The AI Bubble ... and Larry Ellison

[dann]

There is another AI thread, A thread to discuss the real-life dangers posed by AI, but not in Economics, Business and Finance and not specifically about the bubble, so this is about the bubble, the economy side of AI.
It looks pretty grim:

AI bubble about to COLLAPSE? Exposé on MAGA ally Larry Ellison's DEBT BOMB (MS NOW on YouTube, Aug 15, 2026 – 20:17 min.)
MS NOW’s Ari Melber delivers a special report on the tech boom, deregulation and the MAGA allies reshaping AI and media. Melber draws from a New York Times investigation, “Larry Ellison Bet It All on the A.I. Boom. Will He Be the Face of an A.I. Bubble?” New York Times reporter Jim Rutenberg joined Melber later on “The Beat.”

Articles mentioned in the video:
Wall Street is growing increasingly uneasy at the AI investment bubble (npr, July 30, 2026)
The AI Bailout Could Be Baked into the AI Bubble (The American Prospect, Aug 3, 2026)
Is AI a Bubble? It's Starting to Get Soapy (WSJ, May 19, 2026)
Is the Artificial Intelligence (AI) Bubble about to Burst? (yahoo!finance, July 12, 2026)
We're 'Absolutely' in an AI Bubble Right Now, Says Lead Edge Capital's Mitchell Green (CNBC on YouTube, Aug 12, 2026)
'Big Short' Michael Burry Says the Bulletproof AI Narrative Reminds Him of the Dot-Com and Housing Bubbles (Business Insider, Aug 14, 2026)
The AI Bubble Is No Ordinary Bubble (The Atlantic, July 21, 2026)
Wall St. Wants Another Half-Trillion Dollars for the A.I. Boom (NYT, Aug 11, 2026)
Top economic warns that the AI math doesn't make sense (Fortune/Yahoo!, Aug 10, 2026)
The Tech Expert Thinks It's 'Time to Call Bull**** on tha AI Industry' (PC Mag, Aug 11, 2026)
What Happens If the AI Bubble Pops? (Marketplace, July 29, 2026)
Larry Ellison Bet It All on the A.I. Boom. Will He Be the Face of the A.I. Bubble? (NYT, Aug 5, 2026)
Larry Ellison pledges $40-billion personal guarantee for Paramount's Warner Bros. bid (Los Angeles Times, Dec 22, 2026)

Quotations from NYT in the video:

...[Ellison's] big bet on A.i. was built on an astronomical amount of debt... because the more you spent building A.I. infrastructure, the more you would earn, or so the logic went.

...concerns have grown about whether the trillions of dollars being furiously pumped into this global ecosystem of data centers will ever return the promised profits.

One firm] has estimated that the A.I bubble is... four times as large as the 2008 housing bubble.

The financial structure of the data center build-out makes it especially vulnerable to a crash... The hyperscalers are investing heavily in the same companies they are counting on to buy their computing power.

Oracle's annual report... noted that it would not guarantee that it would be able to manage its outstanding debt, which had now grown to $130 billion. It warned that its customers might not be able to pay for its services...

Since founding Oracle in 1977, Ellison had become infamous for his ruthless need to win...

 

[Gulliver Foyle]

If you want proper analysis on the LLM bubble go to wheresyoured.at. Ed Zitron has been explaining what the media are only beginning to pick up for the last two years.

 

[dann]

Looks interesting. I hadn't heard about the site.

 

[dirtywick]

yeah zitron talks about the ai bubble a lot. patrick boyle covers this and related global market news quite often. i'd also recommend learning about passive investment and how index funds work as those organizations are now bending rules to list these ai companies. a guy named mike green explains that concept really well.

imo there's a lot of good arguments that there's some serious issues with the financial side of ai

there's also some concerning evidence that the productivity side of ai isn't living up to it's end of the bargain either. personally i think there's potential for ai for use in programming related fields, but even they aren't showing the productivity gains the market priced in and that's low hanging fruit. trying to turn that into a broadly useful tool to justify that kind of massive investment is a little more questionable.

Seven Myths about AI and Productivity: What the Evidence Really Says

Meta-analytic evidence finds no robust relationship between AI adoption and aggregate productivity gain.

cmr.berkeley.edu

However, a July 2025 systematic review of 37 studies examining large-language-model assistants for software development reveals a far more granular reality (Mohamed et al., 2025). While developers did spend less time on boilerplate code generation and API searches, code-quality regressions and subsequent rework frequently offset the headline gains, particularly as tasks grew more complex. Senior engineers, in particular, found themselves investing substantial time fact-checking AI output for subtle logic errors that junior developers might have missed entirely.

