Nvidia ($960B) is now worth more than Facebook, Tesla and Netflix
twitter.com
twitter.com
Will AI be "big" business? Under what business model. And who will benefit? Will Nvidia be the only one selling shovels? And how many will be buying shovels and to what end?
Does Mr Market project that AI is going to get "democratized"? Who will be running AI models and what have they being doing until now without high end GPU's?
Everything is up in the air and that includes the stock price
For example, if the largest models are trained and shared as open source and only inference is done locally, its dominant position will be erode much faster.
On the other hand if only a few supersized entities train large models and serve them through API's they will eventually have their own datacenter chips.
Ultimately the Nvidia booms and busts reflect the lack of serious competition in this space. Once the previous oligopolist made the wrong calls on multi-core CPU designs, the field was left wide open.
The more general challenge for these players is how their offering will land in the post-Moore's law landscape.
These accelerator type chips are hard to program, consume lots of electricity and are good for very specific tasks. The mechanics and economics of the mass market CPU era dont apply.
Grab your pop corn.
Open source models would still need to be trained on Nvidia’s HW?
I agree with your other points though, if the market becomes heavily concentrated I just don’t see the larger companies not developing their own chips/funding Nvidia’s competitors. Data center seems to be fundamentally different from the gaming market in the way that price/performance is the only thing that really matters. Which would imply that it should be easier for other competitors to challenge Nvidia.
yes, but how much hardware would really be required for training in this scenario and does it support a stratospheric valuation? Training models versus using models are dramatically different processes. Training is very intensive in cycles and memory but it is one-off and centralized to the model developer. Inferencing is in contrast relatively less intensive but its use is continuous and decentralized to potentially billions of users. Consider also that we are likely in the brute-force era and model size will keep shrinking, potentially quite dramatically.
While models have a lifecycle and will, in general, need to be retrained with new data, be specialized to private data etc, once a model is released it may be used for years without change. In fact you want this stability and longevity to release things in production.
My guess (but it is only a guess in what is a very uncertain landscape) is that just like mass-produced general purpose CPU's eventually won the economics game and became the dominant chip design (disrupting both the RISC workstations and HPC supercomputer clusters of the time), inferencing chips will win the mass produced "AI" economics game and will become the dominant accelerator, co-processor or super-sized CPU design.
The important commonality between inferencing and training is that both are numerical linear algebra. The economic model that will successfully commercialize inferencing is thus likely to drive the entire architecture for this specialized type of compute - the mass produced inferencing architectures will subsidize developing the higher-end training configurations.
Of-course, even if that reading is accurate, Nvidia could still somehow compete, but it could also be disrupted fairly structurally.
All these “market cap” exercises just take whatever the last trade happened to be, for some relatively small number of shares, and then multiply it by the total number of shares.
This is an okay, but far from perfect, way to value a company.
This is why when a company is purchased in entirety, it’s often at a large premium or discount to the market cap.
Is Nvidia really worth about 1T? Maybe, likely not though.
I think most people would rather have one third of Apple than all of Nvidia.
All of Nvidia was not purchased in Friday, and neither was 1/3 if Apple. We don’t really know what those things are worth, because we do not have such a transaction. To base that on.
All we know is what a bunch of individuals would buy or sell a small number of shares for. Very different.
The reason no one uses ATI is because their dev support is a disaster.
They've made big noise about future GPU and compute products; We'll see if they're actually "competition" in this sector when they actually get released in competitive SKUs.
Though many of those products may end up using TSMC too, but that's an option Intel have likely chosen for capacity/price/performance issues, rather than a hard requirement.
https://twitter.com/petergyang/status/1662831418882560000
>May 28 >9/ Jensen wrapped up the speech with this:
"Run don't walk. Either you're running for food, or you are running from being food:
1. Have the humility to confront failure and ask help. 2. Endure the pain needed to realize your dream. 3. Make sacrifices for your life's work."
The stock market has become a meme market, today nVidia is hot...10 weeks from now it will be something else entirely that people fall in love with.
I predict that all the crypto-boys will have an epiphany "if you can beat them join them" and will pump JPMorgan and Bank of America to the moon, as always completely removed from any fundamentals.
“At 10 times revenues, to give you a 10-year payback, I have to pay you 100% of revenues for 10 straight years in dividends. That assumes I can get that by my shareholders. That assumes I have zero cost of goods sold, which is very hard for a computer company. That assumes zero expenses, which is really hard with 39,000 employees. That assumes I pay no taxes, which is very hard. And that assumes you pay no taxes on your dividends, which is kind of illegal. And that assumes with zero R&D for the next 10 years, I can maintain the current revenue run rate. Now, having done that, would any of you like to buy my stock at $64? Do you realize how ridiculous those basic assumptions are? You don’t need any transparency. You don’t need any footnotes. What were you thinking?” — Scott McNealy, Business Week, 2002
The simple truth is that the people bidding it up indiscriminately have no clue about the fundamentals. You only win if you get out before reality sets in.
This is not some microcap company that can eventually justify such an extreme multiple via actual growth
Nvidia is in a very good position in a rapidly growing market. For folks who don't understand that, it might be comforting to look at recent history for sales numbers, gross margins etc. Those numbers are practically meaningless for Nvidia 2032 financials.
