Nvidia stock price befuddling analysts

MS is currently number 1,
1709849964016.png
 
GotNoRice said:
How much farther can the balloon inflate?
Click to expand...
I think both the hysteria and the hype are overblown. I'm not so sure it's pure bubble, but it's inflated for sure. Just look at how homogeneous "AI" output is already becoming as the algorithms are trained on "AI" output. This shit just barfs out whatever is most likely to get a response, it's like fuckin' Reddit with venture capital.
 
Twisted Kidney said:
algorithms are trained on "AI" output.
Click to expand...
2 third of alpha fold data on its final training set were synthetic data from the previous best version of alphafold, synthetic data is not necessarily the issue we think and which AI are we talking about that is trained on AI output ? They will tend to be similar because many of them will be trained on similar training set, wikipedia, the same built free of rights open internet package and so on. Once they get good at training on audio-video, this will explode the amount of human generated data to train on, from podcast, youtube to surveillance camera feed...

Do wee see an hysteria and hype around the new compute platform.... good luck knowing something like that how one brain goes around evaluating this... maybe like Internet in 1995, it is way too low, maybe like cloud in 2012, mobile in 2008 or social media in 2010 about right.

This could be electricity level big.
 
Last edited:
Zepher said:
MS is currently number 1,
View attachment 640217
Click to expand...
Interesting that four of the five have a fundamental dependence on the others for their businesses to continue to operate.

Anyway, I think there is some irrational optimism that is contributing to this, but Nvidia is also in the extremely rare and enviable position of having 95% marketshare of both the hardware and the software sides of a multitrillion dollar industry that is also exploding with demand at 150% of production capacity and rising. Even if their marketshare fell to 50%, there would still be gigantic growth forecast all the way out to the horizon. This is the start of a gold rush, and Nvidia owns almost every mine. This means huge growth until the mines run out or AMD finds diamond. So, I think another factor is that there is some growth priced in.
 
Although I'm sure Nvidia is a good stock, my friend ended up investing into Shin Etsu Chemical, due to their upstream utility to everything Nvidia and TSMC does.
Doesn't look quite as "fun," but I'm sure there will be no value lost at least.
 
Nvidia has an hard moat to beat, you need

1) As good/better chips
2) As good/better interconnect-network tech to scale up (Mellanox 6.9 billions was a lot in 2019 and not sure how much for Cumulus, now look really good)
3) Software stack

If you do all that, you still need to both buy a good enough node like TSMC and wafer in good enough volume when there is a waiting line, generate good yield over increment and convince customer at a price that is competitive....

There is many thousand of piece that need to go right to build this:
nvidia-h200.jpg


And to connect many of those together, those kind of margin with that kind of market should never work for that long, but that such an high complex affair, a bit like TSMC or people that make lens for them.
 
Last edited:
LukeTbk said:
Nvidia has an hard moat to beat, you need

1) As good/better chips
2) As good/better interconnect-network tech to scale up (Mellanox 6.9 billions was a lot in 2019 and not sure how much for Cumulus, now look really good)
3) Software stack

If you do all that, you still need to both buy a good enough node like TSMC wafer when there is a waiting line, generate good yield and convince customer at a price that is competitive....

There is many thousand of piece that need to go right to build this:
View attachment 640281

And to connect many of those together, those kind of margin with that kind of market should never work for that long, but that such an high complex affair, a bit like TSMC or people that make lens for them.
Click to expand...
Exactly this. They're in a position where the only real chance of them not growing explosively would be some insane supply chain disaster or a non-GPU-based or non-GPU-bound AI discovery that benefitted from being on something like an x86 extension or something similarly Intel-heavy. Other than that potential disruption, even Intel's best chance for growth is as a supplier to Nvidia.

I guess to counter that a bit, with AI being used to optimize AI, I think that might improve the likelihood of that kind of disruption.
 
