Imagine a $3-trillion company was divided into three trillion-dollar daughter companies that had to compete against each other. Parent-company shareholders would start with equal share values in each daughter company. Who would suffer? Who would win?
But when I look around these years, I suspect my own sanity. Thought of the day (in Mastodon format):
Pied Piper of greedy wannabe super-rats? How else can you explain the orange puppet's fanatics?
ChatGPT gets 600 million visits per month, with 77,2 million monthly active users in the US alone, and about 100 million global active weekly users. Just ChatGPT.
I know The Kids don't know how Slashdot works, but if you want to see the results, you don't need a fake survey option, you just click on the link. CowboyNeal or GTFO.
Undervalued. Unless LLMs go away and are replaced with something with better design.
They were lucky to be at the forefront during the past 15+ years. First it was gaming, then mining, then gaming, now AI. AMD is trying, and their stuff is pretty awesome, compared to where they were in the past, but it's very difficult for them to topple the giant, unless, of course, the giant topples itself.
Been a LONG time since I've felt any need for a faster CPU. I have been mostly power/battery- and network-connectivity-constrained for many years now... I would point to the moribund poll about NVIDIA as evidence of concurrence with my position, but mostly I see the dead poll as evidence of how moribund Slashdot has become these years...
Since returning to Slashdot I have mostly been discouraged to see how "the Slashdot situation" has continued to decline. Only moderately active story was the TikTok thing. I
I think NVidia should be valued, but not as much as they are now. They are very good at using AI for example for optimizing rendering, but they are not AI researchers that make big discoveries. Google is the company that does the big discoveries, and that is mostly because of Deepmind. In 2030, people won't be talking much about LLM anymore.
When there is a gold rush, the people who make money are the people who sells shovels. NVIDIA is making and selling shovels for the AI gold rush. They still have the best ML accelerators cards backed by CUDA. I wish I had an option to use AMD or INTEL but NVIDIA has the widest support and easiest to get running.
I never thought Apple would get so big by making cellphone and ipod but here we are. Apple made 90B last quarter and NVDIA made 26B. Yet they have roughly the same market cap. But I guess NVDIA can
Better design is assured, the question is how well NVidia can stay ahead of the curve vs. competitors. That said, CUDA is their moat for now.
A couple known potential disruptors:
* Ternary transformers: Often called "one bit" transformers, though actually a trit, multiple studies have now shown that they can perform nearly as well as floating point, so hardware implementations could be dramatically faster and more efficient. But this involves ternary hardware, which hasn't been seriously attempted s
We've just discovered that grokking works on real world data sets. that just increased the demand for compute power by a factor of 100x. https://youtu.be/QgOeWbW0jeA?s... [youtu.be]
We will TILL the land in semiconductors and we will plow electricity into it and we will cover the surface in chips. NVDA is clearly undervalued, we can't even pave the roads with how few H100s they sell.
NVida's PE ratio of 71% is crazy high. It's EPS is all over the board the past couple of years. As a comparison, Amazon's PE ratio is 50, Apple's is 30, and Google's PE ratio is 19. That bubble's gonna pop!
You must not remember when Google and Amazon and Tesla had PE ratios in the triple and quadruple digits.
It is not uncommon at all for a fast growing company with a district advantage and a huge market to have a PE ratio much higher than this.
In my opinion, the bottleneck is going to be how fast they can build the processors because they donâ(TM)t have their own fabs. If they had their own fabs, I would value them much higher than I do now.
They are undervalued unless something crazy unexpected hap
You must know something nobody else does. I wonder how would one use a Large Language Model (LLM) to develop a chip? I am sure some model can be use but I doubt it would be an LLM.
I think today it's way overvalued because we're in this weird space in between technologies. I know people are convinced LLMs will save the universe, or at least enough people to keep NVIDIA ticking along for quite a bit here, but at some point we'll figure out a better way to tackle the AI "problem." Most likely with specialty chips not designed by NVIDIA, since they're convinced they just need to keep doing what they do, just slightly faster and with slightly better memory specs.
My desktop at home has a 5-ish years old NVIDIA card that wasn't even then top of the line and still does the job (I do some Windows gaming and use Linux for freelance work - web, graphics, photography). My desktop at work has integrated Intel graphics and does the job (I work in IT, mostly office stuff). From my point of view, NVIDIA is way overpriced.
At this point Nvidia's main business is machine learning and scientific computing, and videogames are just the cherry on top. Which is ironic, considering that it started out the exact opposite.
ML is not just a bunch of MACs which pretty much anyone can do. It is also a software that enables you to write your model and run it on these MACs. NVIDIA has CUDA which is a great moat. Google has their own TPUs but you can only rent them and they are crazy expensive to rent. If you want to train a model on your own machine or at your own location, NVIDIA is pretty much the only option.
Parkinson's Law: Work expands to fill the time alloted it.
The bubble of all bubbles (Score:2)
No reason for any company to be that big (Score:1)
Imagine a $3-trillion company was divided into three trillion-dollar daughter companies that had to compete against each other. Parent-company shareholders would start with equal share values in each daughter company. Who would suffer? Who would win?
But when I look around these years, I suspect my own sanity. Thought of the day (in Mastodon format):
Pied Piper of greedy wannabe super-rats? How else can you explain the orange puppet's fanatics?
