this post was submitted on 25 Jun 2024
45 points (100.0% liked)

TechTakes

1435 readers
97 users here now

Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.

This is not debate club. Unless it’s amusing debate.

For actually-good tech, you want our NotAwfulTech community

founded 1 year ago
MODERATORS
dgerard@awful.systems
45
Nvidia is being a bubble stock again (pivot-to-ai.com)
submitted 5 months ago by dgerard@awful.systems to c/techtakes@awful.systems
25 comments fedilink hide all child comments
top 25 comments
sorted by: hot top controversial new old
[–] self@awful.systems 17 points 5 months ago (1 children)

The silly stock prices are bets on AI going up forever. But bubbles pop — and the shine has been coming off AI. We estimated that the present AI hype would have two years of desperate venture capital funding — but AI industry peons we spoke to earlier this year gave the VC money only until about the end of 2024.

god that’s not even a particularly long bubble, though I feel like the short expected duration of this one goes hand in hand with how much non-tech people hate AI now — it took a few years of well-thought-out criticism for crypto to go from almost universally hyped to the shit you warn your grandma to not waste money on

while I am rooting for it to happen, it feels like the tech industry’s earned a particularly hard crash this time, and it’s the nature of the fucked up beast that it’ll impact tech workers like you and me more than it’ll ever affect Huang or Altman or anyone else responsible for it. hopefully we can all brace appropriately for a thoroughly miserable time.

[–] o7___o7@awful.systems 16 points 5 months ago* (last edited 5 months ago)

These bubbles seem to be inflating and popping with shorter and shorter cycle times, as if the VC-fueled bullshit economy is bumbling toward some kind of ending. It reminds me of a supermassive star burning it's way down the periodic table until its heart is chock full of iron.

I'm probably just seeing patterns in the noise, but i~~t's a nice thought innit?~~ no you're right, it'll probably blast the innocent bystanders, mostly.

https://lco.global/spacebook/stars/high-mass-star/

High mass stars go through a similar process to low mass stars in the beginning, except that it all happens much faster. They have a hydrogen fusion core, but much of the hydrogen fusion happens via the CNO cycle. After the hydrogen is exhausted, like low mass stars, a helium core with a hydrogen shell forms, then a carbon core, with helium and hydrogen shells. Then unlike low mass stars, they have enough mass that gravity contracts the core raising the temperature and carbon can fuse into neon, then neon into oxygen, then oxygen into silicon, then iron. Each stage of burning lasts a shorter time than the previous one. For example, in a 25 solar mass star, hydrogen burning would take about 7 × 10^6^ years, helium burning 7 × 10^5^ years, carbon burning 600 years, neon burning 1 year, oxygen burning 6 months and silicon burning one day. Once silicon has fused into iron, no more fusion occurs, as the fusion of iron requires more energy than it releases. The core therefore collapses and releases a huge amount of energy in an explosion called a supernova. In the centre of the debris from the explosion is an incredibly dense neutron star. If the star is massive enough, the neutron star will collapse further and form a black hole.

[–] Buffalox@lemmy.world 7 points 5 months ago* (last edited 5 months ago) (2 children)

Yeah people have been saying something along those lines since 2016. I admit I had no idea it could get to where it is now, but people called it insane when Nvidia passed a $100 around that time (before the splits), and they called it doubly so when it passed $300 (still before the splits). Now they are at $125 but with a split of 1:10 this year, and in 2021 they had a split 1:4, meaning one stock prior to 2021 would be 40 stocks today. So that's equivalent to a stock worth $100 around 2016 is now 40 stocks worth $5000 total! That's 50x the value in 8 years.

Nvidia has been boosted over several times first by crypto, then by (traditional) datacenters, and now by AI:

I've been thinking this was a bit too much for about 5 years now, but the world constantly hungers for more computing power, and apparently Nvidia is the best company to deliver that.

My bet was on AMD, because I thought they'd catch up to Nvidia (apart from taking market share from Intel). While AMD was still a good bet, Nvidia would have been way better.

For 8 years betting against Nvidia has been the wrong bet. Maybe it's finally at the end of the fairy tale, but I wouldn't bet on it.

[–] dgerard@awful.systems 12 points 5 months ago (1 children)

oh sure, it's an excellent company as a long-term hold. This is still a bubble and I think the numbers will go down when it pops.

[–] Buffalox@lemmy.world 1 points 5 months ago (1 children)

I agree this increase is insane and without parallel AFAIK. It seems to be too good to be true, and normally that means it actually is too good to be true.
It's also hard to believe that Nvidia will be allowed to stay ahead in a market with such profit potential.
But for 8 years they've not only kept their market advantage and profit margins. They've actually managed to increase both.
I thought Nvidia was crazy when they made chips that yielded only one per wafer, I thought AMD would beat them with more agile chips at a fraction the price.

So the question is when you call this a bubble, do you mean demand for compute will stop increasing as much as it is? Do you think AI will flop?

In some ways I think AI already flopped somewhat from the initial hype of ChatGPT. But I do not think that will stop the race among companies to implement AI to keep up or stay ahead og the competition. It seems like AI is like a weapons race, and I bet it will go on for at least 5 years. Then if it turns out not to be the advantage companies hoped for, it will obviously slow down.

But I think Nvidia has a few years of good running ahead of them yet.

[–] dgerard@awful.systems 11 points 5 months ago* (last edited 5 months ago) (2 children)

I think this "AI" (a marketing term) bubble will flop per the article we linked in the post: this is a bubble fueled entirely by investors, mostly VC, throwing money into a fire of unprofitable businesses selling services at a loss that don't actually work.

I spoke to a pile of AI industry peons who think the VCs will get sick of setting money on fire by the end of this year. I woulda given it two years myself - there are hundreds of billions of dollars desperate for a home - but the number I was consistently hearing in March was "three quarters".

Maybe the Nvidia price will weather it. It took a hit when crypto crashed, though not a huge one.

[–] TinyTimmyTokyo@awful.systems 9 points 5 months ago

The next AI winter can't come too soon. They're spinning up coal-fired power plants to supply the energy required to build these LLMs.

[–] curiousaur@reddthat.com 2 points 5 months ago

The problem with betting on or against it is that you have to two factors to place your bet. What and when. You're right about the what, but can you time the top? I'm not betting on that either.