NVIDIA stock to lose $400 billion ( US tech 1 Trillion) after DeepSeek release
- Thread starter Bernoulli
- Start date
diffusionx
Gold Member
market cap is a fake number. I hate this style of financial "reporting"
Guardian Monkey
Member
It was -14% earlier. Looks like people are putting their brains back in their head.
SlimySnake
Flashless at the Golden Globes
I always knew nvidia was inflated. good to see.
Everyone going all in on AI is equally moronic. AI is just a tool. its the products that matter.
Everyone going all in on AI is equally moronic. AI is just a tool. its the products that matter.
Buggy Loop
Gold Member
What we see from openAI and other tech AI companies is what they want the public to see. There's models internally that we're not prepared for so USA in panic for world domination of AI is a cute facade, they'll use this to put fear into government and twist arms to force massive money injection into it.
od-chan
Member
Yeah obviously, but it's easier to built stuff with a hammer than with a spoon.
Unknown?
Member
Eerily reminiscent of Pets.com before the dot com bubble popping.
Drake
Member
The real reason behind it, I think is because tech companies want to use it to replace highly compensated coders. Especially in more niche languages where some of these people are pulling in 300k+ a year.
Last edited:
For all i've heard and read about the IT sector in USA, they were becoming bloated and inneficient, their margins and futre promises to investors were maintaning that bubble, there's a possible reckoning coming where the pressure will demand to cull the fat because they know it won't affect the company or make it productive. It was almost akin to a government body.
diffusionx
Gold Member
It's quite convenient that Trump gets into office with this tech bro support, he announces an initiative for AI development, which is fine, and then IMMEDIATELY after, this Chinese model arises out of nowhere that is somehow vastly superior to what the tech bros have been doing.
This is space race, "missile gap", Cold War type shit all over again. People need to smarten up!
Windle Poons
Made a crappy phPBB forum once ... once.
Nvidia will be the scalpers themselves on the 5090s now to get their money back.
Insane Metal
Member
Nuts. Market cap is a weird thing.
Bernoulli
M2 slut
Isn't this what they have done?
Sell to the government the idea that AI is magic and they can only do it with the 500 billions
Then china releases a superior model for free, open source and doesn't require that much computing
Silver Wattle
Gold Member
I tried to search for technical reviews of deepseek but can only find the mainstream regurgitate, they keep claiming it's great with no depth or independent research.
Makoto-Yuki
Banned
from what i've read they do use nvidia but lower spec ones.
i think the whole thing is that they are matching OpenAI with less powerful hardware and money?
Drake
Member
It's super efficient, so it requires way less resources which means less money for nvidia if this becomes the new thing.
Darkness_Arises
Member
Much older and cheaper Nvidia GPUs, showing that you do not need Nvidia's latest and greatest.
Yeah it's kind of strange although in a way there is a connection. The efficiency gains may mean lower amounts of H100s required by customers for their needs and so less orders.
Bernoulli
M2 slut
It's not from no where, they have been working on it for 2 years, there were some previews and older models from deepseek before R1
No matter how ahead you are designs end up converging
That's why Warplanes for example look similar between us and china , over a long time engineers from each side reach the same conclusion, it's not about copying
Makoto-Yuki
Banned
wouldn't you still need the best hardware? maybe the problem is in education/skills?
if this is what china can do with old GPUs then what can they do with the new ones? What could USA do if they were more efficient?
if this is what china can do with old GPUs then what can they do with the new ones? What could USA do if they were more efficient?
Last edited:
Bernoulli
M2 slut
FalconPunch
Member
Good, Nvidia needs a massive haircut. The stock is greatly overvalued rn.
kruis
Exposing the sinister cartel of retailers who allow companies to pay for advertising space.
It's 20x more CPU efficient than other AI commerical models. That's bad news for Nvidia's and OpenAI business. Why invest 500 billion dollars in new Nvidia super computers and run OpenAI models when you can switch to DeepSeek and get 20 times the performance for free?
ReyBrujo
Member
Cobol coders on suicide watch.
Bernoulli
M2 slut
Ai 4X upscaled stocks
DeaconOfTheDank
Member
$NVDA stock is on sale
Magic Carpet
Gold Member
If only China would start making chips and dumping them on the world. We could all just walk into gas stations, buy gas, burrito, video cards. and a couple of charging cords that will short out my cards usb ports.
Guardian Monkey
Member
lol 
. This is like the SuperZip of AIs. An AI program that's probably actually using other AI sources and just scraping. Ofcourse it uses less resources.
