I think the big picture that many people are missing here is the motivation that all these AI/tech companies have for buying up so many GPU's in the first place: achieving AGI/ASI.
And while some still try to portray a dedication/duty to AI Alignment, I think most have either secretly or more publicly moved away from it in the race to become the first to achieve it.
And I think, given that inference time compute is so much cheaper than pre-training, the first to achieve AGI might have enough compute on hand from having been to first to build it that they would not need to purchase many more GPU's from Nvidia. So at some point, Nvidia's revenues are going to decline precipitously.
So the question is: how far away are we from AGI? Seems like most experts estimate 3-10 years. Did that timeline just shrink by 50x (or at least by some multiple) from these new optimizations from DeepSeek?
I think Nvidia's revenue is going to continue to grow at its current pace (or faster) until AGI is achieved. Just sell your positions right before that happens.
It doesn't feel like DeepSeek has a big enough breakthrough here. This is just one of many optimizations we're going to see over the next years. How close this brings us to "AGI" is a complete unknown.
The large investments were mainly for training larger foundation models, or at the very least hedging for that. It hasn't been that clear over the last 1+ years that simply increasing the number of parameters continues to lead to the same improvements we've seen before.
Markets do not necessarily have any prediction power here. People were spooked by DeepSeek getting ahead of the competition and by the costs they report. There is still a lot of work and some of it may still require brute force and more resources (this seems to be true for training the foundation models still).
Not that I believe it's likely to happen, but it seems incredibly fucking dangerous for there to be an ASI race with one winner. To the extent these companies believe it's possible, what are they hoping will be the outcome for humanity?
That they get to be the trillionaires with an untouchable moat? Wouldn't this be like creating a Kwisitz Hadarach thinking you can control it, to borrow a Dune reference?
And while some still try to portray a dedication/duty to AI Alignment, I think most have either secretly or more publicly moved away from it in the race to become the first to achieve it.
And I think, given that inference time compute is so much cheaper than pre-training, the first to achieve AGI might have enough compute on hand from having been to first to build it that they would not need to purchase many more GPU's from Nvidia. So at some point, Nvidia's revenues are going to decline precipitously.
So the question is: how far away are we from AGI? Seems like most experts estimate 3-10 years. Did that timeline just shrink by 50x (or at least by some multiple) from these new optimizations from DeepSeek?
I think Nvidia's revenue is going to continue to grow at its current pace (or faster) until AGI is achieved. Just sell your positions right before that happens.
(Not an expert. Not even an amateur)