Last month, I did something I almost never do: I tried to put a clock on a market crash. I wrote that I thought the AI bubble, and potentially the broader market bubble that has attached itself to AI like a remora on shark, was likely to burst somewhere between the end of this year and the beginning of 2027.
Normally, trying to time these things is a fool’s errand. You can be completely right about a bubble, completely right about the eventual outcome, and still get your face ripped off for another year while everybody collectively agrees that paying any price for anything with the letters “AI” attached to it constitutes rigorous securities analysis. But the setup now looks so insane, I decided to stick my neck out anyway.
Then, last week, I wrote about the recent Anthropic whistleblower situation and argued that the biggest risk to the AI trade might no longer be competition, valuations or even disappointing revenues. It might be regulation itself. Researchers inside the companies building these systems are publicly warning that the technology was advancing faster than they believed it could safely be controlled.
I wrote days prior that we may only get one chance to stop AI. And suddenly, last week, the idea that governments might intentionally impede AI development, something that would have sounded completely insane to investors a year ago, started looking considerably less insane.
I also wrote last week that I found it fairly sick that 24 year old “hedge fund manager” Leopold Aschenbrenner was being handed another opportunity to speculate on the same AI theme using options after his fund managed to incinerate more than $30 billion during July’s AI stock rout. As a reminder, his fund Situational Awareness had grown to roughly $45 billion before collapsing toward $10 billion, forcing the fund to unload most of its public equity portfolio.
Which brings us to this weekend, because something much more fundamental may actually be changing. And as a result, the market could wind up flying from ludicrous speed to a deadass stop, shooting investors, most of whom are not buckled up, to the front of the ship helmet first.
On Saturday, Anthropic CEO Dario Amodei publicly called for the AI industry to slow the pace at which frontier model capabilities improve. Sam Altman then publicly agreed that the industry needs to “pace the frontier” and said the issue has become a major topic of discussion inside OpenAI. Even more interestingly, Altman suggested that leading AI companies may be getting close to announcing some kind of coordinated arrangement around slowing development and addressing safety risks.
In the same breath, Altman confirmed that OpenAI will not go public in 2026, specifically pointing to the current safety environment and the work the company still needs to do on safety, alignment and cooperation with governments.
Now, think about how bizarre that sentence would have sounded six months ago. The company sitting at the nerve center of the biggest speculative narrative on planet Earth is effectively saying that now may not be a particularly good time to sell itself to public investors because everybody is suddenly having a serious conversation about whether the technology underneath the trillion dollar investment boom needs to be slowed down.
And remember: the massive AI capex buildout is one of the primary engines of this entire market boom, meaning any serious slowdown in AI spending wouldn’t just hit AI companies, it would hit all the companies associated with them, from power to real estate, to data centers, to cooling and the likes…and it could pull a major support beam out from underneath the broader market.
And I have to say, the timing of the IPO postponement smells a little funny to the ole’ Q-Man. Maybe safety really is the entire reason OpenAI suddenly doesn’t want to go public this year, but I have trouble believing that altruism is really calling the shots at any of these companies.
Another possibility has been nagging at me this weekend: maybe OpenAI already needed a respectable reason to postpone an IPO that wasn’t generating the kind of enthusiasm it wanted, and then fell ass backwards into the greatest excuse imaginable when the Anthropic whistleblower story exploded.
There had already been reports that OpenAI’s advisers were concerned a 2026 offering might not be greeted enthusiastically enough by public investors, particularly at the kind of astronomical valuation Altman reportedly wanted.
So if you’re sitting there trying to figure out how to explain that your trillion dollar IPO suddenly needs another year in the oven, “we’re deeply concerned about the future of humanity” certainly plays a hell of a lot better than “public market investors aren’t as excited about buying our enormous losses at a trillion dollar valuation as we hoped.”
Again, that’s my speculation, not something I can prove, and the safety concerns may be completely sincere. But that’s also why the juxtaposition is so funny. OpenAI goes from preparing for one of the biggest IPOs in history to telling everybody this isn’t the appropriate moment to go public because the technology may be too dangerous, just as the market has started asking much more uncomfortable questions about AI valuations, financing and monetization. Maybe those two things have absolutely nothing to do with each other and the timing is entirely coincidental.
And hey, maybe that stripper really does like you.
I have absolutely no idea how seriously any of these guys mean the slowdown talk either. These companies are locked in one of the most vicious technological arms races in modern history. Each of them knows that voluntarily slowing down while a competitor keeps accelerating could cost them hundreds of billions of dollars and potentially their position in the industry.
There is an obvious game theory problem here. Everybody can agree that driving toward a cliff is dangerous. Nobody wants to be the first guy to tap the brakes while the other sociopaths keep the accelerator pinned to the floor.
But from the perspective of the stock market, I’m not sure their sincerity matters nearly as much as people think. The narrative itself has now changed. The market doesn’t need OpenAI and Anthropic to unplug their servers Monday morning for this news to matter. Investors simply need to begin asking a question they haven’t seriously had to ask before: what happens if the AI buildout slows down?
That is a potentially crushing narrative for this market because an extraordinary amount of today’s valuation rests on the assumption that AI development will continue moving exponentially forward with virtually no interruption. More chips, more data centers, more electricity, more servers, more models, more capex, more token usage, more revenue, more productivity, more fucking everything, forever.
That assumption has worked its way through Nvidia, Broadcom, AMD, Microsoft, Meta, Alphabet, Amazon, CoreWeave, Oracle, utilities, data center operators, networking companies, power equipment manufacturers, nuclear stocks, private credit, construction financing and virtually every other corner of the market even tangentially connected to AI. In fact, in a quick analysis I ran this morning, it appears to me that far more than half the top 100 weighted S&P 500 names have some credible ties to the AI buildout.
