Hi, this is Steve Eisman. Welcome to another episode of The Real Eisman Playbook. I'm going to make this intro short because there's just so much to discuss. We have back two recurring guests, Gil Luria of Davidson and Dan Ives, whose new company is called Yorkville.
Yorkville. And before we get into a discussion, I'm going to brag for just one second because I made one of the greatest predictions of my entire life with you guys when you guys were here at the end of July. The very end of the interview when I was saying thank you, I then said I will have you back in three to six months because between now and then the world will have turned over at least 15 times. [laughter] And you know what? That's an underestimate.
So just so everybody knows, today is November 4th. It's the day after the election. The Democrats have taken both houses of Congress. We'll get to that.
Since what I want to do is, from both of your perspectives, the AI narrative has migrated. I mean, even people who, like such as mostly yourselves, think it's very positive, I think everybody would agree that narrative is a lot more complicated now. So let's just do something very dispassionate as opposed to taking a position on it, which we'll come to. Dan, since last we met in the summer, how has the AI narrative in your view changed?
I think it's changed. On one hand, the use cases and the actual — I think not just the visualizations — become more of a reality for companies, that they're actually going down the use case path in terms of ROI. That's become much more front and center. We've seen it on Palantir and so many other tech companies. On the negative, I think it's the view that the amount of capital raises and the amount of money being spent is something that essentially more or less has drowned the market. I think the market screamed the last few weeks on the tech side: no mas, enough's enough. And we're at this sort of tug-of-war point where the AI revolution's accelerating. We see that in Korea a few weeks ago in terms of chips, Taiwan. We see it on the use case side in terms of enterprise. But then the spending — I think investors have said we have to slow this down, not realizing that it was going to take a blue tsunami to actually slow it down, which is why I think tech stocks are up.
Explain that last sentence because I don't quite understand what you just said.
Because I think — and I'm so interested to hear what Gil says — the last few weeks, no matter what tech companies have done, stocks sell off because you have ATMs, capital raises, bond offerings, converts. They're going to Malaysia to raise money, wherever it may be. And I think essentially what's happened is, and I've seen that from an institutional perspective, investors are like, look, I feel like at this level of capital raising ATMs, these stocks almost can't go higher. Finally now, politically, because of what we're seeing in Texas, because of the blue wave, now they're going to have to slow down. CAPEX 27 is probably a peak year for CAPEX. It would start — you probably won't see that till 2029. And for the hyperscalers, they're stronger, stronger because they're the ones that ultimately — neoclouds on the other side of this. But that's why I think hyperscalers are so much higher today.
Every single AI company that's reported, and every company that's reported about their AI business over the last three months, has reported better than expected results, increased their outlook, and increased their visibility. It's actually the non-AI stuff that looks weaker.
PC sales are down. That hurts Microsoft. It hurts Intel. Search is down. That hurts Google. Retail's down. That hurts Amazon. It hurts Shopify. The AI results from all these companies, from the labs, from the hyperscalers, from the chip companies, have all been better than expected. So that's the good thing that's happened. The not good thing that's happened is that it became increasingly political over the last few months. Leading into these elections, it became a slam dunk issue. It became the slam dunk issue. It was a big part of the blue wave. Data centers are bad. It was a big part of the blue wave. Even though it was a bipartisan issue, there were Republicans running against data centers, but it was part of the blue wave. And now we're going to have to deal with this reality that it's going to be harder to build a data center. It's not that we're going to stop, but it's going to be harder to build a data center. There's going to be a lot of downstream impacts from that, but that's an outcome of the elections — it's going to be harder and more expensive to build data centers in the US.
Yes.
Okay. So let me press the point. Forget about Nvidia just for a second. Take like equipment companies like Arista, Cisco, Dell. Those stocks are up today, but the Vertivs of the world, the Quanta's of the world are all down. How do you — does that make sense to you?