LukeTbk said:
Nvidia has an hard moat to beat, you need

1) As good/better chips
2) As good/better interconnect-network tech to scale up (Mellanox 6.9 billions was a lot in 2019 and not sure how much for Cumulus, now look really good)
3) Software stack

If you do all that, you still need to both buy a good enough node like TSMC wafer in good enough volume when there is a waiting line, generate good yield over increment and convince customer at a price that is competitive....

There is many thousand of piece that need to go right to build this:
View attachment 640281

And to connect many of those together, those kind of margin with that kind of market should never work for that long, but that such an high complex affair, a bit like TSMC or people that make lens for them.
Click to expand...
“As good as” won’t do, why change up your known functional environment for what is effectively a side grade? There’s too much at risk and too much unknown, AMD and Intel need to be significantly better and not just cheaper to pull customers away.
If AMD or Intel is unable to deliver that within a generation or two then they have the further problem of then trying to pull away entrenched customers.

Nvidia’s moat has some teeth swimming in it.
 
I'm going to break a lot of brains here, but while everyone seems busy fixating on Nvidia's share price and saying it's in a bubble simply because it's risen so fast, they're losing sight of the valuation which, believe it or not, is actually getting MORE reasonable, simply because Nvidia's blown away their projected earnings consistently which has forced the stock price higher. In essence, Nvidia is actually getting CHEAPER when measured by traditional ways of valuing an equity.

https://www.bnnbloomberg.ca/what-bubble-nvidia-profits-are-rising-even-more-than-its-stock-1.2037826

“Some investors have been scared to buy because they think the stock is too expensive, but that’s been a huge mistake,” said James Demmert, chief investment officer at Main Street Research. “Every time it reports, the P/E shrinks because the E ends up being so much stronger than people expect.”
Click to expand...

This is not investment advice and I'm not telling people to buy the stock at these levels, I'm just reporting the facts. I took some profits on a portion of my position a few months ago thinking things were getting a bit long in the tooth and the stock continues to surge. I really don't know what to do with my remaining shares at this point, but based on the valuation metrics, the stock is growing into this high valuation, and it's cheaper than my last sale, so if I liked it back then, I should love it now.

To me, the risks are that people will project future growth that won't materialize because another competitor will show up, and that will likely have to come from a place we don't expect, IE people using ASICs for AI from another company rather than GPUs and CUDA. That's a black box though, we don't know if this will happen, and Nvidia has a ton of pricing power, mindshare, influence, and ability to control their customers with their order book that they can also use a lot of that power to box out a competitor.

That said, is AI in a bubble in general? Don't know, but what I do know is there's an AI arms race going on, and companies like Nvidia are the picks and shovels. Regardless of your opinion on what AI will be now and in the future, companies are buying an insane amount of Nvidia GPUs RIGHT NOW to get it done. That's reality. This is different from the dot com era in the 1990s or the blockchain insanity of a few years ago where companies would add "blockchain" or "dot com" to their name and surge in value simply because they said "blockchain" or "dot com" regardless of whether or not they were actually getting orders. Nvidia, on the other hand, IS ACTUALLY GETTING ORDERS. Their business is actually, measurably, growing. Semiconductors have always been a cyclical business, but as of right now, they're on fire with orders.
 
philb2 said:
Irrational optimism, not tethered to reality. IOW, a bubble in the making.
Click to expand...

It's hard to call it a bubble when Nvidia is generating so much revenue and earnings that their valuation is actually shrinking with every earnings report. Nvidia's revenue went up 205% from October of 2022 to October of 2023. People might be over-projecting where those earnings will ultimately go, but I'm not willing to say it's "just a bubble" when they company's revenue is exploding and their order book is full. Nvidia isn't Dogecoin.
 
Lakados said:
There’s too much at risk and too much unknown, AMD and Intel need to be significantly better and not just cheaper to pull customers away.
Click to expand...
While I agree, there is still the issue of availability. When Nvidia can't supply what customers need, the customers will look to get it from somewhere else. When the options are to build for a minority player or wait 18mos for hardware from the majority player, the choice is usually pretty easy. This is even easier for new entrants who don't have existing platforms to convert.
 
