The Compound 对 $NVDA 的判断变化

And I know you're going to get to Nvidia but that's trading at what 20 something times forward earnings.

Let's talk Oh, you want to do some Nvidia stuff? Why Why does Nvidia report so late? It's so anticlimatic.

I do think that if to not overthink a bare case, it's just our ability to be less surprised at good news. Just look at Nvidia as a classic example of this.

the stock has traded nowhere for months and months and months. The earning the earnings numbers are astounding, but the market's over it. Like, show me something else. What else you got?

Get uh earnings tomorrow after the close. The valuation's been compressing all year uh for multiple years, actually. And um we have great chart of that. So if there is a bubble, it's not an Nvidia. Now people have said, "Okay, there's no bubble in the valuation, but the actual bubble is in the earnings." And I understand that argument, and I don't I don't laugh at it because I don't want this to age badly.

they're going to report a quarter tomorrow night where the street is expecting them to report 98% earnings growth.

So, and and that's not like they had 5% earnings growth last quarter and they all of a sudden released a new uh product. They've been putting up quarters like this for years already. So, if it is an earnings bubble, it's sure taken a long time.

Do you think it's an earnings bubble? I think there's a higher likelihood that it's a pull forward earnings bubble than it's a everybody was stupid for buying all this stuff and it's just going to evaporate. So I a pull forward earnings bubble to me is like everyone's worried about compute. Everyone's worried about are they going to get their slice of compute? Are they going to be able to deliver the services that they're contracting with their customers? And so they're stockpiling as m as many chips as they can. they are building data centers. You can't build a data center with no chips in it. So, you have to order as much as you can to make sure that you'll have access to the supply. And that's more pull forwardy to me than it is like this massive error of like buying things that no one's going to need.

That's where the if there's if there's an earnings bubble, it's it's the most innocent type of earnings bubble because the demand really is there. Okay. So you you know a lot more about this stuff than I do. If let's just say that the that there's a demand pull forward. Isn't the CUDA operating system a major part of the story that even if the demand for the chips fall there's still demand to operate these things on their platform? Maybe is it sort of similar to Crash Strike or am I making a really stupid comparison?

It's not a stupid comparison. It's just it's it's that's like um orthogonal to the main point. The main point is if you build a steel mill, which is the way people think about like in heavy industry, if you build a steel mill, it's very likely you'll still be using a lot of the same equipment 20 years later. If you build an oil refinery, which we haven't done since the 1970s, it's very likely that most of what you've built is still in operation, and you're making repairs and substituting parts here and there. And that's just like the regular maintenance cost of running a heavy industrial site. Data centers are different. Do you know it's an it's estimated 50% of the cost of a data center is the chips. Do you know that? You know how astounding that is? And the thing is these chips may have a useful life of 5 years, but it's unlikely. These chips may still be in use in 10 years, but given the speed of the advance of what this technology can do, it's highly unlikely. And so Nvidia's portion of this data center center spend is such that even if there's a slowdown in new data centers being built, the amount of maintenance and replacement of the GPUs that are being installed today is so astounding that I think Nvidia is probably in the safest position of all of these data supplier stocks. you you you are not likely to see a situation where they're using 2024 and 2025 era GPUs in 2029. You're very unlikely. So even if the data center construction is cut in half because people get spooked, you're still going to need chips to supply what you've already built. What are you going to do? Not buy the replacements. And all this inferencing burns out the chips um not as fast as training but pretty fast. And they're talking about this explosion in in inference because of all the new things we're doing with AI. The you're going to have to replace GPUs and Nvidia will be supplying these data centers for as far as the eye could see. Not for years, for decades. For decades.

So why do you think why do you think it's trading at 16 times forward earnings? Is it the most obvious long hiding in plain sight? Marll is coming on very strong. Broadcom is coming on very strong. He's comp Let's take Marll. Marll is building custom AS6 for customer. Amazon's their biggest customer. So Amazon's building tranium chips which are application specific integrated circuits. They do not have as broad of a use as a GPU. They're not as powerful. They're not clustered the same way. you know, you take a take a take a cluster of 10,000 chips and and put them in. It's not the way that's being done. But Amazon its tranium line of chips, Alphabet's making chips, Microsoft is making chips, Meta, Apple is making its own chips. They're they're utilizing application specific integrated circuits. And that custom chip making for the hyperscalers is coming out of Marll. This doesn't mean Amazon's not buying GPUs. doesn't mean uh Gemini is running without GPUs, but it is new competition for the buildout of compute. Why would they want to use AS6? Well, if they build them themselves, they can customize them for exactly the uses within their data centers that they see as being important. Okay, so that's what Marll is doing. Broadcom is is in there as well. Obviously, AMD is always nipping at somebody's heels. It's It's not going to be an 86% market share story for Nvidia forever. And so I think competition from 3,000 different players, including their own customers, is part of why we're seeing that multiple shrink. I also think there's fatigue. There's boredom. This is already 7% of the S&P. How much more can investors buy? Like Like if you're not an index, how do you want to be 9% Nvidia? you want to be 10. So there's some element of that. It's a $5 trillion company. We've never seen a company get to five trillion. We don't one day we'll have 50 companies at five trillion and somebody will look back and do a study and they'll say this is the threshold beyond which stocks can no longer trade at 20 times earnings.