It does because Dell, Cisco, they're not building data centers. They're the infrastructure, the install base. The existing data centers, they're going to have more opportunity than ever. But with incrementally, let's call it 10 to 15% of data centers now not being built out — call it 800 to 1,000, let's say 15% of those don't happen — that's negative for Vertiv. It's negative for...
Isn't that also like negative for Arista? Means less equipment sales?
But I would say right now is that the worry is that there was almost going to be an over-building out, an over-capacity. And I think that's not going to be the issue. If you look at the next few years, that's why Dell — I would say probably the next 18 to 24 months, they already see what their sort of prospects are. If you go further out, that's one thing, but I think that's why you're seeing like a delineation that's happened in this market: neoclouds, power, hyperscalers, Apple — that's almost seeing like the kingmaker in terms of where they sit.
So if AI is not taking the oxygen out of the room, that's good for the rest of tech, for non-AI tech, which includes networking equipment, etc. It's good for the non-AI tech because AI has been taking the oxygen out of the room. Enterprises are spending so much on AI, they don't have money for anything else. If they're going to spend less on AI, that's more money for other things. Dan made a really important point though that the hyperscalers win anyway. The hyperscalers win if there's more AI, the hyperscalers win if there's less AI. They win anyway, and you're seeing that today.
Why do you think that's the case?
Because if there's more AI, the revenue continues to accelerate. If there's less AI, their cash flow accelerates.
And so they win either way.
And I'd also add part of the issue with Oracle — Oracle is that we want to be a hyperscaler. SpaceX, we want to be a hyperscaler in Texas. Microsoft, Alphabet, Amazon, they already are hyperscalers. And if they get voted down in like that, they could go to Alabama or anywhere else. Part of the issue — that's the problem with the Oracle and the SpaceX that you're seeing play out in the market real time.
And the hyperscalers have the installed base of data centers. That just became more valuable because we can't build other data centers.
That's exactly it. It's like a collectible. Like why does the value of collectibles continue to go up?
Let's move on. And I got like 3,000 things I want to talk about.
Let's go to Google. So the interpretation, rightly or wrongly, of Google's numbers — and I'm not talking about the cloud part of the business, just talking about the search part of the business — was that Google search revenue growth was 12%, it was 19% in the first quarter. And the fear around Google for the last two years has been that Google has this incredible business called search which has insurmountable moats. I mean, I've never seen a business — no one's ever seen a business like Google search to such a degree it's a paradigm. You couldn't even think outside of the concept of Google search. But now you have alternatives, and those alternatives — we could argue about how high those moats are, but those moats are certainly lower than Google search. So you're exchanging a business where you're basically a monopoly for something where you're competing. It's very capital intensive, and the evidence is now in. Google's — this is the thesis: Google search growth is slowing because it's moving over to AI, and that's bad. That was without question the interpretation of the quarter. I'd be curious as to how you guys actually see it. And again, I'm not talking about the cloud business — was great, the AI stuff was great — but the crown jewel of this company without question has always been search. Gil, take it away.
This is the quarter we feared. To your point, we've been afraid of this quarter for the last couple of years. Ever since we started using ChatGPT more than Google, we were afraid this quarter would come. But they reported quarter after quarter after quarter of good results, and so investors became complacent. But then all of a sudden, out of the blue, abruptly — to your point — they grew 12%.
Out of the blue this happens without them warning us, without them guiding to lower growth, it happens. So it's inconclusive because, by the way, 12% growth is still fine, but it indicates that there may be something broken. Let's not forget, if this breaks, it's going to break in slow motion.
Because it's ad dollars that we're talking about. It's not eyeballs.
Right. Consumers don't pay to search Google, right? So if we're searching less Google, that doesn't matter until the advertisers say, you know what? Instead of spending 90% of my search dollars on Google, I now have another choice.