HockeyJon said:
I'm going to break a lot of brains here, but while everyone seems busy fixating on Nvidia's share price and saying it's in a bubble simply because it's risen so fast, they're losing sight of the valuation which, believe it or not, is actually getting MORE reasonable, simply because Nvidia's blown away their projected earnings consistently which has forced the stock price higher. In essence, Nvidia is actually getting CHEAPER when measured by traditional ways of valuing an equity.

https://www.bnnbloomberg.ca/what-bubble-nvidia-profits-are-rising-even-more-than-its-stock-1.2037826



This is not investment advice and I'm not telling people to buy the stock at these levels, I'm just reporting the facts. I took some profits on a portion of my position a few months ago thinking things were getting a bit long in the tooth and the stock continues to surge. I really don't know what to do with my remaining shares at this point, but based on the valuation metrics, the stock is growing into this high valuation, and it's cheaper than my last sale, so if I liked it back then, I should love it now.

To me, the risks are that people will project future growth that won't materialize because another competitor will show up, and that will likely have to come from a place we don't expect, IE people using ASICs for AI from another company rather than GPUs and CUDA. That's a black box though, we don't know if this will happen, and Nvidia has a ton of pricing power, mindshare, influence, and ability to control their customers with their order book that they can also use a lot of that power to box out a competitor.

That said, is AI in a bubble in general? Don't know, but what I do know is there's an AI arms race going on, and companies like Nvidia are the picks and shovels. Regardless of your opinion on what AI will be now and in the future, companies are buying an insane amount of Nvidia GPUs RIGHT NOW to get it done. That's reality. This is different from the dot com era in the 1990s or the blockchain insanity of a few years ago where companies would add "blockchain" or "dot com" to their name and surge in value simply because they said "blockchain" or "dot com" regardless of whether or not they were actually getting orders. Nvidia, on the other hand, IS ACTUALLY GETTING ORDERS. Their business is actually, measurably, growing. Semiconductors have always been a cyclical business, but as of right now, they're on fire with orders.
Click to expand...
Nvidia is doing what investors, especially older investors, are not used to and if they saw this back in the day (i.e. the 1990's timeframe) it would be cause for great concern, but this ain't the '90's and they need to get with the fast moving technological times. If you look at stocks 30 years ago, they didn't move much, now there is a lot more volatility in stock market and while some companies were overvalued from the IPO get go, nvidia is worth the asking price per share I think.
 
Dan_D said:
I struggle to understand Apple's popularity and stock prices so I can sympathize with them.
Click to expand...
Tesla was once valued at more than most car makers - combined. They were also something like 1000x their P/E. It made no sense. Irrational optimism will take a stock a long way. AI is just getting started.

I went to two bank websites today and instead of searching, the search was replaced with chatbots - what do you think is powering that? Maybe not AI/nVidia exactly, but some kind of machine learning/predictive modeling.
LukeTbk said:
Nvidia has an hard moat to beat, you need

1) As good/better chips
2) As good/better interconnect-network tech to scale up (Mellanox 6.9 billions was a lot in 2019 and not sure how much for Cumulus, now look really good)
3) Software stack

If you do all that, you still need to both buy a good enough node like TSMC and wafer in good enough volume when there is a waiting line, generate good yield over increment and convince customer at a price that is competitive....

There is many thousand of piece that need to go right to build this:
View attachment 640281

And to connect many of those together, those kind of margin with that kind of market should never work for that long, but that such an high complex affair, a bit like TSMC or people that make lens for them.
Click to expand...
This is why.
 