#PiedPiperOfStuporRats [sic]
Maybe the reason I'm so confused is be
Re: (Score:2)
So basically you're suggesting that should Nvidia goes bankrupt, it will be worth less than an NFT or collectable doll?
Re: (Score:2)
> Even bigger than bitcoin.
Hey, when bitcoin busted, at least you had a useable 3090 with 16GB...
Re: (Score:2)
ChatGPT gets 600 million visits per month, with 77,2 million monthly active users in the US alone, and about 100 million global active weekly users. Just ChatGPT.
Re: The bubble of all bubbles (Score:1)
Headline numbers donâ(TM)t tell us muchâ¦.
If itâ(TM)s questions like: list all the words that mean flatulance without the letter f in them?
Iâ(TM)m not sure that is really relevant to productive use.
This isn't Twitter (Score:5, Funny)
I know The Kids don't know how Slashdot works, but if you want to see the results, you don't need a fake survey option, you just click on the link. CowboyNeal or GTFO.
Re: (Score:3)
Seriously, who's writing these polls? You'd think the editors know how Slashdot works.
You must be new here...
Re: (Score:2)
CowboyNeal hasn't been seen around here for years...
Re: (Score:2)
Pffff... noobs.
Unpopular opinion, again (Score:1)
Undervalued.
Unless LLMs go away and are replaced with something with better design.
They were lucky to be at the forefront during the past 15+ years. First it was gaming, then mining, then gaming, now AI.
AMD is trying, and their stuff is pretty awesome, compared to where they were in the past, but it's very difficult for them to topple the giant, unless, of course, the giant topples itself.
Re: (Score:2)
Based on my time at AMD I am always surprised when I hear the company continues to exist. Why is it still around?
(But I worked in a LTHV (Low-Tech High Volume) chips department. Mostly tracing yield problems...)
Re: Unpopular opinion, again (Score:2)
AMD now makes the fastest PC processors and they are still cheaper than Intel, That's why.
Whither Slashdot wander? (Score:2)
Been a LONG time since I've felt any need for a faster CPU. I have been mostly power/battery- and network-connectivity-constrained for many years now... I would point to the moribund poll about NVIDIA as evidence of concurrence with my position, but mostly I see the dead poll as evidence of how moribund Slashdot has become these years...
Since returning to Slashdot I have mostly been discouraged to see how "the Slashdot situation" has continued to decline. Only moderately active story was the TikTok thing. I
Re: Whither Slashdot wander? (Score:2)
I agree with your apparent assessment that slashdot is now a shit show.
This is what happens when cryptocucks take things over. Without exception, they always enshittify.
Re: (Score:2)
Some companies, like some people, are just born to be mediocrities.
Re: (Score:2)
I think NVidia should be valued, but not as much as they are now. They are very good at using AI for example for optimizing rendering, but they are not AI researchers that make big discoveries. Google is the company that does the big discoveries, and that is mostly because of Deepmind. In 2030, people won't be talking much about LLM anymore.
Re: (Score:2)
I never thought Apple would get so big by making cellphone and ipod but here we are. Apple made 90B last quarter and NVDIA made 26B. Yet they have roughly the same market cap. But I guess NVDIA can
Re: (Score:2)
Better design is assured, the question is how well NVidia can stay ahead of the curve vs. competitors. That said, CUDA is their moat for now.
A couple known potential disruptors:
* Ternary transformers: Often called "one bit" transformers, though actually a trit, multiple studies have now shown that they can perform nearly as well as floating point, so hardware implementations could be dramatically faster and more efficient. But this involves ternary hardware, which hasn't been seriously attempted s
Re: (Score:1)
We've just discovered that grokking works on real world data sets. that just increased the demand for compute power by a factor of 100x.
https://youtu.be/QgOeWbW0jeA?s... [youtu.be]
CowboyNeal's Nest Egg (Score:3)
Not enough GPUs for the Dyson Sphere (Score:1)
We will TILL the land in semiconductors and we will plow electricity into it and we will cover the surface in chips. NVDA is clearly undervalued, we can't even pave the roads with how few H100s they sell.
What I really wonder... (Score:2)
What is the aggregate value of every voter's investment portfolio...?
PE ratio of 71%!?!?!? (Score:2)
Re: PE ratio of 71%!?!?!? (Score:2)
Re: (Score:2)
Missing answer: Stupidly overvalued (Score:2)
Amateurs just keep buying shares in it without checking if the current value already reflect what they have just been told months late
Re: Missing answer: Stupidly overvalued (Score:2)
Re: (Score:2)
How to tell me you have bought shares recently, without telling me you have bought shares recently :D
Today: Way overvalued. (Score:2)
I think today it's way overvalued because we're in this weird space in between technologies. I know people are convinced LLMs will save the universe, or at least enough people to keep NVIDIA ticking along for quite a bit here, but at some point we'll figure out a better way to tackle the AI "problem." Most likely with specialty chips not designed by NVIDIA, since they're convinced they just need to keep doing what they do, just slightly faster and with slightly better memory specs.
So, ultimately, it'll fall
Overvalued (Score:2)
My desktop at home has a 5-ish years old NVIDIA card that wasn't even then top of the line and still does the job (I do some Windows gaming and use Linux for freelance work - web, graphics, photography). My desktop at work has integrated Intel graphics and does the job (I work in IT, mostly office stuff). From my point of view, NVIDIA is way overpriced.
Re: (Score:2)
Re: (Score:2)