And by the way, OpenAI has ramped up their advertising very cautiously, very slowly. It's maybe a billion dollar business right now, which is a very small part of the market. But at the point where they start advertising as much as Google, when their ad load catches up to Google, all of a sudden all the advertisers will switch 10, 20% of their budget to ChatGPT, to OpenAI. And so that's not going to happen overnight. That's slow motion. But it feels like this happened abruptly. So first of all, for the first time, Google admitted that they don't generate as much ad dollars on an AI search than they do on a Google search. They'd been denying it up until now.
In this quarter they admitted that that's the case. And that's again, that's one piece. The other piece is there's a lot less searches. More than half of web traffic comes from chatbots now, right? It's not humans searching anymore, it's chatbots. Are advertisers paying when a chatbot clicks through the site? These are the things you have to worry about. At what point do advertisers go to Google and say, I don't want to pay to have chatbots — they're not actually looking at my site. They're looking at your site. I don't want to pay cost per thousand based on a chatbot. I'm going to pay a lot less. This is still stuff that could still happen.
So if I'm going to interpret what you just said, you walked away from that quarter more negative on Google than before.
Yeah, I think there's no way that you could walk away from that quarter feeling incrementally positive. Now, I think the question will be, to Gil's point, is this just an air pocket period, as painful as it is, when Google figures out the monetization engine on Gemini? They ultimately — when you think about the frontier models, what's happened to OpenAI as well as with Claude — do they start to sort of ring-fence things where they feel like, okay, we now have figured out the way that we're going to monetize search and we're going to monetize our ecosystem, that Gemini needs to basically narrow the gap with the likes of an OpenAI and Anthropic? That's essentially what needs to happen because there's no different than another way what's happened with Palantir and C3 AI. Frontier models are here. So many other companies are here that like, should we put a wall up that then this becomes not just our data, but it's our ecosystem to monetize? That's what essentially Alphabet's going through. My view is that the market's baking in a very negative scenario for the advertising. If they just show even 30, 40% success rates, I think it's a stock that could go significantly higher. That's sort of my view along with the cloud business, but no doubt, no doubt this was the quarter...
This was the quarter that people were worried about.
A lot of times it was like bears have called 10 the last two downturns, whatever. Then finally like this was the one. They come out of hibernation mode and they're like, told you. [laughter] And you can't deny. So now they have to prove it otherwise.
Okay. Let's move on because like I said, we got so much to talk about. Imagine I'm an Oracle shareholder. Okay.
You don't sleep well, do you?
I am upset [laughter] because I have lost so much money in this stock over the last year or so. And I come to you guys and I say, guys, help me out. A, explain to me what the hell just happened to my stock, and B, should I still own it? I'll take you first.
That's the look, that's the one that like we're most cautious on because my view is like I know so many people at Oracle over the years — the gold standard. I mean, if I look at like software tech, I could go through every company and like a handful of people started at one point, you know, at Oracle. Many of those people have either been cut or left because of all the layoffs that they've done. They've cut into the muscle, the bone, in order for this data center build out to happen. And that to me is the issue — is like the capital raises never stop, the debt raises never stop. Just a secondary this week. And yet the core business is suffering. The hyperscalers are able to raise money and there's cuts around, you know, across the board, but they're able to continue to sort of build out their business. Oracle's trying to be a hyperscaler. The OpenAI revenue, I get it. It's just in this market right now, they just — I think they just made the wrong bet. I mean, that's sort of my view.
What do you mean they made the wrong bet?
They bet the whole — it's no different than like in baseball when a company spends so much of their salary capital, whatever it is, on one guy. And then after four years and that guy's played — not like Stanton or whatever, but like he's played like 40 games in four years — you're like, we should have just went the way of Tampa and Kansas City and others that have built the farm system. Oracle's a good example. Like they bet hugely on OpenAI with the 300 billion. And the business has suffered significantly. And investors are basically like, how could you own the stock knowing every week there's just going to be another debt raise, another equity raise.
It's a high wire act. Here's the thing. We're going to agree that hyperscaler is very good business.
Neocloud's probably bad business.