Last edited:
Agent_N said:
Nvidia is doing what investors, especially older investors, are not used to and if they saw this back in the day (i.e. the 1990's timeframe) it would be cause for great concern, but this ain't the '90's and they need to get with the fast moving technological times. If you look at stocks 30 years ago, they didn't move much, now there is a lot more volatility in stock market and while some companies were overvalued from the IPO get go, nvidia is worth the asking price per share I think.
Click to expand...

Well back in the day, not everyone had a computer in their hand/pocket, constantly connected to the internet, with multiple services allowing you to trade shares almost instantly for free. You used to have to call some dude on the trading floor and he'd charge you a decent amount to make the trade.

Also, companies are significantly more productive now in terms of cash generation. A few generations ago, things like the number of machines and number of factories you had defined value. That's capital intensive, high costs of entry, a lot invested up front. Now, data is what defines your value. Much less capital intensive to acquire and maintain, also much more lucrative from a profit margin standpoint than, say, a car. The problem is that we anchor ourselves to what used to be "normal" for valuing companies. Being an expert on an earlier version of the world is an expression that comes to mind.
 
HockeyJon said:
Well back in the day, not everyone had a computer in their hand/pocket, constantly connected to the internet, with multiple services allowing you to trade shares almost instantly for free. You used to have to call some dude on the trading floor and he'd charge you a decent amount to make the trade.

Also, companies are significantly more productive now in terms of cash generation. A few generations ago, things like the number of machines and number of factories you had defined value. That's capital intensive, high costs of entry, a lot invested up front. Now, data is what defines your value. Much less capital intensive to acquire and maintain, also much more lucrative from a profit margin standpoint than, say, a car. The problem is that we anchor ourselves to what used to be "normal" for valuing companies. Being an expert on an earlier version of the world is an expression that comes to mind.
Click to expand...
That is why most comments from old time investors are not relevant, the world is changing and they are still clinging to what was. Sawm can be said of some CEO's who are not getting with the times and the company they are leading is dying because they are clinging to old ways and means. I know that sounds like a solid generalization, but there is truth to it.
 
HockeyJon said:
I'm going to break a lot of brains here, but while everyone seems busy fixating on Nvidia's share price and saying it's in a bubble simply because it's risen so fast, they're losing sight of the valuation which, believe it or not, is actually getting MORE reasonable, simply because Nvidia's blown away their projected earnings consistently which has forced the stock price higher. In essence, Nvidia is actually getting CHEAPER when measured by traditional ways of valuing an equity.
Click to expand...
The more you buy, the more you save??


lol
 
Twisted Kidney said:
I think both the hysteria and the hype are overblown. I'm not so sure it's pure bubble, but it's inflated for sure. Just look at how homogeneous "AI" output is already becoming as the algorithms are trained on "AI" output. This shit just barfs out whatever is most likely to get a response, it's like fuckin' Reddit with venture capital.
Click to expand...
Dear twisted kidney, I hope this forum reply finds you well…
 
I'm kicking myself for not jumping on the insanity that is the Nvidia stonks, but I'll make it up when it's time to short. It's getting to the point where companies are trying to figure out what to do with all this AI hardware that nobody has a use for. Even Microsoft might end up putting all these AI features into Windows for free, which doesn't exactly make them money. The question is what are the symptoms that it's time to short Nvidia?
 
I've said it many times over the years and I'm going to say it again.

Jensen is a freaking genius and Nvidia is a well-tuned machine that executes year over year. I admire the man and what he has accomplished.
 
$875 LOL I Should have bought some before COVID @ $45. Insane.
 
philb2 said:
Irrational optimism, not tethered to reality. IOW, a bubble in the making.
Click to expand...
Sadly, You just described global finance since the beginning of time. Also, religions. That's a lot of return on a single statement, well done.
 
Last edited:
philb2 said:
Irrational optimism, not tethered to reality. IOW, a bubble in the making.
Click to expand...
No, not bubble status yet, just that nobody knows how to price AI in yet. In a couple of years bubble might be a thing for Nvidia stock
 
Hulk said:
$875 LOL I Should have bought some before COVID @ $45. Insane.
Click to expand...

