>>107703794
thats actually a dumb move.
ngreedia has gotten hard value out of the shatbot craze. the datacenters all run their cards
and palantir provide surveliance/governance tech and that make every single zogged government extra-hard
theyre 2 of the winners of the shatbot-craze

>>107703848
nvidia is investing in all those ai companies buying their cards, if they fall, nvidia is fucked too

>>107703794
btw palantir is a winner of the shatbot bonanza because they rely on compute to cross reference profile data with outcomes
and theyre gonna get compute for extra cheap once the fad passes, because the datacenters supporting ai-lmaoo are gonna be in dire need for clients
tons of offer with no demand = very cheap shit

>>107703794
>"isn't too early"
Implies that there will actually be a crash, which is not a guarantee at all.

Didn't this guy make a big deal about not investing anymore because he kept being wrong

>>107703864
nah, they have hard value
they make gpus, after all, not vague promises

their stock may get justed, but its not gonna fall down to 0
its hard to tell what will be the timelines
and its not certain to begin with, everything depends on how the hodl plebs will react

its risky and theres not much to gain.
i think its dumb

>>107703794
>Michael you get:
>A) live a normal fulfilling life as a multimillionaire
>B) be a billionaire but only be right ONCE in your entire life about anything
Guess he still made the right choice

>>107703909
>they make gpus, after all, not vague promises
and after the crash a huge load of them that were bought for data centers that never materialized will be dumped on the market for cheap

>>107703888
his opinion is influential and he is probably trying to expedite it.

>>107703924
thats where palantir may intervene
but i sure hope im wrong on this one, and were gonna see a used enterprise-card bonanza

>>107703794

why is he shorting a company tied to the gobbirment

>>107703848
Yes Nvidia is the biggest winner. But also their stock will fall the farthest. That's why it's a bubble.

>>107703794
There won't be a crash, what's gassing the price of DRAM and other components is literally just lack of supply constrained because manufacturers want to PREVENT a crash in the price of DRAM, there is already 1 HMB fab coming online. The simple fact is that DRAM and all other components are still worth less than the value AI provides, and much AI value is still locked away.
1 example, with enough compute and memory you could compress down all files on your system 10-100x and generate or decompress the needed ones on the fly. Not just files, identified blocks of memory can be replaced with the code that generated it, a super garbage collection ability.
GPUs where already shooting up in price before AI literally because of Leisure (video games) and mining. These datacenters could mine or become render farms, HPC centers, they don't have to do AI rendering.
Of course Google is already using AI as part of their DeepResearch program, the value of which already dwarfs the value of computer parts.

>>107705378
>mining
Btc needs asics and eth moved to pos years ago. This makes me suspicious of the other things you're saying, but I don't care enough to research.

>>107705378
The problem was never components (that's just the consumer-gamer's problem) its the datacenter expansion that's the "bubble". Increasingly more datacenters are being built, needing more power/networking infra and causing component shortfalls in a build-out cycle that should have been over a year ago. The problem is all that capital is being financed with debt/circular buyouts instead of hard assets (which tech has been hording for the past 10 years). The debt, like in 2008, might cause a domino affect for anything (land, operator, software, hardware etc) data-center related. AI has been, and still is, an unproven productiviy enhancement; it promises to automate almost everything but the hard automation (self driving cars, robotics) has yet to go mainstream and the easy automation (ai blogs, sass, image gen, slop etc) has shown little benefit, its just automating bullshit jobs that shouldn't be there in the first place..

>>107705482
My point is that there is no line where there are "too many" datacenters. They literally turn sunlight or sludge or gas into intelligence or at worst rendering. It can't be "overvalued" because the inputs are all worthless and the outputs is literally capital.
But to fit my argument into something you're economic theory can understand. Even with this massive rollout datacenters are still massively under produced due to supply and labor constraints.