AI's debt binge can't last, hidden borrowing reaches $1.65T
fortune.com
fortune.com
We have only begun to extract the value of commoditized intelligence. Sure there are arguments on local models and pricing power.. but I think we will be compute constrained for the near future.
Journalists have been eager to call AI "over" since 2022, and yet:
- Models just got good at writing code this year
- Models just got good at editing images last year
- Models just got good at cinematic video this year
This hasn't even played out. It hasn't even started.
Why on earth would this be the end?
The robotics story is just getting started, too.
I literally do not write code anymore.
Rather than it being presented as productivity tool for enhancing human labor and activity, it's presented as an eventual god that will radically transform the rules of economics and everything else, and thus it needs to be forced into everything. That's absurd hype and with it all the absurd VC funding and valuation. Thus it's seen as a huge financial bubble.
This could happen this year or next, assuming you're willing to pay $30k for the hardware.
- Fable
- Seedance
- Nano Banana / GPT Image
- Kimi
- ChatGPT
These tools are 80-90% of my day now.
Google Search? Meh. Chrome? Eh. Mac or Linux? Honestly just input devices now.
The models are the hottest thing in the world.
I am getting so much done. If I told myself from two years ago the progress these models would have made, I wouldn't have believed it.
Absolutely lots of hype but there is lots of value behind generated (unlike crypto) and we are still very early. This is what I was pointing at. There are folks on very extreme both sides, you are a good example, and I happen to believe it’s probably somewhere in the middle.
If any one of these happens, or two, or all three, then the loans for trillions will become worthless while the use of AI can explode. The relationship between cost and ai intelligence output need not be linear over time, which is absolutely what the people financing are assuming.
Personally, I think linear over 5 years is about right, but no longer than that.
Google was not the first search engine. But in a way, it was the last.
I think other industries are use to being continually disrupted by advances in technology and so will adapt easier and faster. Which again, is kind of ironic..
(i am a dev myself but it still makes me laugh)
That's not correct, is it? Opus 4.5 came out in Nov 2025. Some might say models were good at coding even before that.
Here is probably one of the more clear examples. A model trained on video and also robotic simulation/recording (probably ensembled with control systems/mobility models) will likely be at the core of how robots make decisions and plan.
https://deepmind.google/blog/gemini-robotics-2-brings-whole-...
This is way outside my area of expertise though. I've only dabbled in more classic robotics and control systems, but these multi-modal sequence to sequence models are highly adaptive and can effectively transfer learning across very different domains.
It took 10-15 years before the stock market got back to the level it did pre dot-com crash [0].
So AI can become as big as the internet, but that does not mean that the existing "AI" stocks will become big.
but you are taking advantage of a subsidised service. What happens when developers are forced to pay the full price (i.e. $2000 a month instead of $200 a month)
Similarly, majority of people still don't 3D-print stuff they can get cheaply at Walmart or from Amazon. Or use VR/AR as their primary form of interaction.
I too would use "plenty" rather than look at the horribly depressing stats.
Basically, 80% of sales are still brick-and-mortar. That doesn't seem very depressing?
One thing's for certain: There's no way anyone who's come close to Sauron's Ring (made actual use of AI) wants to part with it :')
Although at its peak, CSCO was up ~2500% in a 5-year period, whereas NVDA is “only” up ~1000% in a similar timeframe.
Ever since the 2008 housing crisis, people have been predicting the next bubble-burst/black-swan event.
The one that really crushed the markets was the one almost body saw coming: Covid-19.
Tulips, 1929, Dotcom, Great Recession, 2010's Flash Crash - none of these were in the public discussion before they happened.
The "public discussion" is a whole different thing. They weren't in the public discussion because macroeconomic theory isn't something mom and pop like to chat about on the weekend. They only become dinner-table discussion topics when the impacts hit main street, after they happen. But bubbles in recent history have been pretty reliably identified beforehand:
https://web.archive.org/web/20180330001927/https://www.barro...
https://www.economist.com/special-report/2005/06/16/in-come-...
