, because this seems to be good enough and no one appears to want more. Poison.

2. There are cases of highly repetitious work where an AI agent can be useful cranking out iterations in search of something. It doesn't need to create or even be average if it can automate harder tasks in a repeatable manner. Some of the mathematical "proofs" that have been credited to AI and many of the security vulnerabilities it's found are simply cranking through a multitude of low-percentage combinations looking for the needle in the haystack. It's not creative, maybe it shouldn't even be called AI, but it covers ground cheaper and faster than humans and sometimes that's of value.

[cyberdemon]

Re: I got a hammer, find me nails.

2. I agree, but as I say these "low-hanging fruit" problems are both finite and vanishing - requiring asymptotically more compute to solve as the number of "AI-soluble" problems approaches zero.

And the use of "AI Agents" etc, is currently heavily subsidised by the companies who, just like your local crack dealer, are desperate to get everyone hooked. They are prepared to turn trillions of dollars in capital into tens of billions in revenue, because they know that almost nobody would be willing to pay the real cost of their services, until their brain has atrophied to the point where it's the only option.

It also has a homogenising effect, meaning that all content processed by AI tends to look the same, albeit sometimes subtly manipulated to serve the political interests of its masters. There's only so much of a user's creative individuality that can be added with a text prompt or "unusual set of input tokens" - and ironically that "distinctiveness" is being assimilated by the slop-borg every time a human interacts with a cloud-based chatbot.

[EricM]

Re: I don't understand how the bubble didn't POP yet.

Too many people financially invested, hellbent on stocks going up, raising expectations and promoting stocks beyond reasonable means through all available channels.

Too many people in CxO roles, without practical AI knowledge beyond an occasional ChatGPT session, being easily manipulated by AI sales people in FOMO purchases.

Too many semi-technical people intellectually invested, thinking their "prompting skills" somehow make them special and fearing an environment where you have to possess actual technical skills yourself to make a buck.

At least the last list item distinguishes the AI bubble from "normal" financial bubbles and gives it some kind of inertia that might lead to more inflation before the pop (and more damage).

[BoHu]

Dancing around the AI seesaw

Oracle stock's got all the volatilty of plucked poultry on a roller coaster rotisserie ride lately, $140 in July '24, $190 that November, $130 the next April, $310 in September, $140 again in March '26, $230 this past May, and back to $140 now ... thrilling but not confidence inducing imho. Quite bursty, but with zero net growth in 24 months. Maybe they should focus on less-hyped fundamentals?

[EricM]

Re: Dancing around the AI seesaw

Given the size of their accumulated debt, that already forced them to lay off staff to pay their dues, going back to fundamentals now might not provide a path to survival for Oracle...

[Bebu sa Ware]

Re: Dancing around the AI seesaw

Dancing at the edge of the precipice.

I think Catastrophe Theory^† tells us that as a system edges towards the precipice it becomes increasingly chaotic.

Choose a set of parameters that broadly describes the system you are interested in and look for increasely wilder fluctuations in those parameters.

After a while you will be convinced that a catastrophe is inevitable and use your judgment to decide when to bail if you have any skin in the AI game although if you did that argues a certain lack of judgment in the first place

_{† I am but vague about this as René Thom's work was from the 1970s and enjoyed a certain popularity like Lotfi Zadeh's Fuzzy Logic earlier.}

[Anonymous Coward]

Fingers in ears

Cash hidden inside your mattress.

Hopes and dreams..

[Brett Weaver]

Bloody crap posting of a so called article

Couldn't read was full of speaker identification shit. Was not focussed did have any authority. Is this whats the register has become.. OK .. I'm too old and too intelligent for this.

[ICL1900-G3]

Re: Bloody crap posting of a so called article

Well... I agree with you. I bailed out after all the 'I was like...'. For me, teenspeak tends to undermine any argument.

But, on a larger scale, The Reg has become dull, dull, dull ... with the exception of most of Liam's articles.

[Mike VandeVelde]

Re: The Reg has become dull

Compared to what that currently still exists? I still like it. And the comments sections still retain some value.

[Mimsey Borogove]

Re: The Reg has become dull

the comments sections still retain some value

That's the main reason I come here. The articles are good, but to a non-tech person, the conversations of people who actually do this shit for a living are the most valuable. Thanks, all of you, for helping insulate me against "AI"!

[Stuart Castle]

I do, as a consumer, use various AI chatbots. Primarily Siri 2 (which uses Gemini) and ChatGPT.

