Model inference will be done on the edge. Apple makes their own silicon and will never return to NVDA, so count out all iDevice revenue. Same with Tesla.
OpenAI trained on Nvidia cards, but doesn't OpenAI's ChatGPT run on Nvidia cards as well? So it's not just the training, but the execution as well, right?
I train Dreambooth models for Stable Diffusion, which takes up a lot of GPU time, but then I generate images with those Dreambooth models, which takes less time (but still uses my Nvidia card).
And after they train a model, what are they going to do with it?
Easy: Within 20 years, we'll see the first publicly-traded company > $100B trading at 1000x EBITDA, before crashing down to 200x EBITDA.
As shown in the last few years, money can be loaned out easily by the Fed & banks in times of extreme economic stress. With more crises inevitably coming up in some form or another, more loans will eventually be given out to incentivize the continued existence of the market.
all the major techs are massively overbought because its become a safety trade