Nvidia Shares Resume Fall | Hacker News Reader

Nvidia Shares Resume Fall

ft.com

66 pointsbelter66 comments

This is probably a massive overreaction. No one is cancelling orders and the future of AI is still going to heavily involve GPUs. If not for LLMs, then for image and video generation.

Of course, if the stock was overly inflated from false expectations, this could certainly be a resetting to the normal.

Also, I can't help but feel like DeepSeek is a red herring for the market overall. The stock market's overall dip didn't happen until after the slew of Executive Orders hit. The federal spending cuts are certainly going to affect the bottom line of businesses, and they add to the seriousness of some of the other anti-economic threats like tariffs.

Trying to convince us or you?

I hope it pushes Nvidia to hobble less of the consumer grade offerings for AI/inference.

Part of this recent selloff is going to be the bubble popping effect, but it's clear that inference will be the main demand driver, and that's something where ddr5 is competitive.

This a reaction to well sourced rumors, the US Administration will impose additional restrictions on the sales of Nvidia chips. This is in addition to the TSMC tariffs.

I assume its just a market correction. I don't think AI is a reason, I just think some people come up with really bizarre conclusions and then the rest regurgitate it. Now had a major Nvidia competitor announced some highly competitive GPU or some AI that uses drastically less resources and can run on any GPU or CPU type of thing, maybe then I'd understand. Nvidia is still the key GPU in AI though.

Their PE ratio is nearly down to 50, leaving it only about 10% higher than the S&P 500 on the eve of its crash in 2001.

How this will be called in the future? Dot ai bubble?

The total income of the s&p 500 has never grown by > 100% annually. Not saying Nvidia isn’t overvalued, but your comparison a bit apples to oranges.

> This is probably a massive overreaction.

You may be rights but here is the other side. The AI/software/hardware/data center/energy trade got very crowed by levered money(Hedge funds).

1) That trade has had a huge run up lead by NVidia with alot of funds hedging their gains till the new year for tax reasons. They were al looking for a reason to sell and now could continue to lighten even more.

2) Nvidia is a large portion of the market right now. If you are a hedge fund that is measured against a US index then you probably have bitten the bullet and bought some NVidia exposure so you don't underperform if it continues its growth and the AI story fulfills its promise.

Now that NVidia is falling and you can make a story that it may under perform the market t his year you may see even more fast money dumping it as they no longer need to worry as much about it skewing market returns as much now.

3) if its true that you now need atleast an order of magnitude less power and chips to train a model then Nvidia was due to have a multiple decrease. This knocks onto the data center and power trades that also will have a multiple decrease on their pricing.

You may be right that this is an undeserved sell off, but given the recent growth it also seems like you can make the case that the whole trade got ahead of themselves and we were due a pull back to more properly value these companies earnings prospects.

One thing that is always useful to remember about the markets is that when you are priced for perfection like NVidia was, its signs that your long term growth may be slowing will affect your share price more than short term signs that your are doing what was expected of you.

>This is probably a massive overreaction.

No. I know people working on even tighter optimizations of LLMs. The magic is not in compute, its in design, and its feasible that once all the kinks are ironed out something akin to O1 could be run effectively on a Raspberry Pi; a whole GPU would yield much better results but it probably won't be necessary to buy 3, or 4, or a whole rack of them. Nvidia's stock prices soared on the expectation (from those who weren't in the know) that if massive amounts of capital were simply moved to the right places it continuously improve. The thing that drives tech forward is people, however, not capital.

if o1's can be run from a rasp pi, do you see farms of ARMs in the future?

Are farms the future?

I could see an "AI appliance" like the old Google Search Appliance: A single rack unit with a single $1000 GPU would probably be enough to run a pretty-robust DeepSeek style product, sold as "100% self-contained and on-prem, so you can trust it with propriatery data".

I'm pretty sure Middle-Eastern investors are having second thoughts on sending more money to OpenAI, Anthropic, xAI, etc. And plenty of AI companies are priced out of the market. And a week ago Google priced their new models along the lines of Deepseek.