However, i do not think the current AI models are sustainable. If the planet is going to survive, we need for AI to reduce it's hardware footprint massively. I truly think LLMs are a technological cul de sac because they do require massive amounts of compute power. Which requires massive data centres, which require massive amounts of power, and frequently water. Water which they frequently poison, thus reducing the supply for local towns.

Musk has proposed that the answer is putting the computers in space. It isn't. They won't be able to generate nearly enough power without miles of solar panels, and space, while extremely cold, isn't a good heat conductor.

Admittedly, I am by means an expert on any of this, but I do think we need a replacement for LLMs that uses little enough processing power that it can operate (perhaps with some restrictions) entirely on device, with no cloud accsss.

[Jedit]

"all the ways [Oracle] might suffer if this whole AI thing doesn't pan out."

As if we needed more reasons to hope that AI doesn't pan out.

[Ryan D]

Round and round and round we go

And how it ends we already know.

I do wonder how big a pop this will be? A small flash in the pan, or big bang? Time will tell.

[Dave@Home]

Pensions

Just have a look at any middle to high growth fund, and you'll see NVIDIA, Alphabet, Microsoft, Amazon

When it does go pop you'll see people's pension savings hammered

[David Hicklin]

Re: Pensions

It will be at that point that you will see how well managed your pension funds are.

Having said that 2025 was a record year for pension funds largely driven by the AI bubble which according to my advisor we won't see again for a long time, so it would seem that they are preparing for the big crunch.

[a_yank_lurker]

That was a helluva nap

Artificial Incompetence is an solution in search of a problem. What the AI bros aka con artists claim it can do is not what it actually does. Areas they claim it does well are very narrow and in most cases have automation tools available for many years.

[Anonymous Coward]

well obviously

A "Bubble" seems to have become a more user friendly term for a "Pump and Dump" scheme..at least in my opinion

1. Invent something slightly useful...maybe

2. Build up the hype... by promising the moon

3.a Use the stock-market.... you know when you and your mates need to go public and when to sell --probably?

3b As a bonus use your product to predict when to dump, after all you seem to have all the data

4. Profit for you and your mates

5. buy up an essential item that prevents anyone from easily and cheaply bettering your product

6. see point 4

7.buy some politicians

8 worm you way into government circles

9.get some nice regulations and laws

9 see point 4

Of course It wouldn't happen, people are just not that stupid.... oh wait...

I for one welcome our very very very rich overlords

[ecofeco]

Uh oh

A good con requires confidence of both parties. The conman and the mark.

The mark is getting nervous.

Uh oh.

[JLV]

One problem not mentioned here is the weight of the big firms on stock indices. Even if you’re an AI skeptic, as a fund manager not investing means the index is being lifted without you. That was true when Apple was at $2B. More so for nVidia at $5B.

Back in 2006 or so, IIRC, Wells Fargo’s CEO got into big trouble with stockholders. Fool had been sitting out the subprime lending and underperforming! Warren Buffett too during dot com boom times (but he’s Warren Buffett so he mostly got away with it).

It’s an interesting dilemma for capitalism. How do you avoid something that’s built into its incentives?

[Efer Brick]

Sloppocalypse

now

[xyz123]

The AI bubble is going to take down Microsoft heavily.

earlier they were $500 billion in debt for datacenter construction for AI.

It's estimated to be over a trillion now......

The boom! will be spectacular. Windows will end up being sold off as the company is torn to pieces.

No-one thinks it'll happen to them, then a well-known corp goes down faster than megan fox auditioning for a movie role.

[Mimsey Borogove]

The AI bubble is going to take down Microsoft heavily.

That could actually make the whole sad spectacle worth it.

[Anonymous Coward]

Not boom but revolution

Here are some implausibilities:

People judging AI based on the free services have missed what frontier models can do.

Programmers on this forum afraid of losing their jobs would rather destroy the machinery like the Luddites past.

University student use of AI is high; in a few years those students will enter the workforce and need AI to be productive.

The AI boom is the industrial revolution, except affecting knowledge workers rather than skilled laborers.

More plausible, on the other hand, is that the get-rich-quick schemes of investment bankers and corporations always cause trouble for everyone else.

[Fido]

Not boom but revolution

Here are some implausibilities:

People judging AI based on the free services have missed what frontier models can do.

Programmers on this forum afraid of losing their jobs would rather destroy the machinery like the Luddites past.

University student use of AI is high; in a few years those students will enter the workforce and need AI to be productive.

The AI boom is the industrial revolution, except affecting knowledge workers rather than skilled laborers.

More plausible, on the other hand, is that the get-rich-quick schemes of investment bankers and corporations always cause trouble for everyone else.