It isn't hard for economists to find bubbles, where the market is taking on high levels of risk. What is downright near impossible to do is predict what specific event will cause the dominos to begin dropping, or when it will happen.
Anecdotally, I have family who don't follow the stock market at all and are talking about the "AI Bubble" that's about to pop.
That's why timing the crash is hard. The market has to agree with you but also at the right time
The only people who didn’t see it coming were the capitalists who were invested in the inflated market, and had bought into pseudo-scientific economic theories that served the single purpose of affirming what the capitalists already believed.
There's a very good reason a book (and movie) like The Big Short was such a big hit. It's because it was about the handful of people who actually saw the crash coming and were confident enough to put their money and reputation on the line.
Feel free to cite at least one reputable source.
Sorry to break it to you but you are neither immune nor a bystander to the fortunes of AI going down. You are part of it all whether you like it or not.
there are a lot of older folks leaning on 401ks and IRAs that are heavy in big indexes and I suspect many of them may end up eating cat food to get by if/when the bottom drops out.
the rest of us will, though. I for one am looking forward to cheaper graphics cards, which I'll use to play video games after givin out handies behind the wendys
I'd love to live in a world where AI firms bidding these things up doesn't affect me but I'm really struggling to understand how they aren't impacting the market.
Just because I'm too poor to own stocks doesn't mean stock prices don't affect me. That's indirect exposure not direct exposure.
You're never too poor to own stock.
The financial crash that will happen as the result of the AI speculation bubble popping will be the exact same. You won't see Altman or Dario on the streets, that's for sure.
so zero exposure to any popular index? Even “ex-US” is tsmc and sk-hynix in a trenchcoat. I think it was BHP exclaiming that theyre an AI play because they cover 85% of the raw materials in DC build outs.
In the current mania “no financial stake in AI-related” is a very bold claim.
Of course, this doesn't insulate you from second- or third-order effects, but it does remove the possibility of your money being immediately wiped out. It's really not as bold or crazy as you say it is.
If you are young, impact is less scary because you still have 20, 30 or 40 years for the market to go back up.
Are you aware of the existence of other countries which do not operate like the US? (no mandatory pension fund, etc.)
what will come out of the crash will likely be terrible for the average person, regardless.
Yeah, this is what I'm thinking. New ways of productizing the technology are still be defined as people are using it. The pricing models are evolving in real-time as the providers figure out what the market will bear.
Fiber and railroads don't need tens of billions of dollars in continuing yearly maintenance expenses to keep them from going stale.
Same here - the main value is in dominating AI or something like that, not in the stack of hardware.
Congrats, you're as qualified as all the private equity companies that have been piling into railroads!
That's precisely the sort of attitude that got us East Palestine (and others.)
BNSF (for example) spends billions of dollars a year on maintenance. When trains go over rails enough it can mess with the ballast and they have to come through with machines that lift the rail and jiggle the ballast and place the rail back at the correct height. And non-concrete ties have to be replaced every so often. Signals need maintenance. Switches need maintenance. Sensor/scanner shacks need maintenance. Etc.
Railroads have been dramatically cutting back because of pressure from PE, so the estimate is probably low.
>Fiber and railroads don't depreciate after 3 years of use like AI chips.
Are you seriously arguing that infrastructure doesn't depreciate? Tell me you've never done anything other than push bits, without telling me. Fiber gets eaten by backhoes trying to show off to potential mates in the spring, washed out, run over, knocked down by drunk drivers (if on telephone poles, not all of it is buried), has to be relocated because of other works, suffers water intrusion, amplifiers fail, you name it.
More slowly and with less expense in relative terms than an LLM model.
You get a lot more bang from your buck from a 10 year old set of under-maintained railroad tracks than you do from a 10 year old unmaintained LLM.
To compare the depreciation or ongoing maintenance expense of the two as apples to apples is ludicrous. To equate the capital expenditure and long term value of the two as equal is also ridiculous.
