There is a huge difference between this bubble and the 2008 mortgage crash. In 2008 they made capital products out of debt. When that debt was worthless, so were the derivative products which caused a sudden cascading crash that wiped out trillions of dollars. This time, everything was paid for in cash and results in tangible assets. Even if AI suddenly became worthless, the money spent was real, the resulting goods were paid for with real money that doesn’t evaporate, the goods exist and can be sold or repurposed (granted those ai chips are bespoke and expeeeeensive and worthless for literally any other application than an LLM).
So sure - its a bubble. AI is not going to magically become useful or ever really turn a profit for the companies that invested many billions. It wont replace people. It will crash. But the crash won’t be a catastrophic destruction of capital. It’ll be a slower realization that assets are overvalued and it will unwind over the course of years. And yes, companies will go bankrupt and people will lose their jobs. But not nearly on the same scale as 2008.
Those models are still outrageously expensive to train. They may be free for you to use, but they aren’t free to make and once the existing hype train crashes no one with the resources to create this stuff is gonna wanna spend money training the next Mistral or Llama or Deepseek.
The existing models will continue to exist, sure, but they’ll be frozen in their current state and updated versions with newer training data won’t be released. The ecosystem will go stale — at least until someone figures out how to make training cheap which probably isn’t even technically possible.
When community models were first getting formed (going back to the pygmalion days), there were distributed systems that allowed you to “contribute” gpu power from your personal computer to help update the model or provide compute for others - I would imagine we would return to something similar to those roots.
This would not be deepseek - this would be the niche models that were made by hobbyists for hobbyists.
It’s not that there isn’t tangible value in AI, I mean the subprime crisis is an apt comparison. Packaged tranches of subprime mortgages had value too, hell they are still traded today. They had value too, it just became a question of was what the hell was that value, and the models and assumptions made to assign the value to them evaporated pretty much overnight, because there were a lot of flaws in the assumptions. As there is here. So then all of the sudden there was no bottom to be found, and trillions of dollars evaporated in the ensuing panic.
It’s kind of the same thing here, like what is the actual value of AI? It’s obviously got some uses, but other uses are massively overexagerated, at least in the current conditions of it. Maybe for the foreseeable future. Maybe forever, who knows?
No, the difference is that when those subprime mortgages were bundled as assets, they were intentionally classed as AAA debt when they were worth much less. This created value out of thin air and allowed a shell game to be played by Bear Sterns and Lehman Brothers with the big banks. The AI stuff is paid for in cash. It’s real money. It is still an incestuous investment and contracting game between the 7 major players. But that money came from the sale of stocks that are worth what they are due to market valuations and real trade (which is still bullshit).
When the public and investors come to realize AI was oversold and its abilities were exaggerated and AGI is still a decade+ away, the valuations won’t just vanish. The circular investing game will absolutely cause stock devaluations and they will be sudden. That $560 MS stock will poof go back to $420 when they first invested in OpenAI. That will really fucking hurt the average 401k. OpenAI will probably go bankrupt. Meta will still be fucking worthless as a societal drain but paradoxically worth nearly a trillion dollars. Over 10% of the GDP of the entire US in capital will absolutely be destroyed. But it won’t take place in a moment. Investors will be stubborn and bet on the dead cat bounce and hope and pray for their gambling addiction to pay off and keep those companies on life support.
The same shell game these companies are currently playing trading stock for rapidly depreciatibg assets then borrowing heavily against those assets which each unit represents a daily liability because they operate at a loss. When the investment capital runs out all these companies operating overheads will kill them in days/weeks, and the hardware they actually own, which loses about a third of its value vs purchase price upon first install, will flood the 2nd hand market at probably 5-10% it’s current value.
As usual panic in the markets will be driver of a fast crash. Nothing is certain but this is currently the most likely outcome based on past precedent of such a large majority of all investment capital being funnelled into something with such low (negative in this case) economic productivity.
It’s funny money. The story is always generally the same when it comes to this stuff. You sell to me at a higher value, I turn around and sell it back to you at a higher value, you turn around and sell it back to me at a higher value and so on. All those “gains” get “invested” into something else, and then I sell that at a higher value and you…
There is a huge difference between this bubble and the 2008 mortgage crash. In 2008 they made capital products out of debt. When that debt was worthless, so were the derivative products which caused a sudden cascading crash that wiped out trillions of dollars. This time, everything was paid for in cash and results in tangible assets. Even if AI suddenly became worthless, the money spent was real, the resulting goods were paid for with real money that doesn’t evaporate, the goods exist and can be sold or repurposed (granted those ai chips are bespoke and expeeeeensive and worthless for literally any other application than an LLM).