Guess what's right in the middle of that? Oracle, right? Because they do have a core business that's very cash generative, right? Oracle databases are not going away. It's a 50% margin business. It's paying the interest on the loans that we're taking to build data centers.
Which is a little bit of what the hyperscalers are doing. It's just that the hyperscalers have such better balance sheets, and Oracle already had a challenged balance sheet because of Cerner and other acquisitions they made. So it's a high wire act. If they execute perfectly, the stock is a multiple of where it is today, but it's a high wire act because if they don't execute well, if OpenAI doesn't materialize, there's a lot of downside.
And also you don't come off those conference calls. Like those calls don't instill confidence.
Because it just feels like there's pirates in the plane. You're not sure if...
Just so you understand, we're talking about the quarterly conference calls when they give you an update. Why does that not give you confidence?
Because you don't understand the direction of the business. Some of the stuff is sort of like long and like word salad-ish. You feel like, is Ted Striker flying the plane?
The reference is to a movie called Airplane, but go ahead.
And I think that's part of it. It's part of like the Intel — that last Intel call, you came off there being like, I got to go to the bar or something like that because you felt like it wasn't just like a soft quarter. You felt like they were just deer in the headlights. And I think...
Let's switch that to Intel. So Intel had a terrible quarter. The stock got crushed. Just explain what happened. Why was the call so bad?
For three months, given memory prices and given what we see in the PC market, you would have thought things were great. And everyone's like, we got it under control, we feel good. If you're an institutional investor, you're like, okay, they, you know, the steady hand type thing. Then all of a sudden that was a disaster quarter and you felt like there's no end in sight and you don't have the confidence of management to sort of navigate that. So I think that like when you like look, Gil and I that have done this, you know, for so many years, there's times where companies, they miss a quarter, something happens, but you feel like the confidence is there where it's a one quarter issue or two quarters. You don't throw the baby out with the bathwater. Then there's others you're like, could this just be a sort of catch a falling knife type stock?
What do you think of the call of Intel?
It's a broken company. But it's always been a broken company.
Yes. Yes, that is the point, but they brought in a very capable CEO.
Lip-Bu Tan, very capable guy, comes in, does the best deals you could absolutely do. US government, Nvidia, Apple, Elon, lines up all the best deals, secures the next node of 14A, gets everybody all excited. But those of us that have been following Intel for a while know that it's a broken company. It's broken in the sense...
Department of motor vehicle.
I mean, department of motor vehicle.
I mean, yeah, get a license to motor vehicle. That's the vibe at Intel.
Their gross margins are in the 30s. They compete with companies whose gross margins are either in the 50s or the 70s.
That is bad execution at every level of the company. And you can bring in a great CEO that can do great deals, but if you can't make chips at a high enough yield to get at least a 50% margin, you're a broken company. And what really puts the exclamation point on that is AMD reporting. AMD is in the exact same markets.
And they're in the 50s. AMD is in the PC market. They didn't bat an eyelid. Intel collapsed because of the collapse in the PC market, and AMD barely nudged based on that. And so you just know it's execution at every level. It's very hard to fix a company that's broken this badly.
Yeah, Intel feels like it's going like a White Castle. AMD is like a real nice steak restaurant. And I think it comes out again and again on the calls relative to, you know, and also how Lisa Su's position is.
I mean, we're talking about one of the best CEOs not just in tech, but I'd say really ever as a public company, you know, across the board.
All right, let's talk about where you guys stand on the debate of moats in AI. To create my podcast, I do research on markets, finance, and anything else relevant every day. In doing that research, I use AI all the time to retrieve data and to help formulate my ideas. And I'm not alone. AI is changing how every business runs, and the leaders who pull ahead will be the ones who know where it matters and how to use it. And that's exactly what Masterclass Executive is built for. If you've ever wanted to learn business from the people actually shaping it, not just studying it, you need to hear about Masterclass Executive. Masterclass Executive is the first AI-native business education program built with the University of Chicago Booth School of Business in collaboration with OpenAI. You can learn directly from Fortune 500 CEOs, operators, AI builders, and Nobel Prize-winning academics, people like Ray Dalio, Mark Cuban, and Paul Krugman. It's designed to fit around a full-time job with self-paced coursework and weekly live sessions. If you're serious about leading in the AI era, this is a program I point you to. I look through the program and the instructor list, and you should really check it out. Seats in each Masterclass Executive cohort are limited, and applying is free. Apply today at masterclass.com/isaacson. That's masterclass.com/isaacson, masterclass.com/isaacson.
