News Nvidia Breaks $1 Trillion Market Cap

[Admin]

Nvidia's market cap leaped over $1 trillion today, making it the first semiconductor maker in history to reach the mark.

Nvidia Breaks $1 Trillion Market Cap : Read more

 

[bit_user]

I hope this prompts AMD to triple-down on their investment in GPU compute, and finally support ROCm across all recent hardware models, including APUs.

You can't hope to compete with Nvidia on compute, if you don't have a vibrant ecosystem. And part of having a vibrant ecosystem is making it easy for people to develop on your hardware.

Another thought I have is that Nvidia's Grace represents a real threat to AMD's server CPU revenue, as well. I'll bet quite a lot of servers hosting Nvidia GPUs have been EPYC-based. Grace looks set to change much of that.

 

[Exploding PSU]

Welcome to the club!

 

[InvalidError]

A 'barrier' is a limitation that isn't meant to be crossed. Fluffed-up market valuations passing 1T$ is merely a late-stage capitalism milestone.

Another thought I have is that Nvidia's Grace represents a real threat to AMD's server CPU revenue, as well. I'll bet quite a lot of servers hosting Nvidia GPUs have been EPYC-based. Grace looks set to change much of that.

Nvidia has its Grace Hoppers, AMD has its MI300 and Intel has its Data Center GPU Max. No shortage of large integrated CPU+GPU solutions in the DC space.

 

[mhmarefat]

Only 1 trillion? Ok then that explains the predatory pricing behavior of its recent GPUs.
"Only" 1 trillion! Their very foundations are at stake here. They NEED to release a mainstream TRASH card and charge $400 for it because times are really rough. It is a matter of life and death for nvidia.

 

[bit_user]

Nvidia has its Grace Hoppers, AMD has its MI300 and Intel has its Data Center GPU Max. No shortage of large integrated CPU+GPU solutions in the DC space.

Grace is new and only beginning to finally ship. We don't know how much of AMD's revenue projections for EPYC were based on the assumption AI buildout would continue on trend, but Grace might invalidate such assumptions.

One place where MI300 falls short is that it's not CUDA-native. Unlike x86, you can't just deploy MI300 or Falcon Shores and run your CUDA workload on them. In theory, AMD and Intel can just support PyTorch, Tensorflow, etc. and you won't have to think about what's underneath, but I doubt most sophisticated AI customers would agree with that. They probably have custom layers written in native CUDA code.

Perhaps an even bigger reason is that Nvidia just keeps a tight grip on the performance crown. Every generation, they manage to defy being leapfrogged. Furthermore, they know how to scale like nobody else - aside from securing their performance superiority, it's also a major factor in selling massive numbers of devices.

Basically, the situation we're in is that everyone wants Nvidia. I tried to find some market research, but didn't find an easy answer to the AI marketshare Nvidia currently holds. I'm sure it's massive, though.

BTW, it sounds like AMD got the wakeup call. From earlier this month:

AMD Says AI Is The Number One Priority Right Now

When it comes to an economy, you get what we collectively expect. And given all of the talk and weird things that have happened in the early part of 2023,

www.nextplatform.com

 

[gg83]

I hope this prompts AMD to triple-down on their investment in GPU compute, and finally support ROCm across all recent hardware models, including APUs.

You can't hope to compete with Nvidia on compute, if you don't have a vibrant ecosystem. And part of having a vibrant ecosystem is making it easy for people to develop on your hardware.

Another thought I have is that Nvidia's Grace represents a real threat to AMD's server CPU revenue, as well. I'll bet quite a lot of servers hosting Nvidia GPUs have been EPYC-based. Grace looks set to change much of that.

I was thinking that GPUs would be replaced by a more AI processor in the near future. Do you think GPUs will be the processor of choice going forward? Can Nvidea stay on top? Or is their stock insanely overrated?

 

[Elusive Ruse]

Jesnen will be even more under pressure to deliver on growth and profits which means bad news for gamers since our needs are now truly an afterthought.

 

[InvalidError]

Jesnen will be even more under pressure to deliver on growth and profits which means bad news for gamers since our needs are now truly an afterthought.

That depends on how sustainable the AI gold rush will be. My gut feeling is demand will level out within two years, so I expect about two years before AMD, Nvidia and Intel go "Oh crap! We need something else more sustainable to dump all of our excess wafers on. Maybe we shouldn't have golden-showered gamers two years ago."