Sure, in a while demand will get back up. But by then maybe Nvidia will not be the only game in town anymore.

Yeah, I think NVDA could fall another 20% and its stock price would still imply "massively profitable company primed for continued explosive growth." Its previous valuation was a premium even above that story.

>This is probably a massive overreaction.

yeah for sure, but its an over reaction to an over reaction (the AI bubble and Nvidia's role in it)

I can't help but wonder what fraction of this is import tariffs on Taiwan. Also the initial drop, can't help but wonder if someone leaked that info to some private traders.

pelosi's husband sold millions in nvidia stock https://www.newsweek.com/nancy-pelosi-sells-nvidia-stock-wee...

Pelosi isn't even in the top 5 most prolific traders in Congress, despite being one of the most high profile ones. Congressional trading should be illegal and Pelosi needs to retire yesterday, but the constant focus on her isn't particularly justifiable.

Then ignore this information.

For others, like me, it's important.

Justifiable because she's consistently beaten the S&P500's gains by a factor of 3-4x over the last decade.

Exercised 500 call options purchased on Nov. 22 for 50K shares, purchasing on Dec. 20 when the options were set to expire. Sold 10K of those 50K shares on Dec. 31.

Purchased 50 call options Jan. 14, 2025

Why purchase another 5,000 shares Jan 14 if they have insider information they should sell?

If we only look at the hardware (not the software) for inference, does Nvidia have any advantage over Intel, AMD and Google?

They all build GPUs/TPUs that can run LLMs, correct?

Nvidia GPUs over significantly more performance per card than Intel and AMD. In general the limiting factor is total computation, so people will pay a premium for the best. I'm not familiar with Google's hardware, and don't think it's generally available.

What about AMD's MI300X? I see reports all over the web that it runs LLMs at a similar speed as an H100.

It’s available on GCP through Cloud TPU, though Nvidia on GCP is still a far bigger business.

Can you run LLAMA on Google's TPUs?

Yes. You can serve Llama on TPU: https://cloud.google.com/vertex-ai/docs/predictions/serve-ll...

You can also train Llama on TPU: https://cloud.google.com/tpu/docs/v5p-training

Nvidia is the only one that does this on the sort of scale these AI companies need. They do boxes packed full of AI accelerators with custom high speed bus connectors and high end interconnectivity to make super computers full of GPUs very easy to purchase combined with the software to make it all work. No one else has that sort of integration to the very large that is necessary for the sort of training the top firms are doing.

Yes, that's why I asked about inference.

For inference, Nvidia's strengths seem to be not very important.

You can do inference on a single GPU. And AFAIK the software stack is not important for inference either. Because you don't have to experiment with the sofware. You just need to get it to run and then you will run it for a long time unchanged. Groq for example runs LLAMA on their custom hardware, correct?

And I expect hardware for inference to become a bigger market than hardware for training.

Deepseek v3 needs 16 H100s or 8 H200 GPUs for inference.

Or a single 2 processor AMD EPYC rig, for less than $6k.

https://xcancel.com/carrigmat/status/1884244369907278106

The only reason you need all those GPUs is because they only have a fraction of the ram you can cram in a server.

With AMD focusing on ram channels and cores the above rig can do 6-8 tokens per second inference.

The GPUs will be faster, but the point is inference on the top deepseek model is possible for $6k with an AMD server rig. 8 H200's alone would cost $256,000 and gobble up way more power than the 400 watt envelope of that EPYC rig.

I thought the whole story about Deepseek was that Deepseek does not have H100s?

This is all a bit hyperbolic. The price is still over twice what it was a year ago, and the same price as in October, three months ago.

yes, it was in the 90s last August.

The US Administration is preparing new restrictions on sales of chips to China.

