Buggy Loop
Member
And voila
This is just an accelerator to strike it rich faster on robots / autonomous cars / etc. They won't simply stop buying or slow down because a model is more efficient
Its an arm's race. There's no slowing down.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
NVIDIA stock to lose $400 billion ( US tech 1 Trillion) after DeepSeek release
- Thread starter Bernoulli
- Start date
And voila
This is just an accelerator to strike it rich faster on robots / autonomous cars / etc. They won't simply stop buying or slow down because a model is more efficient
Its an arm's race. There's no slowing down.
Schmendrick
Member
Pff in a few days Nvidia will announce their last numbers and they'll still be printing money like there is no tomorrow.
Panic reactions to literally nothing are just a stock thing. We're way past the point where stocks in general have more than a loose correlation to what is actually going on.
Panic reactions to literally nothing are just a stock thing. We're way past the point where stocks in general have more than a loose correlation to what is actually going on.
E-Cat
Member
Seriously, did people's brains fall out of their ears or something??And voila
This is just an accelerator to strike it rich faster on robots / autonomous cars / etc. They won't simply stop buying or slow down because a model is more efficient
Its an arm's race. There's no slowing down.
"Hey, did you hear there's a new, more efficient model out of China?"
- Yikes, I guess that means the scaling laws for foundation models and inference no longer apply, stop buying GPUs!
GHG
Gold Member
And voila
This is just an accelerator to strike it rich faster on robots / autonomous cars / etc. They won't simply stop buying or slow down because a model is more efficient
Its an arm's race. There's no slowing down.
The arms race looks a lot different now though.
Due to the open source nature of deepseek it means companies can adapt and intergrate it at will. You no longer need to use Nvidia hardware, you no longer need to go to one of the big players for an AI solution, and you certainly don't need huge nuclear power plants running in the background feeding it energy in order for it to be viable.
Yes you can sit there with millions of Nvidia chips and say you have the most capacity - but who exactly are you selling to and at what price?
Buggy Loop
Member
"Yea! Let's slow down on spending". Said nobody ever when it comes to disrupting tech and even worldwide domination when they are on the cusp of having robotics, self driving cars and AGI models aiming towards singularity and change the world (for better or worst is anyone's guess).
Someone found a gold vein that's easier to dig. He ain't gonna stop buying the best shovel and slow down. Total nonsense. But I guess that's what wallstreet wants to portray right now on media. Speculate a crash and then buy low. Oldest trick in the world with media reporting on stocks.
Little Mac
Member
Is this a “buy the dip” situation? I only invest in broad market ETFs and crypto.
DeaconOfTheDank
Member
Not really. DeepSeek's approach implies that you can get much more out of your hardware resources for a fraction of the cost - a massive boost to efficiency.
Nvidia still holds the bleeding edge when it comes to the hardware required for building AI, and this news doesn't change anything. If anything, AI development will accelerate even more rapidly than it had before now that compute limitations have been mitigated.
This is still an arms race, and all sides of the competition won't want to be left behind. They'll still go for the best resources available, especially since the value proposition is much more enticing since they can squeeze even more out of the high-end GPUs. Nvidia is actually very well positioned considering how much they're banking on massively increasing tensor cores as opposed to compute units...
GHG
Gold Member
AI Industry May Never Earn Back Its $150 Billion Nvidia Chip Spend
With AI revenue $500 billion short of AI hardware investment, David Cahn believes we are witnessing the growth of a financial bubble.
www.forbes.com
And 15% down in a single day with this kind of volume being traded is not "speculate a crash". It's a crash no matter how you want to spin it.
Buggy Loop
Member
You're thinking like someone doing AI in his basement at home. That's neat and all, yea he has a free model, yea he doesn't need to buy blackwell to generate hentai porn.
That's not how the big boys tech firms play.
The new model scales with inference speed and scalability on GPUs. They didn't invent an ASIC solution ala bitcoin here GHG, the model scales fully the more you throw inference computational power at it. They found a more efficient model BECAUSE they are restricted on using bigger GPUs, they didn't find a way to do it without GPUs.
There's as much chance as a snowflake in hell that these companies will lift the foot from the pedal, in fact the panic is now who is gonna be using this model and the biggest AI farm to beat everyone first to robotics or autonomous cars, or even AGI. The panic is going towards USA going even more FULL IN to not let China get the advantage.
AI Industry May Never Earn Back Its $150 Billion Nvidia Chip Spend
With AI revenue $500 billion short of AI hardware investment, David Cahn believes we are witnessing the growth of a financial bubble.
And 15% down in a single day with this kind of volume being traded is not "speculate a crash". It's a crash no matter how you want to spin it.
I mean peoples are dumb, what can you do about it. If they invested in AI without understanding the principles, no wonder they are reactionary to this.
Like I said, the biggest ones that should be hit with a trading blow are the tech firms behind the models.
