Creative destruction

This has been a big 2 weeks for local models with 3.8 27B and Next coming out, a lot of "frontier labs are done" comments which, in general, could certainly be true. Could be trillions in economic value evaporating here over the next few months as "normies" catch on to what just happened in the model space. However, one thing that keeps coming up over and over is the "We need the frontier to keep the open labs working/releasing models". And while I don't disagree with that, I think it's important to realize what actually happens if the frontier labs go under. The models don't disappear. The code doesn't evaporate. And, perhaps most importantly, the chips don't blink to another dimension never to be seen again. What would very likely happen if one or both of them go under is that there would be a massive glut of hardware suddenly available that people will buy up at pennies on the dollar (Coreweave/etc), drastically lowering the price of inference. Their models will be bought up and development will continue, perhaps at Google or another "strong hand" in the AI model business. The important thing to realize, none of this stuff disappears. What happens is that the value gets baselined again at a MUCH lower number. The model that cost 100B to develop is sold off at 5B. The rack of chips that cost a few million bucks sell at a million bucks. The strong hands that come in to take these companies apart will be in a VASTLY better position to compete with and develop new, strong models to compete with China. The only scenario where these companies and their technology just "disappears" is if there's no value at all in what they have. I think we all know that's not true, even if you discount the models to 0 they still have a lot of hardware that has a lot of value. Now, on a negative note, yes, it's going to crash the economy to some degree if/when it happens. There so much circular dealing going on that one of them going under is going to spread the hurt around all over the place, but, out of the other side will hopefully come new companies that aren't burdened with an absurd cost basis and can actually compete in the market against the open models. Oh, and yes, hopefully/likely hardware will get a lot more affordable without the 2 goliaths committing to buy the next 58 quadrillion chips that NVDA can produce. :)