Something I did not expect, apparently nvidia is the stock with the 9/10 largest stock losses in a day. So it’s quite volatile. People just don’t understand if this meteoric rise will hold or not and so suspicious. Makes sense they try to take money out if they can. On the other hand, retail investors also put 900mil into the stock yesterday

Just let me know when it's at the bottom

"Shares resume fall" - I think you really mean it resumes being a super volatile stock. On a weekly basis it gains and drops 10-15%. I know because I buy when it drops 10-15% then sell when it pops back up. I've made over $20,000 trading on Nvidia volatility alone.

It used to stress me out, until I realized that I could profit off the volatility.

I don't think this is an effect of the bubble bursting. This week it swung from Friday closing around $148 to this week bottoming out at $118. Currently it's trading at $121 which is fairly normal for its range within the last 3 months. It's been averaging around $120-135 for the last 3 months. I know because I trade on it weekly. The sudden "investor confidence" is mostly the excitement of the AI super investment by Trump, hence the valuation up to $148 (Wish I had a sell order at 145 not 150, I would be in the money so much).

What about wash sale

What about wash sales? Why would that matter? If OP is always selling at a profit then wash sale doesn't matter. Even if it did, then it only matters if you want to try claim a loss at end of year...

I trade in retirement accounts, so taxes don't apply.

Given that both "shares" and "resume" can function as either a noun or a verb, this headline keeps messing with my brain.

Also „Fall“ may be a noun or verb.

I read this as Nvidia shares resumé fail. I sometimes think headline writers treat confusing headlines as an artform

What a great headline! I love the idea of someone pausing and resuming their fall :)

California Dept. of Finance panicking.

If I had the money I would be buying right now.

I suppose that may be in fact why you don't have the money

Good one.

My $100k line of credit against my house was approved today. I plan to pay off my house with the gains.

Only if you are shorting :-)

Realistically, in the long run Nvidia's stock will soar. I can't believe people will think otherwise.

I wish I was so confident about future events.

Careful what you wish for, sounds like a recipe for disaster to me.

Realistically, in the long run everything goes to zero.

The market can stay irrational for longer than...

NVDA's market cap was already 12.5% of the entire yearly US-GDP, at a P/E of 10 or 20 once their growth slows to me that implies that their profits are expected to be 0.5-1% of the US GDP. Just nvda alone, not the LLM foundation models + application layers and so on built on top their hardware.

Even in the most optimistic case, I find it hard to imagine that nvidia alone captures much more than 1% of the US GDP as profits for the next 20 years or whatever their investor's horizon is.

Nvidia's P/E numbers are lower than almost all of their major competitors. You can't look at Nvidia's ratio alone and say "this seems overvalued" without also acknowledging the fact that the entire industry looks ridiculous by the same analysis.

I'd even agree with that argument, I just don't think it's the one you're making.

I think we agree - my response was narrowed to respond to "NVDA is going to soar". The whole segment looks overvalued and the implication (at least to me) seems to be that AI will capture 10-20% of GDP in profits (not even revenue) starting soon for 10-ish years.

I don't have the numbers, but I wonder how "internet company" (very loosely) profits compare to GDP - I could imagine the AI ecosystem reaching similar profits eventually.

Btw, imo their P/E is mostly this good because they're currently selling hardware at margins better than many software companies, I doub that this will/can last. For example, when FB/MS each are spending $50B+/a on GPUs, can they justify say $100M/a on a crack software dev team to make their stack work on other GPUs (ie replace the oft-argued CUDA moat within a year or two)?

Please consider getting in touch with a problem gambling hotline. This isn't Internet snark--going into debt to play the stock market game is a serious red flag.

I assumed it was a joke.

But yeah, absurdity:10 does not mean insincere!

You sound like the people who were buying "shitcoins" in the cryptocurrency frenzy.

Best of luck (you'll absolutely need luck, because that's a stupid move).

People who share links behind paywalls didn't understand how the internet works.

need archive link

https://archive.is/zOrRr

Reply on news.ycombinator.com