So that's my answer. Chartkin and I made a chart uh earlier in the year showed the reason why Nvidia is trading at uh below market multiple it has a size problem. If it were trading at 18 times forward, 20 or or 25, what 30, whatever it was, whatever the best growth stocks in the market have historically traded at, it would just be too big. And the market can't digest a stock that of that size.

Last thing, I don't give a [ __ ] what anyone thinks. These stocks are going to be cyclical in the end. And this derating of the multiple over the last three years I believe is an acknowledgement on the part of the investing public that 99% earnings growth quarters are awesome. Their products are awesome. Their competitive position is awesome etc etc etc but these are cyclical companies. They are selling a semiconductors are cyclical.

I maybe I'll end up wrong. Maybe this will go on for 20 years uninterrupted. But the history um if you read chip war uh chip wars or if you have been on Wall Street for 20 30 years, you know that what goes up must come down in this space.

Funnily enough, funnily enough, is it funnily? Either way, laughably enough, ironically enough, one of the reasons software stocks have historically had such a higher multiple than semiconductors up until two years ago when the semis just went wild is that investors know that software is less cyclical than hardware. And investors know that historically software can outgrow a cycle in a way that semiconductors never have. Th that may prove to be not true this time, but historically software company growth software companies have gotten higher multiples than growth semiconductor companies for precisely that reason.

And a lot of people either aren't doing this long enough to have learned that or forgot about it. But I do think there's a gravity that is pushing down on the multiples of Marll, of Nvidia, of Broadcom. They understand that these companies are going to go into a phase where they rip each other's throats out for the next upgrade cycle and not everybody can win and that is why they're not selling at 30 or 40 times earnings the way that they were three years ago.

But I do think there's a gravity that is pushing down on the multiples of Marll, of Nvidia, of Broadcom. They understand that these companies are going to go into a phase where they rip each other's throats out for the next upgrade cycle and not everybody can win and that is why they're not selling at 30 or 40 times earnings the way that they were three years ago. Anything to say about Nvidia tomorrow?

I'll give you the the story lines. These are the things that the people that care about the stock and care about the AI trade are watching the most closely. Let's start by saying the Wall Street consensus price target for Nvidia is now 305.

Let's start by saying the Wall Street consensus price target for Nvidia is now 305. Where is it right now? 212. That is almost 50% higher than where the stock is.

So, you're not going to get bailed out by a by upgrade uh price target lifts. You could get a lot of reiterations if Jensen kicks ass on the conference call, but like what are they going to do? Take the target to 320, 330. So, that's number one.

You could get a lot of reiterations if Jensen kicks ass on the conference call, but like what are they going to do? Take the target to 320, 330. Um the here uh that's its own story. Here are the seven story lines. Number one, the revenue bar is staggering. Consensus is $95 billion in quarterly revenue. Data center is going to be 85 to 87 billion or almost all of that. For context, Nvidia did 26 billion for this quarter a year ago.

Number one, the revenue bar is staggering. Consensus is $95 billion in quarterly revenue. Data center is going to be 85 to 87 billion or almost all of that. For context, Nvidia did 26 billion for this quarter a year ago. Wow.

Jensen to your point. That's that's in that's over. Like investors are no longer a warning of that. We know, right? So Jensen's guidance is 91. So the the 93 to 95 is the street.

Um storyline two, the Blackwell ramp and chip yields. Um the Blackwell is the new central product. If you call up Nvidia, I want the newest hottest [ __ ] That's what it is. Um investors want to hear about production scaling. That's obviously Taiwan Semi is is who's making these things on the three nanometer chips. um they want to hear about supply constraints. They want to yield is like how many chips go bad in the manufacturing process like how many do we get out of each turn? Um so they're listening ve very granularly for any sign of a slip up in execution.

storyline two, the Blackwell ramp and chip yields. Um the Blackwell is the new central product. If you call up Nvidia, I want the newest hottest [ __ ] That's what it is. Um investors want to hear about production scaling. That's obviously Taiwan Semi is is who's making these things on the three nanometer chips. um they want to hear about supply constraints. They want to yield is like how many chips go bad in the manufacturing process like how many do we get out of each turn? Um so they're listening ve very granularly for any sign of a slip up in execution.

