low demand? Got a source for that? Anthropic is paying crazy sums to direct competitors to meet demand, so it seems that demand is quite high. Could be wrong, though.

> What bubble is left? Really. Just people with OpenAI or Anthropic exposure? Not that high.

The stock prices of 9/10 of the top S&P 500 stocks have significant AI exposure. A drop in valuation of them will be widely felt.

▲

simianwords a day ago | parent [-]

I agree there's no low demand.

> The stock prices of 9/10 of the top S&P 500 stocks have significant AI exposure

not true.. only ~2-3% exposure at model layer. rest are at hardware layer..

▲

bryanlarsen a day ago | parent [-]

And there's a significant bubble in the hardware layer. The Nvidia overvaluation alone far exceeds that of both OpenAI and Anthropic. When the bubble pops people are going to want dedicated inference chips without an 80% profit margin rather than the B200.

▲

simianwords a day ago | parent [-]

why would the hardware layer bubble pop if the demand for hardware will still increase (or not change)?

▲

bryanlarsen a day ago | parent [-]

Because everybody and their dog are designing inference chips. Same as tokens, high demand at commodity prices means low profits.

TSMC will still make massive bank because they're the bottleneck. That bank will be squeezed out of companies dependent on TSMC.

▲

simianwords a day ago | parent [-]

> commodity prices means low profits.

I still don't understand how models being commoditised implies the hardware (NVIDIA) itself being commoditised.

What is the exact mechanism that causes NVIDIA's profit to go lower?

Do you mean to suggest that the bubble popping will simultaneously commoditise models as well as inference hardware?

They are completely different layers and have almost zero correlation. I see no indication that hardware also gets commoditised.

	▲	bryanlarsen a day ago \| parent [-]
		everybody and their dog are designing inference chips. Even currently, Google's Tensor can do inference cheaper and more efficiently than a B200. There are dozens of companies designing inference chips; when they hit the market the pricing power of B200 will drop significantly. And it's correlated with the other bubble dropping because chip cost will matter a lot more once the margin of tokens is tight.