In 2004 "value" guys were looking at Google's IPO and calling bs. 20 years later, sales are up 100x. Market cap is up 60 or 70x. Those value guys were looking at LTM.
King of value guys, Warren Buffett:
"Future profitability of the industry will be determined by current competitive characteristics, not past ones. Many managers have been slow to recognize this. It’s not only generals that prefer to fight the last war. Most business and investment analysis also comes from the rear-view mirror."
So you make an educated guess on what the future earnings will be and work backwards from that to determine whether something is a good investment.
So what is your revenue/earnings projection for NVDA and terminal multiple to justify today’s valuation. You do have one right?
Not a great return from current prices, but this situation is nowhere near detached from fundamentals. The only people making that claim aren't grasping the situation we are in with respect to demand for GPUs from AI and the competitive position Nvidia have managed to get themselves into. The world has changed in the last few months as far as computing is concerned.
Current market cap $1T, projected market cap in 10y $1.5T given 50B net income * 30.
That’s quite terrible given the commensurate risks. You aren’t pricing in at all that CPUs can be used for inference, FAANGs will compete, AMD cards will surely become viable too if the market is growing that quickly. Apple’s chips today can be used for fast LLM inference for large models, and they weren’t even designed with that intention in mind. Competitors didn’t care before because it was a small market.
So how is it logical at all to invest at these prices? Even if you double the revenue projection to $400B, the total return is not very compelling over 10 years, and carries a large amount of downside risk versus alternatives.
I have no doubt that people will make money playing hot potato with it over the next few months, but the stock price is likely to go nowhere over a longer timeframe. Eventually the greater fools run out
The only way to make money on nVidia is shorting, mega-caps hit a ceiling and the value of shares simply stops going up.
It's the classic S-curve phenomenon where the ceiling is 1T-ish
There is effectively no rational fundamental argument for nvidia to go materially higher such that compensates for the risks. Rationality is not what’s at play though
The claim for AMD has been there for years. They can't write software. Cuda and Nvidia chips are the only real game in town and have been for longer than most expected and there is really nothing that looks like it will take over. Custom ASICs were going to take over, until they didn't.
Anyone making these claims isn't close enough to the market. The real fundamentals are in the details, not in hand wavy nonsense.
Investing has nothing to do with whether technology is legitimate, real or cool, and everything to do with the amount of money you can make from that technology. And the numbers show NVDA will be a poor to middling investment in the long run even if the strongest bull case to the fundamentals materializes. Even if you 2x, 3x, 4x the numbers you provided. If you have to use extremely optimistically bullish numbers to get to a 10% CAGR (matching the index), there's a big problem.
Please show me the math otherwise.
Those numbers are my estimates. I don't own the stock. I don't think it's a great buy. They do illustrate that the situation isn't detached from reality. People who own it aren't idiots, they are just a little more optimistic than I am. A little more optimism on margins (which may be warranted) and market size and it starts to look good. Just increase my 3 assumptions 20% each and it goes to $2.5 tril instead of $1.5 tril, ie ~10% return. Doesn't need 2x, 3x or anything of that nature.
https://trends.google.com/trends/explore?geo=US&q=chatgpt&hl...
Week chatgpt: (United States)
11/6/2022 0
11/13/2022 0
11/20/2022 0
11/27/2022 2
12/4/2022 40
12/11/2022 50
Henry Ford first signed off the Model T and then got rich.
I could go on, we used to reward practical improvment in quality of life, not just expectations . The thing about rewarding people for creating expectations is that false dawns can be natural (and they do happen), but they can also be manufactured artificially to get rich.
And also quite frankly Chat-GPT is being developed in the wealthiest area in the world where everybody hangs out together in one giant Fed fueled party. If this was something as revolutionary as the wheel like many commentators said, surely the movement would have preceded the unveil not the other way around.
There is nothing wrong with that logic. NVDA are extremely well placed with their chips and cuda. Demand is high. It doesn't appear to be going anywhere but up. Their competitive position is good. Maybe the price is too high, but it's not anything even approaching a meme stock. There are real fundamentals behind that valuation.
AI may not lead to a promising future, but a blind insistence on viewing everything through the lenses of class and history definitely won't.
If this was something as revolutionary as the wheel like many commentators said, surely the movement would have preceded the unveil not the other way around.
What does that even mean?
We're on the threshold of the mother of all gold rushes, and everybody else's shovels are made out of either rubber or Chinese pot metal.
So, yes, Nvidia is in a very good position right now. About the only things that can derail them are war in Taiwan or the appearance of new techniques that make inference computationally very cheap.
So long as training (computation of the gradient and update) remains expensive, nVidia - and others - will be fine. At the moment, between the global cloud providers and smaller players, there appears to be effectively "unlimited" demand for GPUs. This is likely to continue, as ML gets used for more and more things.
Remember - the GPTs are the new hotness, and certainly hog their share of GPUs right now, but there are a lot of models just quietly working without making anywhere near the splash; all of these need GPUs too.