Last edited:
ozz ex machina
Member
I'm upset because my stocks have gone from being over 1000% up to only being 950% up.
E-Cat
Member
It's impossible to calculate a ROI on the Singularity; except to say there will probably be only one or two winners, make your pick carefully...AI Industry May Never Earn Back Its $150 Billion Nvidia Chip Spend
With AI revenue $500 billion short of AI hardware investment, David Cahn believes we are witnessing the growth of a financial bubble.
And 15% down in a single day with this kind of volume being traded is not "speculate a crash". It's a crash no matter how you want to spin it.
viveks86
Member
This is what I'm thinking as well. Seems too speculative. Stock market gonna stock market, I guess.
viveks86
Member
Except this is historically not how an arms race would work right? If they are more efficient, they will just max out performance and sell even more throughput and better response times so they can leap frog the competition. This only makes sense once the technology reaches some saturation point (singularity?) and there is no point in improving further.
Last edited:
Buggy Loop
Member
I see you are a man of culture
That's what USA / China governments are chasing for
If they fear China even sniffing the idea of beating them to the punch, well let's just say ...
You'll never have seen a funding as big as what's about to come. Forget the Space shuttle / Apollo program / Manhattan project. It'll look cute compared to the funding AI will get.
Singularity is the single most disruptive thing that will happen to mankind. The country being first to it will leap everyone and its exponential once you have it.
Its not because you find a more efficient way to reach your goals that you lift the foot from the pedal. AI transformers for youtube videos is not the end game, far from it.
Last edited:
GHG
Gold Member
You are missing the forest for the trees.
- There is no "China advantage" - it's open source and every single company can learn from and adopt deepseek's methodologies.
- Yes of course this new model scales, but if you are an AI solutions provider are you really going to run 1000 GPU's at full tilt when your current customer base only dictates you need to run 10 at full tilt?
Bottom line is as follows - if you give a company an opportunity to cut cost with no decrease in output then they are going to take it. Remember, that was the primary selling point of this whole AI revolution in the first place.
Ask yourself this - who is buying and at what prices are they willing to buy at?
When the mooted costs are higher than just employing people for certain disciplines, would a business rather opt for a person or an AI solution?
They were already struggling to monetise things as is. Scaling up doesn't solve that problem. First they need to figure out what customers will want from all this, then you are free to scale up accordingly.
Last edited:
MiguelItUp
Member
I'm sure things will work out for them regardless, I mean, they're sitting pretty on quite a lot at the moment.
That being said, I wish they would get pushed around some more, they need it.
That being said, I wish they would get pushed around some more, they need it.
Bernoulli
M2 slut
But what is singularity? we are already seeing exponential advancements in AI
GHG
Gold Member
And would you say the eventual winner will be a single hardware solutions provider?
Because that is precisely how Nvidia has been trading over the last 18 months or so.
Buggy Loop
Member
kiphalfton
Member
A lot of stuff is way down today. Nvidia, TSMC, AVGO, Marvell, etc.
Buggy Loop
Member
We're not even at AGI. It's a long road. A chinese model being more efficient is welcomed news. Doesn't change the outlook in financing and mega AI farms with nuclear power.
E-Cat
Member
Are you aware of the o3 model? It cost like $350 grand and ~16 hours for a single SOTA solution to ARC-AGI... if they put 10x more money in, the results would have been even better. How much money and inference time would you be willing to spend in order to solve the Riemann hypothesis? Or devise a better way of building microchips? Or a more efficient AI learning algorithm? There's literally no limit, the fact that it's now available to everyone will only highlight the importance of compute supremacy. A lead of 1 month could make or break the winner, the US gov is gonna go absolutely apeshit pedal to the metal on this one...
Last edited:
It's open source bro. Go read the github files instead of making up conspiracies on a game forum. This what our youth has become and we wonder why China has outpaced us technologically.
AtomicStarving
Member
This will hit them hard. But competition is always good. Get f***ed
Mister Wolf
Member
Massive drop for Nvidia. I'm still in the green though.
E-Cat
Member
We don't even have autonomous research organizations comprising human genius-level AI agents, that's still 2-3 years away. Talk of the singularity before that is premature
Last edited:
James Sawyer Ford
Member
Nvidia’s days at these valuations are numbered
Massive bubble
Massive bubble
GHG
Gold Member
I know its open source
New model means Stargate can reach AGI faster and beyond AGI. Why would they cut costs.
I feel like I've detailed enough how scalability and arms race will not slow down things. If speculators are that dumb with stock market then so be it, they created a dip.
It's not about things "slowing down". Meta is up, Microsoft is attempting to rally back, Apple (whose chips will benefit the most from this due to their architecture) are up. These (and all other competing) companies will benefit greatly from deepseek and can now use what they already have (along with looking at alternatives to Nvidia - yes, this may well bring AMD back in to the picture if the methodologies can be adapted to get around AMD's chip to chip issues) to scale up at no extra cost.
