Nvidia’s $589B DeepSeek rout • Hacker News Board

Re: Nvidia’s $589B DeepSeek rout

#41

Post by spacebanana7 » Mon, Jan 27, 2025, 11:32 AM UTC

InkCanon wrote: post #20

Discordian93 wrote: post #5

It's funny the bubble seems popping because the tech can be made better or cheaper instead due to loss of enthusiasm for the tech itself. AI is here to stay I suppose then.

If anything Meta and other foundation model companies should be dropping. What's the moat? DeepSeek is releasing stuff for free.

Meta should actually go up from this. If deep seek is perfect they don't need to pay for an expensive Llama team. And even if deep seek isn't perfect, the low cost training strategies they've invented could be used by Meta to reduce the cost of Llama training.

Although Meta develops models they don't sell them. So a world where foundation models are free is fine for them.

Re: Nvidia’s $589B DeepSeek rout

#42

Post by torginus » Mon, Jan 27, 2025, 11:32 AM UTC

pjc50 wrote: post #26

The whole AI valuation was based on being able to rent-seek a significant chunk of all white collar salaries, with a permanent monopoly moat because nobody else would pay hundreds of billions to train models. And yet again a cheaper Chinese product turns up and everyone loses their minds. Expect a ban incoming to preserve the AI valuations.

Banning wouldn't work imo - now that the cat's out of the bag, with the architecture being open source, anyone can replicate their results to a compete with them for a relatively small investment.

Re: Nvidia’s $589B DeepSeek rout

#43

Post by zo1 » Mon, Jan 27, 2025, 11:32 AM UTC

Wow, literally bought more Nvidia shares last week. Just goes to show the stock market is 80% gambling. Me believing in the value of a company is overshadowed by the hype of growth and "future valuation".

Re: Nvidia’s $589B DeepSeek rout

#44

Post by cbg0 » Mon, Jan 27, 2025, 11:32 AM UTC

dogma1138 wrote: post #14

diegocg wrote: post #12

Earlier quoted context omitted.

Nvidia is way too overvalued regardless of deepseek or the success of AI. This is just some correction (not even too big even considering the current bubble), these traders are not stupid.

NVIDIA has a P/E ratio of 56 that’s double that of the S&P 500 but half that of AMD and the same one as Meta. And whether it’s overvalued or not isn’t relevant that selling a stock because the product the company produces is now even more effective is mind bogglingly stupid.

> selling a stock because the product the company produces is now even more effective is mind bogglingly stupid.

No it isn't. Investors are most likely expecting there will be less demand for Nvidia's product long-term due to these alleged increased training efficiencies.

Re: Nvidia’s $589B DeepSeek rout

#45

Post by fvv » Mon, Jan 27, 2025, 11:33 AM UTC

based on meta declarations , amd is better at customized inference than nvda ,

meta uses amd mi300x for all inference and which bought 173.000 gpu ( 40% of their total gpu count ) according to that report https://www.cnbc.com/2023/12/06/meta-and-microsoft-to-buy-am...

Re: Nvidia’s $589B DeepSeek rout

#46

Post by dogma1138 » Mon, Jan 27, 2025, 11:33 AM UTC

locallost wrote: post #33

dogma1138 wrote: post #14

Earlier quoted context omitted.

NVIDIA has a P/E ratio of 56 that’s double that of the S&P 500 but half that of AMD and the same one as Meta. And whether it’s overvalued or not isn’t relevant that selling a stock because the product the company produces is now even more effective is mind bogglingly stupid.

Well the price has a built in presumption that the earnings will keep growing. That's why PER is not that relevant for them, it's been over 50-70 since forever, but the stock went up 10x, which means earnings went up as well. DeepSeek might be good for their business overall, but it might mean earnings will not continue growing exponentially like they have been for the past two years. So it's time to bail. You should…

Why? The compute requirements would still continue to grow the more efficient and more capable the models become.

If it’s cheaper to inference you end up using the model for more task, it it’s cheaper to train you train more than models. And if you now need only 1000’s of GPUs instead of 10’s or 100’s of thousands you’ve just unlocked a massive client base of those who can afford to invest high six to low seven figures instead of 100’s of millions or billions into to try their luck.

Re: Nvidia’s $589B DeepSeek rout

#47

Post by gsky » Mon, Jan 27, 2025, 11:33 AM UTC

It doesn't matter whose model is better. AI is here and it won't go anywhere. As a result companies that feed AI will see better valuations

Re: Nvidia’s $589B DeepSeek rout

#48

Post by ppsreejith » Mon, Jan 27, 2025, 11:33 AM UTC

fire_lake wrote: post #7

Fascinating. I think Meta is a big winner from this - they still control the content and now mining it for value has been proven even cheaper by DeepSeek.

Agree. Apple should be as well. The only con I can think of would be their (Meta's) data center investments but seems this will make them more efficient?

Re: Nvidia’s $589B DeepSeek rout

#49

Post by bythreads » Mon, Jan 27, 2025, 11:33 AM UTC

nabla9 wrote: post #15

gostsamo wrote: post #11

Earlier quoted context omitted.

Consider that the chinese might be misrepresenting their costs. A newsletter was implying that they might do it to undermine the sanctions justifications. Agree that the AI bubble should pop though and the earlier, the better.

Their model is open and they published paper describing it https://arxiv.org/pdf/2412.19437 The can't be far off, or it would be noticed. Even if they are heavily government subsidized for energy and hardware, I don see how the cost of training in the US would be more than double.

Doesnt that say they based it on llama?, sooooo not really a bottoms up training - since the cost of llama is 100% surely not part of their quote.

Re: Nvidia’s $589B DeepSeek rout

#50

Post by Cumpiler69 » Mon, Jan 27, 2025, 11:33 AM UTC

Etheryte wrote: post #38

dogma1138 wrote: post #6

Good time to buy then, I don’t understand how stupid some traders can be. A more efficient model is better for NVIDIA not worse. More compute is still better given the same model. And as more efficient models proliferate it means more edge computing which means more customers with lower negotiating power than Meta and Google… This is like thinking that if people need to dig only 1 day instead of an entire month to ge…

While everything you say may be true, this shows a fundamental misunderstanding about how the modern stock market functions. How much value a company creates is at best tangential and often completely orthogonal to how much the stock is worth. The stock market has always been a Keynesian beauty contest, but in the past few decades, strongly shaped and morphed by attention economies. A good example of this is DJT, a c…

>In other words, NVIDIA is in the red not because the company is suddenly doing worse, but because traders think other traders think it will trade down.

Well put. People need to unterstand that some stocks are basically one giant casino poker table. There was a comment with a link here that a lot of Nvidia buyers don't even know what products Nvidia is making and they don't care, they just want to buy low and sell high. Insert old famous comment abut shoe shine boy giving investment advice to Wall Street stock traders.