Rebecca Hinds, head of the Work AI Institute, says there are three problems creating this productivity paradox. Number one is that the impressive time gains are not what they seem. (Or: look behind the headlines, as journalists say .) Rebecca says: “We find a lot of the [time] gains are being absorbed by what we call ‘botsitting’, the hidden work of making AI usable, feeding it context, checking outputs, rerunning prompts and cleaning up mistakes. Workers report spending 6.4 hours a week on botsitting.”
The second issue hampering productivity is what Rebecca calls “the toggle tax”. I hadn’t articulated this before, but it’s familiar to many of us: nearly eight in 10 workers have to juggle multiple AI tools every week, which is time-consuming; 60 per cent are running the same queries across different tools in search of better outputs.
Finally, many staff are spending time on the AI version of “workplace theatre” — visibly performing work for bosses and colleagues, rather than focusing on the actual grind of getting things done. They may even deliberately downplay the help they get from AI (a third of workers do this). All of this performative work takes a toll, both emotionally and in terms of time wasted. “Workers are managing the optics of AI as much as the work itself,” Rebecca points out.

 

[jeremyp]

imo there's a lot of good arguments that there's some serious issues with the financial side of ai

The most serious issue IMO is that almost nobody is paying the true cost of the LLM tools. For example, at the top of every Google search I do is an AI summary. I don't pay anything for that even though it has cost Google (or rather Google's investors) quite a lot of money to put that summary there. I do find it quite useful sometimes, but if Google said "you need to pay us a penny for each search with an AI summary in", I'd turn it off, or move away from Google.

Companies that use Anthropic tools to help with their coding report significant cost reductions because they can fire developers. However, not only are there hidden costs associated with that in that nobody in the company understands the code and there's no way to Google the StackOverflow question it was pasted from, but the LLM model is subsidised by Anthropic's VC investors. As soon as you start paying a realistic price that gives Anthropic a positive margin, it's going to look much less attractive.

The reason AI companies are pushing the technology so hard is that they have got to get us all dependent before the VC money runs out.

 

[The Great Zaganza]

The underlying issue is that AI doesn't scale the way Amazon or Facebook, Uber other systems that profit exponentially from extra customers.do.
The extra costs per added customer is the same if you add one or one million - it doesn't go down.
And there is no reason to use one system because everyone in your network does - in fact, it makes a lot of sense to use different systems to detect mistakes.

You can't even make propitiatory inventions with it, as everyone can figure out how your prompt worked once you submit the Patent.

 

[dirtywick]

speaking of proprietary stuff, giving an ai company access to proprietary data so you can train it's ai is also pretty risky

 

[bigred]

"The sky is falling, the sky is falling"

I'm leery myself, but not hopping off the bandwagon just yet. When/if the AI bubble bursts is anyone's guess.

 

[Darat]

There has been quite a bit of discussion about the economics of AI in the AI thread in the "Science..." section but probably makes sense to have a separate thread.

The thread in the "Science..." section: https://internationalskeptics.com/forums/threads/artificial-intelligence.369280/page-100

If you want proper analysis on the LLM bubble go to wheresyoured.at. Ed Zitron has been explaining what the media are only beginning to pick up for the last two years.

He's been on a similar trajectory to me, once I found him I've been recommending him because obviously if he has come to the same conclusions as I have he must be a brilliant bloke and be totally correct...

It was the circular financing that started me to ask "where are the new mega-datacentres?" and I wasn't finding them, I wasn't even finding new ones that had broken ground. This led me to the thought of "are there hundreds of thousands or even millions of these GPU cards sat in warehouses?" and "where are we seeing the depreciations in the companies' books?"

It is a house of cards, but an inverted one!

 

[Darat]

Long interview/discussion with Ed Zitron. It's a bit of a ramble but covers all the negatives.

If you want to hear about depreciation etc. you can jump to 30 minutes in.

 

[ahhell]

There has been quite a bit of discussion about the economics of AI in the AI thread in the "Science..." section but probably makes sense to have a separate thread.

The thread in the "Science..." section: https://internationalskeptics.com/forums/threads/artificial-intelligence.369280/page-100



He's been on a similar trajectory to me, once I found him I've been recommending him because obviously if he has come to the same conclusions as I have he must be a brilliant bloke and be totally correct...

It was the circular financing that started me to ask "where are the new mega-datacentres?" and I wasn't finding them, I wasn't even finding new ones that had broken ground. This led me to the thought of "are there hundreds of thousands or even millions of these GPU cards sat in warehouses?" and "where are we seeing the depreciations in the companies' books?"
It is a house of cards, but an inverted one!

They are being built, I know of two that my organization is building. There have also been several that never got out the planning stage. The folks that own are pretty hush hush about it. I've add to sign some NDAs so, I can't say anything else about it.

ETA: For whatever my opinion is worth, this is a bubble, it will burst, and there will a number of companies that fail spectacularly as a result. I think there will also be a smaller number of companies that succeed just as spectacularly.

 

[Roger Ramjets]

The most serious issue IMO is that almost nobody is paying the true cost of the LLM tools. For example, at the top of every Google search I do is an AI summary. I don't pay anything for that even though it has cost Google (or rather Google's investors) quite a lot of money to put that summary there. I do find it quite useful sometimes, but if Google said "you need to pay us a penny for each search with an AI summary in", I'd turn it off, or move away from Google.