What about customer customers broadening out? Last last quarter they broke out the number of the hyperscaler revenue right

in May for the first time they categorized what they sell to the The Vera Rubin architecture that's that's what's coming. That's going to be the next next thing. Um, Jensen has already said he has a trillion dollars in revenue visibility through calendar 2027 for the Vera Rubin upgrade. So, let's take him at his word.

Three is the launch timeline for V Rubin. Ver Rubin. Um, so a delay would be very problematic for the stock because a delay would force people to change their near-term quarter uh outlooks. So we want to hear that that's on schedule.

Four, China export restrictions. So during the course of this quarter, they did get approval and allegedly started selling in China. It's I don't think that's in the numbers, but again, it's a story line that people are listening to the commentary on.

Five is the hyperscaler custom silicon threat which you and I just spent 10 minutes on. I won't go terribly further. Google's TPUs, Amazon's Tranium, uh, which again that's Marll, Microsoft and Meta are building custom accelerators or XPUs. People are going to want to hear Jensen answer the question about that competition.

Six is gross margin trajectory. Um in they they margins this pressure would be the wrong way to say it but margins had come in a bit as they were developing new products. They want to see 75% plus again. And so when we talk about yields yields will be a part of that.

Number seven I don't know if this comes up on the call or if analysts are too uh or are too prissy to ask. 28 insider sells in the leadup to earnings. This could be part of why the stock is uh down 12 straight days. Uh uh I don't know, but I think the insider selling has to stop at a certain threshold before we get to the earnings. So maybe not. Um it's not a huge dollar amount relative to the market cap, but it's also not like the greatest vote of confidence at the levels that that those insider sales um I made this point with Sean today on T.

Um the stock has not been been treated well after earnings recently. This chart. So, that's yeah, that's that's really I think that's a really key point. Tell people what they're looking at here.

So, the last four times it reported and the fiscal year is a little bit weird. That's why you see 27 in here. Uh the market has not liked it. Josh, you and I were on What are your thoughts? I believe during one of these or I can't remember what show it was, but uh stock got smoked.

Yes, this is a terrible post earnings reaction stock. That could change and it hasn't always been that way as you can see on this chart, but it really you have to go back a while to find a quarter that people were super enthusiastic about the next day. All you're looking at is next day returns on this chart. You know what the market so I don't know what he could say for the stock to go up 8%. What hasn't he already said? But sometimes investors just change their mind.

Okay. Well, SpaceX did SpaceX did say that they're all in on Nvidia. If I were in the investor relations suite at Nvidia, I would what I would be telling them is listen, if you don't care about the stock price reaction, no problem.

Talk about space. You don't have to be specific. talk about robots because if there's a new leg to the Nvidia story, not that it needs one with 99% earnings growth, but if there's going to be an expansion of the TAM or a change the subject from ASIC competition, it's going to be automation and self-driving cars and rockets and data centers in orbit and robots. Let's talk about robots because that could be a whole new TAM that's not in the stock or not meaningfully in the multiple.

So, I would just tell people who are in the name going into this quarter, of course, anything can happen and Michael and I don't know, but over the last 26 quarters that they've reported, so that's back to Q1 2020. So half of those quarters in the AI era, 12 times the stock has moved up or down by 5%. That's remarkable, which means it almost never does a thing. If you look at the average next day return, it's plus 2% after an earnings call over the last 6 years. Um, but there were two massive reactions skewing that. In May of 2023, the stock exploded 24%. Yeah. and then in February of 24 plus 16 if you pull those out the stock is typically flat on average the next day.

So here's the t let me set the table for you. 92 billion in revenue is the midpoint of that guidance. That's 97% year-over-year growth. Earnings $29. That would be 99% growth. EBIT 61 billion. That would be 102% year-over-year.

And the final point to make here, and I think this is applicable for everybody listening and watching for as long as they live, just because you identify the stock with the most insane growth rate, that does not guarantee you a reaction in the stock to earnings or other news that's going to make you money. If it were that simple, we would just automatically buy Nvidia, let them report 102% growth in EBIT and sail off into the sunset with unlimited wealth. That is not how things have gone for shareholders in this stock for a very long time. So knowing the numbers, fetishizing the growth rates, these are interesting things. They do not guarantee you upside in the share price because the market is smart. It is way ahead of this. Back to the Dudley conversation. When we talk about anticipatory, nobody gives a [ __ ] that Nvidia grew earnings by 99% over the last 90 days. They're worried about 27 and 28. And that's what the stock is trading on.