So sure - its a bubble. AI is not going to magically become useful or ever really turn a profit for the companies that invested many billions. It wont replace people. It will crash. But the crash won’t be a catastrophic destruction of capital. It’ll be a slower realization that assets are overvalued and it will unwind over the course of years. And yes, companies will go bankrupt and people will lose their jobs. But not nearly on the same scale as 2008.
How can we verify if things are paid for in cash?
Debt has to be issued somewhere and the banks arent saying that enormous loans are being made to pay for AI development
Ah yes, banks, the bastions of truth and moral conduct.
Well they’re required to release public reports to both the stock market and to the fed about debts issued. That’s not really a moral subject.
My confidence that these reports aren’t distortions of the truth is slim considering banks run on greed. Their track record doesn’t help either.
Yeah, about that… Nvidia is renting their GPUs to many bullshit generators.
https://www.cnbc.com/2025/09/24/nvidia-openai-investment-in-cash-mostly-used-to-lease-nvidia-chips.html
Calling it now : AI use is gonna become quite expensive all of a sudden, after that bubble bursts…
Not my thing, but I’d imagine liquidated hardware for self-hosting would be quite cheap for those who care about it that much.
Hopefully video cards get cheaper again.
Nah. There are models you can run at home that are nearly as good as the ones you pay for.
Those models are still outrageously expensive to train. They may be free for you to use, but they aren’t free to make and once the existing hype train crashes no one with the resources to create this stuff is gonna wanna spend money training the next Mistral or Llama or Deepseek.
The existing models will continue to exist, sure, but they’ll be frozen in their current state and updated versions with newer training data won’t be released. The ecosystem will go stale — at least until someone figures out how to make training cheap which probably isn’t even technically possible.
When community models were first getting formed (going back to the pygmalion days), there were distributed systems that allowed you to “contribute” gpu power from your personal computer to help update the model or provide compute for others - I would imagine we would return to something similar to those roots.
This would not be deepseek - this would be the niche models that were made by hobbyists for hobbyists.
So if 2008 was like ripping a bandage off quickly, this will be like ripping a bandage off really slowly?
So slowly that it heals as the bandage is pulled and causes a really weird scar
It’s not that there isn’t tangible value in AI, I mean the subprime crisis is an apt comparison. Packaged tranches of subprime mortgages had value too, hell they are still traded today. They had value too, it just became a question of was what the hell was that value, and the models and assumptions made to assign the value to them evaporated pretty much overnight, because there were a lot of flaws in the assumptions. As there is here. So then all of the sudden there was no bottom to be found, and trillions of dollars evaporated in the ensuing panic.
It’s kind of the same thing here, like what is the actual value of AI? It’s obviously got some uses, but other uses are massively overexagerated, at least in the current conditions of it. Maybe for the foreseeable future. Maybe forever, who knows?
It’s going to be one for the text books.
No, the difference is that when those subprime mortgages were bundled as assets, they were intentionally classed as AAA debt when they were worth much less. This created value out of thin air and allowed a shell game to be played by Bear Sterns and Lehman Brothers with the big banks. The AI stuff is paid for in cash. It’s real money. It is still an incestuous investment and contracting game between the 7 major players. But that money came from the sale of stocks that are worth what they are due to market valuations and real trade (which is still bullshit).
When the public and investors come to realize AI was oversold and its abilities were exaggerated and AGI is still a decade+ away, the valuations won’t just vanish. The circular investing game will absolutely cause stock devaluations and they will be sudden. That $560 MS stock will poof go back to $420 when they first invested in OpenAI. That will really fucking hurt the average 401k. OpenAI will probably go bankrupt. Meta will still be fucking worthless as a societal drain but paradoxically worth nearly a trillion dollars. Over 10% of the GDP of the entire US in capital will absolutely be destroyed. But it won’t take place in a moment. Investors will be stubborn and bet on the dead cat bounce and hope and pray for their gambling addiction to pay off and keep those companies on life support.
The same shell game these companies are currently playing trading stock for rapidly depreciatibg assets then borrowing heavily against those assets which each unit represents a daily liability because they operate at a loss. When the investment capital runs out all these companies operating overheads will kill them in days/weeks, and the hardware they actually own, which loses about a third of its value vs purchase price upon first install, will flood the 2nd hand market at probably 5-10% it’s current value.
As usual panic in the markets will be driver of a fast crash. Nothing is certain but this is currently the most likely outcome based on past precedent of such a large majority of all investment capital being funnelled into something with such low (negative in this case) economic productivity.
Nvidia is financing OpenAi using the loans against the stock value though?
Then Open Ai uses that money to pay Nvidia, which inflates their stock value.
Does not seem like cash to me.
It’s funny money. The story is always generally the same when it comes to this stuff. You sell to me at a higher value, I turn around and sell it back to you at a higher value, you turn around and sell it back to me at a higher value and so on. All those “gains” get “invested” into something else, and then I sell that at a higher value and you…