That's not a problem. Google already supplies a free search engine, free video channels, free maps etc. - all paid for by advertising. Google also uses their AI in house to save money on programming, so they would have it anyway.

The problem is other AI companies which need paying customers. The vast majority of users are only attracted to AI because it's cheap or free. The AI companies may think they can get people so hooked on it that they will pay more, but this won't happen. So they need to bring the cost down. Building more data centers won't help, it just makes the goal harder because then they have to charge even more to make it profitable.

Brute forcing it isn't working. The AI companies that survive will be those that bring costs down by improving the models. There's no guarantee that the current players get there first. They are probably less likely to make the needed breakthroughs because they are committed to maintaining what they have and won't risk doing anything radical.

This is nothing new. It happened to IBM with the PC, and to Kodak with digital cameras. In both cases they got in early with products that were functional but overpriced, and by the time they made them cheap enough other companies which didn't have the deadweight had outcompeted them with better products.

The reason AI companies are pushing the technology so hard is that they have got to get us all dependent before the VC money runs out.

I think the original idea was that with enough investment they could produce a product that was worth paying for. This turned out to be a lot harder than expected, but they crossed their fingers and hoped that eventually a breakthrough would occur. Now the sunk cost is so high that they have to keep going.

This is what happens when people have more money than sense - tech bros and investors who don't understand the technology and just assume that anything is possible if you throw enough money at it. Not the first time, won't be the last.

 

[dirtywick]

i agree that companies like google and microsoft, and even meta, are in a much better place being as how they have some actual revenue streams

it's a pretty tried and true practice of tech companies to promise something they can't do, collect a bunch of investment, and figure it out on the way. it's just not working that well with ai and it's gotten out of hand

a big part of the problem is with the market though. too much algorithmic trading, too much passive investing, too much dumb money

 

[The Great Zaganza]

The problem is that there is too much money in the market, caused by massive government support of the shaky economy and financial system, without the claw back with higher taxes and interest rates when the economy is doing better. The other difference is the lack of transaction costs, making it possible to move trillions around daily instead of having to commit to an investment.
A %-based transaction tax, no matter how small, would force investors to be more smart and less hectic.

 

[dirtywick]

European Central Bank Warns That AI Crash Is Looming

A new analysis by economists and financial researchers at the European Central Bank warns of an impending AI bubble crash.

futurism.com

europe central bank sees crash as likely in two scenarios

Which of the two scenarios is playing out right now is difficult to say — and it’s likely an unholy blend of both. In any case, the authors write that both views “imply a boom followed by a correction, or a pullback from wherever valuations have risen, at some point in the future.”

 

[dann]

Long interview/discussion with Ed Zitron. It's a bit of a ramble but covers all the negatives.
If you want to hear about depreciation etc. you can jump to 30 minutes in.

WH nightmare: AI bubble COLLAPSES! A TECH INSIDER shows the TERRIFYING WAY it may all CRATER (MS NOW on YouTube, Aug 21, 2026 – 12 min.)
MAGA allies are making huge bets on AI amid new signs that a tech “bubble” could rattle the economy. MS NOW’s Ari Melber reports and is joined by Facebook co-founder Chris Hughes and journalist Ed Zitron.

 

[Beelzebuddy]

WH nightmare: AI bubble COLLAPSES! A TECH INSIDER shows the TERRIFYING WAY it may all CRATER

That kind of breathless histrionics should be a red flag, even if you end up agreeing with the conclusion.

 

[TragicMonkey]

That kind of breathless histrionics should be a red flag, even if you end up agreeing with the conclusion.

I've been seeing videos and articles all about the imminent collapse of the AI bubble for at least nine months now. I'm sure it'll happen eventually, but then I'm also sure I'll eventually die. Even certainties about future events can be useless if you don't know how far in the future they are.

 

[Beelzebuddy]

I've been seeing videos and articles all about the imminent collapse of the AI bubble for at least nine months now. I'm sure it'll happen eventually, but then I'm also sure I'll eventually die. Even certainties about future events can be useless if you don't know how far in the future they are.

The market has always been irrational, but lately I feel like it's suspiciously so. This behavior, openly passing a buck between them all and everyone claiming it as earnings, isn't something that would have happened in the recent past without knives coming out. Just like Musk shouldn't have been able to take an already grossly overvalued stock, walk out on stage, yell "datacenters on the surface of the sun!" and have people actually buy it, without something being deeply broken.

I wonder how much irrational exuberance isn't actually hoping for a greater fool, but just trying to keep the status quo going long enough for someone more responsible to be behind the wheel. That as soon as the next president (D) even hints at replacing the SEC with people willing to do their damn jobs instead of kleptocrat vultures, the whole apparatus is going to lurch over the threshold and collapse in a grateful puddle of recession.

 

[dirtywick]

too much dumb money trading off algorithms analyzing tweets and index funds. these guys have figured out how to exploit that stuff, rather than building strong businesses imo