For once I made a move at the right time and bought a decent amount right before it split.

Almost sold at like 400 but decided fuck it I'm hanging onto this rocket for a bit more.

It has been good.
 
Agent_N said:
No, not bubble status yet, just that nobody knows how to price AI in yet. In a couple of years bubble might be a thing for Nvidia stock
Click to expand...
Maybe longer than that. If Jensen decided right now to completely stop selling hardware and he kept that going for two years straight, Nvidia would still have over 60% market share in AI.

That is how far ahead Nvidia is in AI.
 
Thunderdolt said:
Maybe longer than that. If Jensen decided right now to completely stop selling hardware and he kept that going for two years straight, Nvidia would still have over 60% market share in AI.

That is how far ahead Nvidia is in AI.
Click to expand...
It may be longer, but I don't know the future or how well received AMD's Instinct lineup will be in the coming years. In a couple of years Nvidia might become the most valued company, beating even Microsoft.
 
Or the profit will be captured at a different place on the stack (like Amazon-facebook-Google-Netflix with Internet bandwith vs broadcom, cisco, Nortel, etc...), the challenge to turn what Nvidia do into a low margin commodity do seem much bigger, but could always happen.
 
Agent_N said:
No, not bubble status yet, just that nobody knows how to price AI in yet. In a couple of years bubble might be a thing for Nvidia stock
Click to expand...
I'm still not sure who expects anyone to make use of this AI technology? Take away accountants jobs? Make cheaper B movies based on Pixar and Disney films? I can see this technology used to make Photoshop better, or video editing better, and help programmers with coding better, but beyond that I can't see this being a thing you can sell. Anything that needs AI hardware will probably run on the GPU, with a dedicated AI accelerator that just uses less power and still doesn't perform as well as the GPU. What everyone is doing is training their AI models until it specializes for a very specific task, which at that point they try to find a buyer. Microsoft will certainly try to integrate AI in Windows through the cloud, but nobody would realistically pay for it. I can see open source projects do it on local hardware with better results. AMD and Intel are going to integrate AI accelerators on every CPU they make, so again who needs to pay for anything AI?
 
DukenukemX said:
I'm still not sure who expects anyone to make use of this AI technology?
Click to expand...
Robots, self driving will be some, seismic data analysis of all sorts has been going on for more than a decade, oncology to detect cancer, anything for which data and result exist (personal assistant will be a common one), try to think about who will not sound much harder than the other way around.

One of the first will be call center that speak to the customer language-accent (before it get integrated into the product themselve)


DukenukemX said:
AMD and Intel are going to integrate AI accelerators on every CPU they make, so again who needs to pay for anything AI?
Click to expand...
Are they selling them for free ? People are paying for those cpu.

The performance gap currently is giant, Intel has put iGPU in almost every cpu they have sold for a very long time, no one asking who need to pay for a discrete GPU, people with large compute need at what are specially good GPU being the obvious answer.
 
Last edited:
LukeTbk said:
Robots, self driving will be some, seismic data analysis of all sorts has been going on for more than a decade, oncology to detect cancer, anything for which data and result exist
Click to expand...
I swear this is the same list I've heard with IBM's ROSS. Someone certainly tried using ChatGPT for lawyering and that didn't go well. Not saying this won't eventually happen, but who's paying?
(personal assistant will be a common one),
Click to expand...
How long have Siri and OK Google been around?
One of the first will be call center that speak to the customer language-accent (before it get integrated into the product themselve)
Click to expand...
You mean like this? He stepped in for NorthRidgeFix since he hasn't posted for a week. Before that he had Morgan Freeman do a repair. Not sure what he's using to change his voice so well but I can't imagine it costing much and probably running on local hardware.


View: https://youtu.be/cXAH-eGkElg?si=cKQRy6G0tWSZEbLb
Are they selling them for free ?