By comparison, OpenAI got another $110 Billion alone for their next round of funding. U.S. private AI investment climbed to $285.9 billion in 2025.
the difference here is that we know how the railroads work, and how to make a profit off them -- PE squeezes notwithstanding. the sports car I drive probably came here via train. meanwhile AI uses more electricity than Bolivia to make poorly formatted Powerpoint Presentations for useless meetings.
For those without accounts, given faded body
chuckle
Now if the number was ?? and labeled "undisclosed"... that would present a more serious problem.
The Wall St vs Silicon Valley showdown that’s setting up here looks like it will be quite epic. If last week was any preview, get your popcorn ready.
Pure play companies, startups, and investors are looking a lot less safe. For example there are other pure plays where debt service alone is like 25-30% of revenue, which is just insane numbers. There are also many investors and funds with extremely precarious positions in AI that are at risk of unraveling with a bang like we saw last week.
It's not like their collateral is a bunch of NFTs.
And is it collateral if it has yet to be built?
(I'm asking because I want to correct my own ignorance.)
Still no credible long term solution to the so-called "UBI" for all and the abundance fantasies and the utopia that was supposedly "promised".
Let's assume the extreme worst case scenario where the bubble pops so comprehensively that the entire AI business is written off, without any change to the debt owed, and these companies return to whatever they were doing before i.e. their previous levels of free cash flow. Naively, they could still repay the $1.65T, with interest, in ~6 - 8 years.
They will, of course, not do that, and will instead try to protect their plummeting stocks and get into a series of lawsuits as they try to claw out of their commitments (hey, maybe the circular investments even cancel out... it's a feature, not a bug!) and a lot of smaller companies go under, and some may angle for bailouts. But even then, the damage to the broader economy seems limited, and this debt doesn't seem that extreme?
If the hyperscalers needed to wipe out most of their income on interest expense they’d lose a large amount of their market capitalization. This could drop the stock market a huge amount, and a lot of spending is driven by the “wealth effect” of households feeling wealthy.
But it seems to me that if a stock market wipeout triggers a recession, it's because of deeper, pre-existing problems with the broader economy (inflation, jobs, war) and the stock market (concentration, unrealistic valuations) that are unrelated to the AI spending.
The GFC “proper” was the dramatic crash in the liquidity of credit markets, not strictly a corollary of the losses on property and mortgage-backed securities.
Won't the debt-holders have some claim to that future cash flow to be made whole?
If the debt is serviced, nothing.
Any historical precedent for this all occurring together with technological hype/fast growth?
FWIW Enron was also a „sophisticated company“ at the time
For the AI bubble too many people assume that large companies having a stake in it will of course know what they are doing, be careful and not expose themselves too much or do wild bets that don’t pay off. But looking at the level of capex from hyperscalers, the amount of circular financing by NVIDIA/google/microsoft, the level of debt raised for datacenters (and its associated raising interest rates), the lack of moat for AI labs, the absurd AI labs valuations, OpenAI ever increasing infra expenditure commitments (we are at more than $750B for 2030), Oracle dire situation (to say the least), the mounting pressure from China/open models, and the fact that 2 companies represent the vast, vast majority of the compute demand. None of that looks like a healthy, sustainable industry. In fact it looks like the most obvious financial engineering ever, where the only ones benefitting are NVIDIA, memory manufacturers, and hyperscalers. And they are doing what is necessary to keep the game going. If the demand for AI vendors isn’t increasing massively in the coming years the whole thing will go down. And the level of demand required need to be pretty much the AI booster dreams where everything becomes agentic everywhere. Short of that we are very likely to see things go downhill
There's an old WSB saying: the market can remain irrational longer than you can remain solvent. The AI craze is that but on 'roids.
> These are incredibly sophisticated companies so presumably they wouldn’t let themselves get into a company ending bind.
The problem is, company C-levels don't care about the long term health of the company. They only think about next quarter (in a misguided interpretation of "shareholder duty/fiduciary duty") and their bonuses tied to their KPIs.