"Unlike many Chinese AI firms that rely heavily on access to advanced hardware, DeepSeek has focused on maximizing software-driven resource optimization," explains Marina Zhang, an associate professor at the University of Technology Sydney, who studies Chinese innovations. "DeepSeek has embraced open source methods, pooling collective expertise and fostering collaborative innovation. This approach not only mitigates resource constraints but also accelerates the development of cutting-edge technologies, setting DeepSeek apart from more insular competitors."
www.wired.com
. This is like the SuperZip of AIs. An AI program that's probably actually using other AI sources and just scraping. Ofcourse it uses less resources."Unlike many Chinese AI firms that rely heavily on access to advanced hardware, DeepSeek has focused on maximizing software-driven resource optimization," explains Marina Zhang, an associate professor at the University of Technology Sydney, who studies Chinese innovations. "DeepSeek has embraced open source methods, pooling collective expertise and fostering collaborative innovation. This approach not only mitigates resource constraints but also accelerates the development of cutting-edge technologies, setting DeepSeek apart from more insular competitors."
How Chinese AI Startup DeepSeek Made a Model that Rivals OpenAI
When Chinese quant hedge fund founder Liang Wenfeng went into AI research, he took 10,000 Nvidia chips and assembled a team of young, ambitious talent. Two years later, DeepSeek exploded on the scene.
Buggy Loop
Gold Member
AI is like a gold rush
Nvidia sells the shovels
China, USA, they both use the same shovel
Models are much like miners trying to find a gold vein. All this stock speculation should have the biggest impacts on peoples making models like openAI, even though I still think that their public models are not what's on the other side of the curtain and where the billions have been injected.
DeepSeek uses Nvidia GPUs. It's not because that it doesn't require as much computing that an AI farm will think of reducing GPU count, he just got more computational power on same hardware with a new model. He'll want everything he can. Especially since newer generations typically have better power efficiency and racks. These are costly for AI farms, regardless of model.
In no way this is a message that Nvidia will sell less GPUs, that's insane. On the contrary, this "panic" mode tech bros are doing on TV is to inject even more cash into it. This is an arm race. Although, DeepSeek is open source so its a matter of time in finding out how they optimized it.
Last edited:
Jevons paradox - Wikipedia
E-Cat
Member
YepI don't think this will hurt nvidia. It just allows more companies to buy nvidia chips.Jevons paradox - Wikipedia
en.m.wikipedia.org
Like, we just had the revolution called test-time compute paradigm, you can literally throw infinitely more exponential compute at any reasoning problem for linear gains in performance. It's perpetually insatiable, now you can solve even harder problems for the same cost
Last edited:
Bernoulli
M2 slut
Google, Apple and Meta are all rushing to make their own AI chips to not rely on the expensive nvidia ones
With this new model showing that you need much less compute it's easier for them to switch away from nvidia and use your own chips or alternatives like AMD
Buggy Loop
Gold Member
Oh yea, because they have a new model, they'll want less computational power and let a competitor buying Nvidia get the upper hand. Nonsense.
No it doesn't work like that. All Google, Apple, Meta and Microsoft's GPUs outside of Nvidia are always for previous models. They put Nvidia for the most top line research.
It's more about extracting more value out of your high-performing resources. Salary is inconsequential in many regards, it's more about how many $ does each employee generate.
Lower level/performing folks are the ones who should be worried.
BennyBlanco
aka IMurRIVAL69
justiceiro
Marlboro: Other M
AI buble popping,?
BennyBlanco
aka IMurRIVAL69
Forget the coders. That NVidia CES presentation was nightmare fuel. AI cameras watching warehouse workers. AI robot chef arms.
If robots can efficiently do skilled labor jobs like a chef, we are so fucked as a species. It's gonna eliminate a ton of middle class jobs and make billionaires more rich.
GHG
Gold Member
GHG should I buy bro?
I posted about this potentially coming a few weeks ago:
All I will say is to combine the above with the information regarding efficiency gains that deepseek has managed and come to your own conclusion.
The days of businesses specifically needing a high volume of specialised Nvidia chips to run AI solutions may well be long gone. If everything in the research paper rings true, it's a complete paradigm shift and the first of many to come. This is akin to the first general purpose chip coming along.
Using their solution you no longer need to use Nvidia chips to get it running well. And even if you decide to use Nvidia chips, you certainly don't need as many.
Last edited:
E-Cat
Member
You don't "need less compute", you need as much compute as you can muster. All this means is there's now a double-exponential instead of an exponential.
