I wake up, I read articles about how the AI crash is coming. I figure that it's not actually coming because everyone is talking about it, so it's priced in.
And then I pull up HN / LinkedIn / X / Substack where everyone is telling me that this is the new AI economy and the tech can't possibly flop, so I figure that the crash must be coming because there's too much optimism.
The public opinions on AI are split between two camps
the folk who don’t like AI slop forced in every product and media (creative/artistic AI backlash)
And the people who see the power of LLMs for coding, investigative work, science
I think there will be a crash with the creative side and all the companies that just adopted it without a real plan (FOMO) and sunk millions into it too quickly
But the actual tech is real and never going away. This is not a dot com bubble. This is a paradigm shift for people who know how to leverage an LLM effectively
> decide that it's not coming because everyone is expecting it and pricing it in.
I'm in the same boat. I'm hoping that the crashing is moving slowly enough that it can unwind carefully, and that not enough data centers got started that things can't simply be cancelled without much harm.
A bunch of fake money borrowed on commitments over data centers never built and plowed back into the ouroboros rather than escaping into the larger economy seems like it could be unwound with only a few people going broke. The only wasted real production would be in all of the chips - and they might be able to go somewhere, somehow. The data centers currently built will certainly be well-utilized - the overcapacity hasn't actually, physically been done, right?
If a bunch of circular financing went into doing something that never fully happened, is that a crash or a lifting of a mask over a real economy that has continued its previous slide?
And if capacity ends up fully utilized and prices do go up, won't there be dozens of viable half-done projects, the best of which that could be dusted off and rolled out again?
People should put money at stake on their idle speculations instead of being another doomer posting consequence-free wishful bias confirmations, hoping AI will fail and leave them be at their now-pointless jobs. This is also why prediction markets are useful signal, which I'm sure is another topic that rustles your jimmies. https://polymarket.com/event/ai-bubble-burst-by
Well, now the CFO is in charge of nVidia we may see some problems.
Because while a founder CEO can play fast and loose with gray finance rules, I don't expect an appointed CFO can do the same. At least not one that expects to work again if they are found liable for it. It is kind of interesting in this way that the board has made this choice. It almost says they are looking to insulate against some liability here.
The more cynical framing, especially on this site, is that the "Founder CEO" has a presumptive immunity from prosecution for banal stuff like bookeeping. They aren't perceived as beholden to the shareholders in the same way, because we in the peanut gallery think they "own" the company more than mere investors do.
But obviously no, in the real world fraud is fraud and if you get caught cooking books with other people's money you go to jail.
It’s no longer smart to predict an AI bubble pop. The question is only when. Any article trying to sound smart without giving dates and justification for those dates isn’t saying anything.
Hasn't the AI bubble already popped a bit? I mean Oracle is at less than half its ATH. SuperMicro (another company suspected of smuggling hardware to China) is not worth 1/4th of its ATH.
What does TFA mean by an "AI Crash"?
That the rushed Anthropic IPO at 3% of the US GDP won't 10x to 30% of the US GDP?
That Micron (30% down from ATH) won't be selling much RAM anymore? Or that NVidia (30% down from ATH) won't have buyers for its shovels?
We have seen complete irrational exhuberance in the stock market but some stuff are already coming down back to earth (Datadog which reached 550 P/E TTM just took a more than 50% slap in the face if I'm not mistaken).
What are the candidates for the pop? Palantir? What else?
It is going to break the back of SaaS. We can't outspend AI on infrastructure. We've had a global application deployed on GCP then AWS and I've never seen a stockout until this year, and now we get them so often we are being forced to reserve capacity. I also can't wrap my head around the basic economics. If you spend 10x yearly global gross revenue sales on building out infrastructure, software can only get more expensive to make.
I don't think there is going to be a crash anymore. I have witnessed some capabilities with custom harnesses that make it really difficult to buy the doomsday story.
"custom harnesses" are a manufactured construct for a loop - the compounding error will never go away due to limitations on transformers (larger context doesn't fix this nor any of the shell game tools that rotate on your x/twitter) - this is old news (we're a few years into this).
I mean, from personal experience, larger context, smarter models, and better designed harnesses have transformed LLMs from "lol this piece of shit hallucinated a function signature" to me not having written a single line of code since fall 2025, compounding errors be damned.
The dotcom bubble didn't burst because the web was pointless and doomed, it burst because it was massively overvalued.