The bull case embedded in these valuations is that the spending curve keeps going up and to the right at a completely absurd rate for years. Instead, we’re talking about slowing it *now. *
Lest we forget, the AI boom is no longer being funded merely by gigantic technology companies casually reaching into enormous piles of cash. The financing ecosystem underneath it has become enormous. Goldman has estimated that hyperscaler capital spending could reach roughly $1.1 trillion in 2027, with an upside scenario approaching $1.4 trillion if AI investment reaches levels comparable to previous historic infrastructure buildouts. The problem is that hyperscaler cash generation increasingly isn’t enough to comfortably finance spending on that scale, which means the system has been turning toward debt, leases, private credit and increasingly elaborate financing vehicles.
And the publicly visible debt is only part of the story. Morgan Stanley has estimated roughly $1 trillion of long term purchase commitments across the AI ecosystem along with more than $800 billion of lease commitments that haven’t even commenced yet. Call it roughly $2 trillion of off balance sheet economic commitments associated with the buildout. And sure, companies can enter enormous future purchase commitments without immediately recording conventional debt, and signed leases that haven’t commenced can similarly remain outside the headline balance sheet liability figures.
But somebody has still promised to spend the money. And increasingly those promises are being used elsewhere in the financial system as the foundation for more borrowing. A hyperscaler signs a giant long term agreement with a data center operator, chip supplier or infrastructure provider. That supplier can then use the agreement to help obtain financing from banks or private credit firms. The leverage migrates into SPVs, suppliers, leases, private credit vehicles and other corners of the financial system where investors have a much harder time seeing the entire picture.
Wall Street has once again discovered that if you put debt somewhere complicated enough, everybody can temporarily pretend it isn’t debt. We have absolutely never seen that movie before.
There is also an ugly timing problem buried underneath the capex boom. Companies are spending enormous amounts of money today on assets whose full earnings impact may not appear for years. Much of the money currently sitting in construction in progress isn’t yet flowing through depreciation expense, but it will. (For more on this, and the accounting behind this circus in general, watch this great interview).
Eventually those data centers, servers and other assets get placed into service and depreciation starts showing up. Morgan Stanley has estimated cumulative depreciation across Microsoft, Oracle, Meta and Alphabet could exceed $520 billion over three years. That’s manageable if AI revenues explode fast enough to absorb it. It becomes significantly less manageable if the industry’s own CEOs decide the technology needs to slow down right around the time the depreciation bill arrives.
So there are effectively two deferred risks sitting underneath the boom. Some of the leverage is hidden in space, pushed into suppliers, SPVs, leases and private credit, and some of the cost is hidden in time, sitting in construction in progress today before eventually flowing through depreciation and pressuring margins tomorrow.
They both start to unlock the second it becomes clear that AI is not developing at the speed that every junior analyst wandering around drunk on a Friday night on Stone St. was forced to integrate into their DCF models in order to back into whatever bullshit price target his firm has for any given tech company.
This is why I think this weekend’s news could wind up mattering much more than it initially appears. Because slowing frontier AI development has moved from something discussed by safety researchers and terrified former employees into something being publicly entertained by the CEOs of the companies actually building the models. Now, politicians are listening too. And we all know…once an idea enters Washington, God only knows what monstrosity eventually comes back out.
Markets, especially bubble markets, are creatures of narrative. Once everybody already owns the story, you don’t need catastrophic news to cause a selloff. You merely need marginally less perfect news. Just ask Leo.
That’s how bubbles usually start coming apart. Not because everybody wakes up simultaneously and announces the emperor is naked, but because a few people begin quietly wondering why his pants look funny. In this case, we’re much closer to everyone pronouncing all at once that the emperor is naked, if you ask me.
I’m not predicting that stocks are definitely going to crash Monday morning. But I am saying Monday morning is going to be worth watching very closely. Last month, I said I thought the window for the AI bubble bursting was somewhere between late 2026 and early 2027. Last week, I argued that regulatory and safety concerns could become the catalyst Wall Street wasn’t properly pricing.
Now, over the course of a weekend, one of the most important frontier AI CEOs has explicitly called for slowing capability development, another has publicly agreed with him, OpenAI has shelved a 2026 IPO while citing the safety environment, and coordinated industry action is suddenly being discussed openly.
And all of this is happening while inflation is still hot, oil is above $100, gasoline has surged, the 10 year Treasury is basically at 5%, the Fed could hike again within days and nearly $2 trillion of off balance sheet commitments are sitting underneath an AI capital spending boom whose economics depend heavily on demand continuing to explode.
Monday morning just got a hell of a lot more interesting.
QTR’s Disclaimer**:** Please read my full legal disclaimer on my About page here.
Contributor posts, guest posts and curated posts have been hand selected by me, but have not been fact checked and are the opinions of their authors. They are either submitted to QTR by their author or reprinted under a Creative Commons license with my best effort to uphold what the license asks, or with the permission of the author.
I cannot guarantee the accuracy of any or all facts and figures included in this article though I made an effort to get them right. I have been wrong before and will be wrong again, and encourage you to always double check, do your own research and speak to a licensed financial professional, which I am not.
This is not a recommendation to buy or sell any stocks or securities, just my opinions. I often lose money on positions I trade/invest in. I may add any name mentioned in this article and sell any name mentioned in this piece at any time, without further warning. None of this is a solicitation to buy or sell securities. I may or may not own names I write about and are watching. Sometimes I’m bullish without owning things, sometimes I’m bearish and do own things I’m bearish on. Just assume my positions could be exactly the opposite of what you think they are just in case. If I’m long I could quickly be short and vice versa. I won’t update my positions.
*Starting in 2026, I have been attempting to no longer actively trade as much as I once did ( read my story here). My goal is for my investing/s…