 

[bit_user]

I was thinking that GPUs would be replaced by a more AI processor in the near future. Do you think GPUs will be the processor of choice going forward?

I would love to see a perf/$ and perf/W comparison between their GH200 and Cerebras CS-2. Related:

​

Cerebras Smashes AI Wide Open, Countering Hypocrites

We could have a long, thoughtful, and important conversation about the way AI is transforming the world. But that is not what this story is about. What it

www.nextplatform.com

​

Can Nvidea stay on top?

Hard to say, but they have as many resources and as much expertise as anyone in the field. They won't be easy to unseat.

A couple years ago, I'd have thought Intel could do it, but I haven't heard a lot about Habana and the failures & shortcomings of Xe are known all too well. AMD is forever playing catch-up. Other AI startups don't seem to have gotten that much traction, but I'm not positioned very well to know.

Or is their stock insanely overrated?

Well, they have their eggs in several growing baskets:

• Gaming & graphics
• Deep learning: training & inference (AI)
• HPC
• Metaverse
• Self-driving
• Robotics
• Datacenter networking

That's a rather broad portfolio. That paves many paths to continued revenue growth. However, I can't tell you whether they'd be a good investment. My hunch is that they're rather over-valued, based on their near-term growth prospects, but nobody should take my investment advice.

 

[bit_user]

That depends on how sustainable the AI gold rush will be. My gut feeling is demand will level out within two years,

It will definitely be a fad, just like the Internet was. Trust me boys: we're going back to sneakernet any day now!
: D

 

[InvalidError]

A couple years ago, I'd have thought Intel could do it, but I haven't heard a lot about Habana and the failures & shortcomings of Xe are known all too well.

I still think Intel can do it, though not necessarily in time to achieve prominence based on the amount of stuff it had to scrap for being too little too late.

 

[JamesJones44]

That depends on how sustainable the AI gold rush will be. My gut feeling is demand will level out within two years, so I expect about two years before AMD, Nvidia and Intel go "Oh crap! We need something else more sustainable to dump all of our excess wafers on. Maybe we shouldn't have golden-showered gamers two years ago."

The fact that Nvidia is severely over bought on an RSI bases and still goes up makes it fell like we are heading deep into a new bubble with the AI craze. 1 to 3 years is a good guess on the pop timeline.

 

[InvalidError]

It will definitely be a fad, just like the Internet was. Trust me boys: we're going back to sneakernet any day now!
: D

I'm not saying that AI will disappear, only that demand scaling will hit a ceiling relatively quickly: how many people and companies can afford to throw 100+TB models at a problem? That is what Nvidia is pitching its new stuff at.

And then you have the whole issue that since AI is just networked of weigh matrices trained by throwing data at it, reinforcing correct answers, deterring incorrect ones until the desired accuracy is achieved, it can be quite difficult to figure out why it randomly gets some trivial stuff wrong.

 

[helper800]

I'm not saying that AI will disappear, only that demand scaling will hit a ceiling relatively quickly: how many people and companies can afford to throw 100+TB models at a problem? That is what Nvidia is pitching its new stuff at.

And then you have the whole issue that since AI is just networked of weigh matrices trained by throwing data at it, reinforcing correct answers, deterring incorrect ones until the desired accuracy is achieved, it can be quite difficult to figure out why it randomly gets some trivial stuff wrong.

Well even google gets 100k queries per second 15% of which are entirely unique. AI getting trivial things wrong, to me, seems completely reasonable. You can basically never train AI enough to be perfect.

 

[Deleted member 2731765]

To put into perspective how much is 1 trillion dollars ? That would be well over 600 million units of RTX 4090 GPUs, or a wardrobe full of NVIDIA CEO’s leather jackets. lol

Claims Vidoecardz, not me.

NVIDIA is now worth 1 trillion dollars - VideoCardz.com

NVIDIA hits a $1T milestone thanks to the popularity of AI NVIDIA has reached an unprecedented milestone by achieving a market capitalization of 1 trillion dollars. This achievement not only solidifies NVIDIA’s position as a powerhouse in the tech industry but also demonstrates the increasing...

videocardz.com

 

[FunSurfer]

Nvidia please don' t forget PC gamers, I bought 12 graphics cards:
Geforce4 4200 Ti 128mb
Two 7800GS
8800 GTS 320mb
8800 GTS 512
8800 GTX
GTX 460 768mb
GTX 570
GTX 760
GTX 970
GTX 1070
RTX 3060 ti
...waitng for the RTX 4070 16gb...

