I agree in principle. It’s getting harder to believe that is possible at this point.
A combination of 1) companies maturing their AI policy and ending the “infinite play time” model most have had since 2021 e.g. cutting spending and 2) open source and open parameter models becoming better and better and cheaper. Data centers will still be in demand for #2 but the ai providers will hurt.
Some have, most haven’t. And those that have had mostly adopted the philosophy of “Every token spent is a a token spent well; except the ~10% of staff that abuse/misuse them”.
And with that philosophy, you naturally want to set the per-person budget at around the 90th highest spender. Which is what most did (some version of it).
So very few companies have actually set a truly restrictive policy. Just a “don’t abuse this” policy. But that will change IMO.
Calling a bubble is easy.
Calling the top of the bubble or the bottom of the crash is the more or less impossible part.
It could very well be a bubble, and come crashing down to a low point that is still higher than it is today.
I agree in principle. It’s getting harder to believe that is possible at this point.
A combination of 1) companies maturing their AI policy and ending the “infinite play time” model most have had since 2021 e.g. cutting spending and 2) open source and open parameter models becoming better and better and cheaper. Data centers will still be in demand for #2 but the ai providers will hurt.
I don’t follow this extremely closely, but didn’t most companies end number 1 at least 6+ months ago?
Some have, most haven’t. And those that have had mostly adopted the philosophy of “Every token spent is a a token spent well; except the ~10% of staff that abuse/misuse them”.
And with that philosophy, you naturally want to set the per-person budget at around the 90th highest spender. Which is what most did (some version of it).
So very few companies have actually set a truly restrictive policy. Just a “don’t abuse this” policy. But that will change IMO.