For nvidia though, the days of them hard selling dozens of fully equipped AI server racks at a time may be well and truly gone.
While an unintentional consequence, this directly attacks Nvidia's most. You no longer need to use CUDA and you no longer need to make sure you have access to the massive bandwidth Nvidia's "virtual GPU's" offered. It's ironic, because the only reason this has come about is because of the restrictions placed on China when it comes to nvidia chips. Funny what can be achieved in the world of software when you're forced to be efficient
Everyone in tech wants AI so far up their asses so it can physically wipe them cheeks, it doesn't make any sense
sendit
Member
What is weird about it? Literally one of the easiest financial concepts.
With that said, everything is down. This is a nothing burger. I'm buying 2 5090s given this news (/s).
Last edited:
daninthemix
Member
Up! Down! Up! Down!
Don’t you guys know how the stock market works yet?
Don’t you guys know how the stock market works yet?
sendit
Member
Last edited:
Jaguar Victory
Member
I'm assuming the concern will be what the administration does and how it cracks down on NVIDIA exports again after new reports.
SweetTooth
Gold Member
Lol
People who sold are smart, Nvidia stock is inflated as f***
People who sold are smart, Nvidia stock is inflated as f***
Last edited:
GHG
Gold Member
The bold is a horrible assumption.
What this has proven is that if you want more gains in a short amount of time then you should focus more on improving the efficency of your software instead of waiting on and relying on more (or better) hardware.
In some tasks we have a 25x improvement - how long would we have to wait (or how much more would you need to spend) to get the same result via hardware alone?
Last edited:
E-Cat
Member
No one is waiting on better hardware, the megacaps are simply spending exponentially more each turn (yes, obviously this will hit a wall sooner or later). And, yes, there are also the 3-4x generational leaps in hw every couple of years.
It's cool that we got a one-time improvement like this, but the vast majority of compute in the future is going towards test-time inference compute. We don't even know the inference cost/performance curve of DeepSeek-R1... more data is needed.
OpenAI are now iterating every ~3 months on the o-series, they can incorporate whatever RL lessons there are to be had from the DeepSeek paper on o4. o3 got 25% on the research-level math FrontierMath benchmark, do you realize how much compute is going to be required for ~90% on this test -- with, or without, this new efficiency gain?
Last edited:
Buggy Loop
Member
Look, you're hoping for cheap GPUs and that's fine
But you're way off with "the time of fully equipped AI server racks is gone", like completely delusional.
Has nothing to do with Nvidia or AMD. Tech firms will pick the best inference available and scale it as high as they can with whatever power they have available to make it run, nuclear power and all.
Can the shovel seller switch hands in the future? Sure. But as of now, even AMD for free is still more expensive than Nvidia for these firms when they consider the whole picture of total cost of ownership.
While an unintentional consequence, this directly attacks Nvidia's most. You no longer need to use CUDA and you no longer need to make sure you have access to the massive bandwidth Nvidia's "virtual GPU's" offered. It's ironic, because the only reason this has come about is because of the restrictions placed on China when it comes to nvidia chips. Funny what can be achieved in the world of software when you're forced to be efficient
You always need an interlink for any massive GPU scaling. What are you even on about. You're talking about things you don't understand. DeepSeek doesn't change GPU scalability. Its incredible the misinformation.
Oh look
AMD Instinct™ GPUs accelerators are transforming the landscape of multimodal AI models, such as DeepSeek-V3, which require immense computational resources and memory bandwidth to process text and visual data. AMD Instinct™ accelerators deliver outstanding performance in these areas.
Requirements for bandwidth and computational ressources not going away.
Tech firms forever have managed to do without Cuda, how else is AMD selling anything?
That's not what they buy Nvidia for, not the big tech firms at least. Even OpenAI which uses Nvidia GPUs has its inhouse solution, Triton.
When you're that big you have the software inhouse. Cuda again, you're thinking advantages like someone making hentai porn in his basement.
ChuckeRearmed
Member
Time to buy I guess
P:axMagellanic
Member
Got to love seeing Jensen's nose being rubbed in the shit.
Also, trumps trade war and bidens continuation of it will prove to be the biggest geopolitical bungle of this century. It forced a sleeping giant to awaken.
Also, trumps trade war and bidens continuation of it will prove to be the biggest geopolitical bungle of this century. It forced a sleeping giant to awaken.
noob smokes
Member
jason wangs gonna take this out on pc dudes the price of a brick going up
Buggy Loop
Member
Yea me too
How many times I can explain the same thing over and over
Peoples have never heard of the Jevons paradox it seems.
Last edited:
ChuckeRearmed
Member
It just means that with more powerful hardware you can achieve even more. This will drive hardware adoption faster across all the lines.