So I'll give you the negative. There's a very simple, simplistic negative case, and you guys tell me what you think. The bear case about LLMs and agentic AI is it's incredible technology, going to change the world, but it's very, very easy to switch from one model to the next model to the next model. And with respect to Anthropic and OpenAI, who charge a lot more, you don't need a Mercedes-Benz to go to the grocery store. You could take a bike. So you can use these Chinese models. So therefore, a lot of market share is going to go away. There's going to be a price war. There's word that Anthropic cut prices recently by 50%. You can have a technology that's wonderful, but the business itself is not so great. And that's the negative case about Anthropic and OpenAI. And I would add that if you look at Anthropic's third quarter, the revenue growth rate slowed pretty dramatically, probably because token maxing has ended. And I'm very happy that I know enough to be able to say the word token maxing.
Actually, when you say it, you actually say it like it's like a...
It's almost like I understand what I'm saying.
And you almost say it in like such a confident way too.
So that's the bear case. Tell me what you think about that, and tell me what you think about that.
I mean, I don't disagree. Look, my view, models are going to commoditize over time. You're going to have more and more, whether it's open source, vertical, countrywide, in terms of you have hundreds of models over the coming years. I think Anthropic and OpenAI, they recognize that, which is why they're so focused on the enterprise, which is why every enterprise salesperson I know has tried to get picked off by one of them because they understand like their moat is the initial on the model side and trying to max out their essentially install base. But it's about enterprise implanting themselves and ultimately, you know, and this kind of goes to the debate with Sovereign AI, what's happened with Palantir and Nvidia and others, they're using it as like to open the door through their models, but their monetization opportunity is going to be multitudes of that on the enterprise. Like when you think about Anthropic, if they could do 500 billion, 700 billion in revenue, get to a trillion over the next, you know, 8 to 10 years, it's not going to be charging people like me and you per month. It's going to be on...
Oh yeah, and that is their like, that is what...
So you agree with the concept that the moat is not great.
100%. I agree, but they're using it as a front door to get in, then entrench themselves.
Okay, what do you think about this whole debate?
I have a lot to say. So this is why I'm happy this is long form because I think this is at the core of everything that's going on right now.
So let me start with the good news. They do have a moat. Their moat is the team. They have — there's only 100 people in the world, I'm not an exaggeration, 100 people in the world that are smart enough to move AI forward.
They are heavily concentrated at Anthropic, OpenAI, and Google DeepMind, heavily. The balance is in China of those 100 people.
I noticed you didn't mention Grok.
Elon is very smart too. He's got a good team, so does Zuck, but very few in terms of really being able to compete head-to-head with Anthropic and OpenAI.
So here's where we're at. That means that those two will continue to build the best models. And Anthropic made a really smart move, as Dan is pointing out, to pivot to the enterprise, and by the second quarter they had an overwhelming share of the enterprise market in terms of tokens, right? And then a couple of things happened. One is OpenAI caught up, and in the third quarter they gained share. And the other thing is CTOs were spending so much on tokens that they realized we can't keep doing this. We have to have open weight alternatives, and they started diversifying their AI spend towards that. And another thing is happening, which is the emergence of diffusion layer, control plane, orchestration layer, harness. I'll introduce those concepts in a second, but the point is that if...
Anthropic started with a market that was this big, right? Now their market is narrowing because a big part of it's going to open weight, because to summarize my email, I don't need the frontier model. I can do any acquisition.