AI being a transformative technology and AI companies being massively over leveraged betting on returns they can never realize are not mutually exclusive scenarios.
That is only one side of the equation. Whether something is overvalued has two elements: What is it's true value (unknown) and what is the price people are paying for it.
If you have witnessed miracles with a custom harness that's awesome, but it could still mean the valuation is too high, or maybe it's too low? Without putting specific numbers on it, it's an incomplete picture.
52 comments
[ 0.23 ms ] story [ 39.8 ms ] threadNobody seems to be able to read the article.
into https://codebeautify.org/htmlviewer
And then I pull up HN / LinkedIn / X / Substack where everyone is telling me that this is the new AI economy and the tech can't possibly flop, so I figure that the crash must be coming because there's too much optimism.
I go to sleep. Then I wake up and read...
Really? It feels like most people are talking about an AI bubble, with a vocal minority in opposition.
the folk who don’t like AI slop forced in every product and media (creative/artistic AI backlash)
And the people who see the power of LLMs for coding, investigative work, science
I think there will be a crash with the creative side and all the companies that just adopted it without a real plan (FOMO) and sunk millions into it too quickly
But the actual tech is real and never going away. This is not a dot com bubble. This is a paradigm shift for people who know how to leverage an LLM effectively
The Internet and personal computing was also a paradigm shift. Paradigm shift and bubble are not mutually exclusive.
I'm in the same boat. I'm hoping that the crashing is moving slowly enough that it can unwind carefully, and that not enough data centers got started that things can't simply be cancelled without much harm.
A bunch of fake money borrowed on commitments over data centers never built and plowed back into the ouroboros rather than escaping into the larger economy seems like it could be unwound with only a few people going broke. The only wasted real production would be in all of the chips - and they might be able to go somewhere, somehow. The data centers currently built will certainly be well-utilized - the overcapacity hasn't actually, physically been done, right?
If a bunch of circular financing went into doing something that never fully happened, is that a crash or a lifting of a mask over a real economy that has continued its previous slide?
And if capacity ends up fully utilized and prices do go up, won't there be dozens of viable half-done projects, the best of which that could be dusted off and rolled out again?
That's correct. The big companies may fall, but thanks to research papers, open models, etc, LLMs aren't going to disappear.
AI hype organizations like Latent Space (+ others) ban people who disrespect any aspect of AI today, and articles like this get buried.
(...The design hack to make the geocentric models work. Maybe it's also a metaphor...)
By definition, the conventional wisdom cannot predict big changes, whether positive or negative.
Because while a founder CEO can play fast and loose with gray finance rules, I don't expect an appointed CFO can do the same. At least not one that expects to work again if they are found liable for it. It is kind of interesting in this way that the board has made this choice. It almost says they are looking to insulate against some liability here.
But obviously no, in the real world fraud is fraud and if you get caught cooking books with other people's money you go to jail.
What does TFA mean by an "AI Crash"?
That the rushed Anthropic IPO at 3% of the US GDP won't 10x to 30% of the US GDP?
That Micron (30% down from ATH) won't be selling much RAM anymore? Or that NVidia (30% down from ATH) won't have buyers for its shovels?
We have seen complete irrational exhuberance in the stock market but some stuff are already coming down back to earth (Datadog which reached 550 P/E TTM just took a more than 50% slap in the face if I'm not mistaken).
What are the candidates for the pop? Palantir? What else?
Automatic research will create large amount of value without a way to capture it, because there will be no moat for anyone any more.
Everything cheap, with no buyers.
We could be in perfect storm of demand shock due to failure to redistribute gains to people, insufficient societal change.
IP value is already showing cracks.
Overbuild, very similar to fiber overbuild in 2000 will lead to defaults of established players.
Monetary system will have to print unprecedented amount of money and put it into hands of people; will it actually?
Do you think there is no value in the Internet? Because that crashed.
Do you think there is no value in housing? Because that crashed
A market correction, aka a crash, merely implies that the valuation is is unreasonable for it's current impact.
The "crash" isn't going to happen because ai has no value, it's just going to be a chain reaction from some companies having overextended themselves
In 10 yrs a 1T valuation may be plausible, but today isn't 2036
AI being a transformative technology and AI companies being massively over leveraged betting on returns they can never realize are not mutually exclusive scenarios.
If you have witnessed miracles with a custom harness that's awesome, but it could still mean the valuation is too high, or maybe it's too low? Without putting specific numbers on it, it's an incomplete picture.