Note that the mantra doesn’t say the market can stay irrational, it’s just a warning that if you are predicting when irrational failure will produce the expected results, you can be in for a bad time. So if someone is thinking about shorting a bunch of the heaviest bubble stocks and expecting to win, well, they can get screwed as the bubble sustains long enough to screw them and then pop.
To be clear, that doesn’t mean AI is going away. It just means no one is actually going to pay for AI models anymore because open-weight free models will be extremely cheap and powerful.
In this scenario, you have a number of these AI companies contributing to the hoarding having their equipment handled through ‘asset recovery’, which means that at least companies that can drive 15kw to a system and water cooling will probably get them on the cheap and run it on their premise, or in a colo. Maybe some of those parts will trickle down, but admittedly a good chunk of the stuff is hard to accommodate in a residential setting.
Longer term, the hardware becomes obtainable as supply chains re-calibrate back to identcal or more close solutions. Ten years ago, a datacenter GPU was likely to be same hardware as consumer, but with a different thermal solution, firmware, and the video ports unpopulated. The AI rush has made them shift to exotic packaging so they can have absurdly unreasonable wattage in small places that doesn’t work in home settings. I anticipate a swing back that way eventually.
I really hope so but I can’t help but to think that they are going to drag it for as long as possible, because no matter how bad the situation is for the common folk, they are still going to make a profit off of it.
Yeah. The current job market shows a different trend.
With people saying “people using AI will replace people not using AI”.
And the reason for that is not because people using AI will produce better work, but because AI usage will be preferred over usable output.
And because the flow of money is such that all those having money can easily choose to give most of it to AI users[1] while non-AI users don’t have the same ability once most providers turn to AI use.
Now as long as you get the Governments on board (they are already buying up GPUs too) you can get all taxes into AI use, hence starving the non-AI market of ability to procure computing hardware (or anything else to work on) and that is how you get AI supremacy without providing anything better.
with the only exception being base material products like agricultural produce, which has a much lower margin and their costs again go to AI users ↩︎
We’re in the late stages of the AI bubble.
The market can remain irrational longer than you can remain solvent.
Except then you read this https://www.usatoday.com/story/money/economy/2026/02/05/most-job-cuts-2009-economy-faltering/88525316007/
Note that the mantra doesn’t say the market can stay irrational, it’s just a warning that if you are predicting when irrational failure will produce the expected results, you can be in for a bad time. So if someone is thinking about shorting a bunch of the heaviest bubble stocks and expecting to win, well, they can get screwed as the bubble sustains long enough to screw them and then pop.
God I hope so
To be clear, that doesn’t mean AI is going away. It just means no one is actually going to pay for AI models anymore because open-weight free models will be extremely cheap and powerful.
It also means that AI in places where it brings nothing and in many cases makes the product actually worse will disappear
How is anyone going to run those powerful models locally once the necessary hardware is unaffordable/unobtainable?
No one is paying for AI because no one wants it to begin with.
In this scenario, you have a number of these AI companies contributing to the hoarding having their equipment handled through ‘asset recovery’, which means that at least companies that can drive 15kw to a system and water cooling will probably get them on the cheap and run it on their premise, or in a colo. Maybe some of those parts will trickle down, but admittedly a good chunk of the stuff is hard to accommodate in a residential setting.
Longer term, the hardware becomes obtainable as supply chains re-calibrate back to identcal or more close solutions. Ten years ago, a datacenter GPU was likely to be same hardware as consumer, but with a different thermal solution, firmware, and the video ports unpopulated. The AI rush has made them shift to exotic packaging so they can have absurdly unreasonable wattage in small places that doesn’t work in home settings. I anticipate a swing back that way eventually.
I really hope so but I can’t help but to think that they are going to drag it for as long as possible, because no matter how bad the situation is for the common folk, they are still going to make a profit off of it.
Yeah. The current job market shows a different trend.
With people saying “people using AI will replace people not using AI”.
And the reason for that is not because people using AI will produce better work, but because AI usage will be preferred over usable output.
And because the flow of money is such that all those having money can easily choose to give most of it to AI users[1] while non-AI users don’t have the same ability once most providers turn to AI use.
Now as long as you get the Governments on board (they are already buying up GPUs too) you can get all taxes into AI use, hence starving the non-AI market of ability to procure computing hardware (or anything else to work on) and that is how you get AI supremacy without providing anything better.
with the only exception being base material products like agricultural produce, which has a much lower margin and their costs again go to AI users ↩︎
All of that is speculation, none of it true. I’m an AI user, where’s my money?
While I was actually looking for someone to break my model with a better one with better info, this is a pretty useless argument.
To get money, you need to: