775
Banks are now trying to offload the AI data center debt they just underwrote
(startupfortune.com)
this post was submitted on 14 Oct 2026
775 points (99.0% liked)
Technology
88823 readers
3821 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Owning a debt from the AI industry is a fairly insane amount of risk
Is it? They've been outperforming every other debt class for five years running.
Even Enron was outperforming, just saying...
Spoken like a person familiar with the scandal
I would not be here on Lemmy but on a remote island to mind my own business
Their yields going higher and higher is not "outperforming", it's the opposite. Higher yield = higher risk.
Do you think they can maintain that for another 5 years? Another 20? Is the product AI companies providing worth the valuations?
Are 5 AI companies providing more value than every other tech company in the last 50 years? Thats what the valuations are trying to say. Im still waiting on ChatGPT to cure cancer.
I don't see it happening; Copilot was a failed proof of concept that crippled the Windows ecosystem and permanently tarnished Microslop's reputation as the premier operating-system developer; the Linux community doesn't openly embrace AI much at all, and gamers couldn't care less about it... All the LLM services have been running at a consistent loss/debt, and CoreWeave is propped up by the only 2 companies that couldn't even successfully launch a finished product. They need to cut their losses and pack it in; call it a day!
TBF windows has been going steadily downhill since XP, but yeah, the copilot shit certainly made a whole lot of people start looking for the exit. I think a bunch of people had been getting increasingly sick of the windows experience for years but kept using it because it was just there, and they couldn't be bothered to change. Actually, I was one of them. Copilot was just the final straw.
I think expecting NVIDIA to fail inside the next five years is a wild reach. You might expect Oracle to take haircut in the next five years, based on their atrocious debt load. But even then, the idea that they'll miss a credit payment?
Twenty years out, these firms are an even safer bet. Long term demand for compute and hosting services is only going up. And they're so heavily consolidated. Too Big To Fail in every sense of the word. Are you going to bet against JP Morgan and Goldman Sachs on the eve of the financial crisis, even? Twenty years later, how are they doing?
It's not a question of value, it's a question of earnings growth. Will these top 5 AI companies be seeing stratospheric earnings growth in the next ten years? All signs point to "Yes". The race to cement these data center footprints is a scramble for limited resources. The Capex budgets won't be inflated to this degree forever.
Dismissing them for their rapid expansion would be like betting against New York and Chicago real estate markets because you didn't trust skyscrapers to be profitable in another 20 years. Or betting against U.S. Steel and Standard Oil at the apex of WW2, because the resource demand from the conflict can't go on forever.
AI is just the NFT version of the GFR, again, in 2026 and much worse. No one will be hype farming it after this crash. All the larping "AI bro" channels will go. Influencers sitting on the beach with their laptops, shilling how it makes them $100,000 a month while they sleep. Those people will go... What I suspect will remain of the industry is a completely niche sector for specialised applications in schools, hospitals, higher education, and film & sound design; it can reduce some of the burden at work if implemented well. But most of the public will remain staunchly opposed to a subscription-based life where AI drives every IoT device in their home or vehicle and generates sloppy, hallucinated content that doesn't help them but harms them.
NOTE: In a survey, over 60% of Microsoft users say they actively seek ways to bypass online account processes for local accounts and encourage others to do the same. If you can't sell the idea of mandatory sign-in just for syncing desktops because everyone understands why a PC is exactly that "a Personal Machine", everything is supposed to stay & run locally if it is indeed "YOUR COMPUTER"... I don't see the big 5 generating the revenue through online subscriptions to get them out of that debt spiral; local open-source AI models and cheaper hardware are the only way I see plausible mass adoption happening.
I'm not saying they will fail I'm saying they will not maintain their rate of growth. Very, very different things but i guess you can pretend i said something else and argue against that or whatever idk dude lol
I think they'll shrink by half over a few years and revert back to the consumer market by repackaging all the hardware they couldn't get the power and data centres built for. They'll probably make a bunch of AI gadgets, then move on to the next thing.
They'll hit a saturation point eventually, sure. But demand for capacity continues to outstrip supply.
I think demand is coming only from a small group of people, though; it's definitely not the rest of the S&P 500, since they're going bankrupt from it! lol
Who is going bankrupt?
If you will hit a saturation point, at some point the demand will stop outstripping the supply. Once you cannot exapand, the supply will catch up simply because the demand decreases, not considering that the supply could simply increase.
Can you point to a five year span in the last century when demand for compute has decreased?
We are not at a saturation point for computer, but if you look at other fields, like smartphone, in 2026 we are set for a -7.4% so it is not impossible or never happened.
This can have a lot of causes, not last the increase in price, but in the case of the smartphones we somewhat hit a saturation point: there are very few people who can afford to have a phone who do not yet have it.
Same with computer and/or compute power: at some point, given that the users are a finite number, we will reach a point where no new people (users) will use it and at that point the demand will stabilize: you have only so much time to do things, both works and free time, so the demand flat out or decreases.
The demand decreases even going from building 10 new datacenter to building only 8 new datacenter.
Consumer electronics and business compute demands are wildly different markets.
That's not a measure of demand
So did .com in the early 00's and houses in the late 00's
If you bought into Microsoft in '01 and have owned a REIT since '07, you've made oodles of money.
Yes, if you invested in the right company you made a lot of money, if you invested in the wrong company, you lose all your money. The onvestment was high risk back then it is high risk now.
That's how investing works.
I would be curious to see how you try and calculate EV on a stock.
The risk back then was high. It is high now as well, if you invest in a sector that is deeply unprofitable and profitability predictions are based on a lot of speculation, ridden with circular financing, lots of obvious oversupply signs and contrary to most previous bubbles, a product (GPUs) that is depreciating so fast that no lasting value will be created by the bubble.
What about pets.com?
In the end, Chewy.com was the better investment.
I'd argue that Microsoft really didn't do anything .com-ish. Amazon would have been a better choice for your argument.
And noone's arguing that no companies flourished during those years. Just that a fuckton of them drug the market back to hell with their overpriced evaluations. (people laying fiber and buying 1.2M sock puppet ads in the super bowl)
They've been the pack leader in the tech sector for decades.
The MAG7 have been outperforming the rest of the market since COVID, often by 5+%
This is very different than betting on some small cap IPOs
So what did they do that was crazy long bad bet .com? Them existing in the .com days leaves nothing to the discussion at hand.
Amazon was at least standing up storefronts and cloud at that point.
What does this have to do with Tech bubble vs AI bubble? Are you interested in being on topic or just arguing?
Windows ME leaps to mind. Internet Explorer giving up enormous market share to Chrome and Netscape. Fumbling Search for twenty years. Take your pick
The Tech Bubble was largely small cap startups. AI is prevailing B2B and SaaS.
Also "bubble" only means something in hindsight. You keep saying AI is going to crash. Where's the crash? I've been reading this prediction for half a decade
bye!
Do you have any statistics?