 

[bit_user]

And then you have the whole issue that since AI is just networked of weigh matrices trained by throwing data at it, reinforcing correct answers, deterring incorrect ones until the desired accuracy is achieved, it can be quite difficult to figure out why it randomly gets some trivial stuff wrong.

It only has to be better than a human, and that often turns out not to be too hard. I see things that aren't there or misinterpreted what I saw with enough regularity that I could easily believe a good computer vision model would have better accuracy than me.

As long as you don't overfit your model, accuracy tends to increase at a somewhat predictable rate. The bigger issues are things like bias and hallucination.

[1712.00409] Deep Learning Scaling is Predictable, Empirically

You can basically never train AI enough to be perfect.

It just has to be good enough.

 

[Deleted member 2838871]

Jesnen will be even more under pressure to deliver on growth and profits which means bad news for gamers since our needs are now truly an afterthought.

... and somebody said recently that now is a bad time to build a PC. I disregarded that advice and can now afford to sit back with the 4090 for the next few years without a care in the world while all this drama unfolds.

Don't even need a 5090. The 4090 is that good.

Nvidia please don' t forget PC gamers

Nvidia only cares about profits... which doesn't make them all that different from any other company.

 

[Blessedman]

I am not sure a lot of the commentors here are understanding the changes that AI will continue to make especially given enough breathing room. People aren't seeing the massive jumps we are about to take in so many different directions with AI finally starting to progress it's own path and within two years AI will have changed every aspect of our lives. We have only been capable of iterative jumps in technology where AI has the ability to take gulps instead of sips. Truly a little scared and a little excited.

 

[JamesJones44]

I am not sure a lot of the commentors here are understanding the changes that AI will continue to make especially given enough breathing room. People aren't seeing the massive jumps we are about to take in so many different directions with AI finally starting to progress it's own path and within two years AI will have changed every aspect of our lives. We have only been capable of iterative jumps in technology where AI has the ability to take gulps instead of sips. Truly a little scared and a little excited.

From the bubble perspective, it doesn't mean it's not a bubble just because it has a chance to be impactful. Computers were just as impactful in the 70s and 80s, the internet in the 90s, etc. All bubbles in stock market land that eventually popped.

People right now are treating Nvidia, Microsoft, Google and a few other companies like they are the only games in town, when in fact there are a lot of potential players and how much money this will actually generate for a lot of these companies is questionable. So while it makes some sense to be speculative, endlessly bidding up companies like Nvidia and Microsoft is likely to lose people money when the bubble pops.

It's always different this time, until it's not every time.

 

[JamesJones44]

I was thinking that GPUs would be replaced by a more AI processor in the near future. Do you think GPUs will be the processor of choice going forward? Can Nvidea stay on top? Or is their stock insanely overrated?

There is a lot to unpack here, lets tart with processor of choice going forward. AI centric chips are in development while there are some custom designs that already ship with "neural engines/cores", but most are very specific to a use case. However, Google for example is getting ready to make its own for training Bard and other AI centered services. The question for Nvidia isn't whether GPUs will be the AI choice in the future, it's whether they can they make an AI "chips" that are competitive and stay competitive.

On the stock side, Nvidia is over valued and over bought at the moment. However, like most bubbles, this could last a while. Intel traded at ridiculous evaluations all the way up until the .com bubble burst because their chips where going to be needed everywhere!!! Fast forward to 2001 and it turns out, the market isn't as big as everyone thought and new players and options made it less likely Intel was going to be inside everything. Nvidia is in the same boat now. If you play, be nibble is my advise. Sell rips, buy dips, watch the trend lines and indicators to make the best guess on when the bubble has popped (or just buy the Qs).

 

[derekullo]

When Price/Earnings doesn't matter anymore your market cap can be whatever you want!
https://finance.yahoo.com/quote/AMD a casual 626
https://finance.yahoo.com/quote/NVDA 210
https://finance.yahoo.com/quote/TSM 16