The performance gap currently is giant, we could do, Intel has put iGPU in almost every cpu they have sold for a very long time, who need to pay for a discrete GPU ?
Click to expand...
Doesn't Meteor Lake and Apple's M3 come with AI hardware? I think even AMD has it now with their CPU's. The reason everyone needs massive hardware for AI was to train it. Level1Techs Wendell was able to use an AMD $100 card to do images with Danny Devito, which was nearly a year ago. What people are going to try and do is sell that trained AI data to the highest bidder. If this reduces the cost of robots from Boston Dynamics then sure, but I really doubt it. If this makes Siri and OK Google work better, it'll still be free for users. As far as I know, AI has always been involved in self driving and it isn't getting any better. I've heard of cancer detection from dogs to special cancer smelling machines. The way I see it, AI makes a great hammer and now people are trying to find a nail. When they do find that nail, they'll realize it was a very expensive hammer and the huge investment wasn't worth it.


View: https://youtu.be/t4J_KYp0NGM?si=HItOp6XpP0fqVoN5
 
DukenukemX said:
I swear this is the same list I've heard with IBM's ROSS. Someone certainly tried using ChatGPT for lawyering and that didn't go well. Not saying this won't eventually happen, but who's paying?
Click to expand...
This is a bit of a ridiculous example, like someone tried to make a comic book with Bing Image creator and it was not a good one..., do you think there is many big lawyer house that do not use AI to find precedent and what not ?
DukenukemX said:
How long have Siri and OK Google been around?
Click to expand...
Exactly, this is not a fad I do not think, they will continue to get better and to be around, personal assistant will book plane ticket, get you out of Netflix and other deals just by simple voice command, help you with tax, law situation, anything clerical and will deal with other people-organization agents. You give it all your emails and can ask it question.

DukenukemX said:
I can't imagine it costing much and probably running on local hardware.
Click to expand...
Not sure the nuance you make, yes voice inference could very well run on local hardware (or not or a mix), I mean more like text to speech type voice inference, with the ip location and the first word said by the customer you can respond to them with a voice that sound natural for them. Voice inference is quite easy to run even at large volume, the fact that it will probably not cost much is why it is certain it will be used, the part that generate what the voice need to say, that has little limit on how much it will cost, depends on what you need it to do, but we can expect company saving money versus human support yes.

DukenukemX said:
Doesn't Meteor Lake and Apple's M3 come with AI hardware?
Click to expand...
Yes amd they come with 3d graphic hardware as well, the gap (as of now) in inference power between a 300 watt gpu and a small NPU is quite large, could be software stack issue, but their performance for meteor lake not that impressive:
https://www.tomshardware.com/laptop...-workloads-and-amds-chips-sometimes-beat-them

Would they compete anything close in FP8 inference for CUDA and others as a discrete Lovelace GPU, it would be all over the place and it would not be hard to find info about their performance.

DukenukemX said:
I've heard of cancer detection from dogs to special cancer smelling machines.
Click to expand...
Yes, dogs can be really good at it, AI now can beat team of veteran oncologue at some aspect of cancer detection:
https://www.cnbc.com/2020/01/02/goo...doctors-in-breast-cancer-screening-trial.html
“In an independent study of six radiologists, the AI system outperformed all of the human readers,” claimed the report.

The cheaper it gets to run, the more AI will be used, so at least at first instead of lowering demand it will increase it, as it make way more new usage and application versus what it is saved by the hardware being able to do more.

DukenukemX said:
The way I see it, AI makes a great hammer and now people are trying to find a nail. When they do find that nail, they'll realize it was a very expensive hammer and the huge investment wasn't worth it.
Click to expand...
I feel we will find a long list of example, (cloud was not a good idea for everything people used it, Internet as well), but in some domain like analyzing terabyte of data in the oil and gaz, this has been going on for more than a decade now.
 
Last edited:
You must log in or register to reply here.
Back
Top