> But what are the chances this is actually an MBS type situation where the system is truly overloaded and a few sacrificial lambs are needed?
The system definitely is overloaded to hell and beyond after well over a decade of ZIRP. That money never got deflated out of the system in a healthy way and now everything is looking to fall apart.
Unfortunately, such events are already "priced in". VC essentially is built on 1 of 100 investments striking it big and 99 going bust. A market correction won't hurt the big guys, but it will definitely hurt all the small guys.
We need to stop thinking that just because they have money they're incredibly sophisticated. We have a few examples like Mark Zuckerberg, who had early success with FB, but he seems to be incapable of investing in profitable products. E. Musk: great at selling his companies, but laughably bad at making profits at the same level of expenses. Sam Altman: never had a real job he did well other than raising money. This is the kind of people that control these companies.
NVDA had the foresight two decades ago to invest in CUDA. That's not next quarter thinking.
eh trolling for clicks. It's just not on the balance sheet (if i have my terms correct) so you have to look in a different report to find the numbers. If it was truly hidden then discovery of the debt would trigger lawsuits from investors. Major investors know about it already that's why no one is getting upset over it except for laymen. btw, laymen in the stock market (retail investors) just serve as red meat or cannon fodder for actual traders with real money and real information.
edit: there will def. be significant winners and losers, the stakes are very high and the dollar amounts are very large.
A lot of publications pay attention to that.
A lot of people love reading things (often only reading things) that make then feel right/correct/justified.
A lot of publications live or die on ad views.
And just like that we have a viable media business model!
Headline doesn't really match the facts in the article. The article seems to say "hyperscalers are borrowing an enormous amount and so far people are lending to them. Other people are worried that this will stop".
My future's so bright I gotta' wear million dollar shades.
You say this as if when "lending stops", it isn't a big deal. What you're describing is a concern for a collapse in finance markets.
The article is about hyperscalers which are massively profitable irrespectively of AI. If the lending stops and this leads to paused or cancelled infra projects, but these companies are still able to service their existing debts, how does this inevitably lead to a collapse?
Hoping for a more nuanced analysis than “la la la we all know this time isn’t different”. Every time is different, and it’s instructive to know how different each time is.
That setup isnt true for the US, not even close.
All of this is kind of beside the point though, because the issue was leverage not exposure. In south korea ppl were forced to sell at the low point of the market due to margin calls. For a retirement account, you can just choose to take a disbursement next month or next year (assuming youve managed personal cash flow with sequence of returns risk in mind).
Systemically i guess thats more a question of 1) is US equities just an AI trade, and 2) how much of all these investment vehicles are just such assets. Regardless of the answers, the nice thing is that unlike debt, there isnt a positive feedback loop here (i.e. margin calls increase volatility)
>[deepmind exec] said today’s enormous AI capital expenditures are not yet supported by current revenue, but argued that betting against the long-term trajectory of the technology would be a mistake. ... he noted that “the revenues from AI don’t sustain the capital expenditures we’re making so far,” while emphasizing that the early foundations of [recursive self improvement] are already emerging.
>Sekhon compared the evolution of AI to earlier industrial breakthroughs, saying, “Steam engines were used to create the next steam engine,” suggesting that today’s AI systems will increasingly be used to develop more capable successors. https://www.citybiz.co/article/883339/google-deepminds-jasje...
Unless someone stops finally US. And even then I am not sure that it will bring good, because before drowning, they will try to take all the others with them. After all, "it's theirs"..
they're still not your friends, and you need to buy the hardware from other assholes, who are also deep in this game, e.g. NVDA
I find this to be a pretty good split for all my side projects. A real developer wouldn't find this sufficient I would imagine but with the recent discounts on ChatGPT 5.6 Luma and Deepseek Flash V4 0731, I tend to have plenty of left off.
I might be an odd case, tech stuff is my hobby so I enjoy playing with these tools. I do not have any media subscriptions (no Netflix, Amazon, etc) but 20-50$ a month isn't bank breaking.