Where Brad Gerstner Is Investing Billions

Great to be here, guys. Great to be here.

I mean, come on guys, we didn't even have any foreplay and you're already driving me down the SpaceX IPO. I mean, first, I haven't seen you guys. Did you sell this thing?

Did you guys get shares in OpenAI when you sold this thing?

So we're all on AI. Okay.

Well, I mean, thank you. And I recall that conversation and the truth of the matter is we've gone through several many corrections over the course of the last two and a half years, right? There's been a wall of worry. I mean, on my podcast with Bill Gurley, you know, we debated this. Bill was saying, will the revenue show up? Will there be gross margins? Will there be ROI? We're overbuilding. Are we, you know, every supply constraint turns into a glut? And we saw all that wall of worry last year. I mean, I think for me the turning point was when we hit inference time reasoning and we really had this whole other vector of scaling intelligence. And I remember having Jensen on the podcast and he said, 'Brad, inference isn't going to 100x. It's not going to 1000x. It's going to 1 billion x because agents are going to be talking to agents.' And so I got very pilled. And then when we saw Opus 4.6 in the beginning of December, it was clear that we had crossed a threshold of intelligence that offered a level of utility that was fundamentally different. And if you were paying attention early in December, you could see that coming. But we started the year with the market very skeptical as to whether or not AI revenues would show up. And let me tell you this, had Anthropic not delivered its revenues that it's delivered this year, I think the stock market would be down 10 or 15%. I think it's that important to the entire narrative because the fact of the matter is OpenAI has not blown away their numbers. Google has not blown away their numbers. Like, numbers have been good but the fundamental driver of outperformance in terms of offtake of AI revenues has been Anthropic which is the fastest growing company in the history of capitalism. So that buoyed the entire AI segment and it was when they started posting those numbers and then they said on top of that we're doing it at high gross margins in a way that in Q2 may in fact actually lead to some positive free cash flow. The market really ascended. Remember two months ago the market was basically down on the year and a lot of these returns, we just had two of the biggest months in the history of Altimeter's public funds. That's 18 years. I mean, we're going back a long time but that's, you know, listen, I think we picked some pretty good stocks, memory, logic, etc. but I also think it's just a function of the market delivered, these companies delivered. You saw Dell's, I mean, listen, Michael Dell, one of my best buds, and you watch the act that he's delivering with Dell. They just had AI server revenues up 750% year-over-year. Went from a $1 billion business to a $16 billion business. This stuff is real, but in order for it to stay real, we have to continue to see usage by consumers that they're willing to pay for and growth in the enterprise, small, medium, and large that they're willing to pay for and growth in the sovereign domain. I think it will occur, but oftentimes, you know, there'll be some pockets along the way here where revenues won't be as strong as people think. We'll have some pullbacks. We could have 10 to 20% pullbacks in the semiconductor stocks as just like run-of-the-mill consolidation.

Run-of-the-mill consolidation. I mean, Micron has gone from a couple hundred bucks to a thousand bucks. Dell this time last year was, I think, 80 or $90. It's now at $400. These are seismic moves.

Yeah. Well, that's the most extreme version. The other one is, yeah, we actually did spend half a billion dollars in a month and, you know, a quarter billion was super effective. So that's what we're doubling down on in the coming quarter. But how do you think that shakes out?

On the other side are the people who are super AI pilled and they're like, oh no, this is perfectly, you know, Pareto optimal. Everybody's spending the exact right amount of money on tokens, which we also know is not true. And the truth lies in the center. When you have millions of independent actors all making self-rationalizing decisions like Altimeter on buying tokens, right? I don't like to waste money. I'm spending money because I'm getting a return. Now, will we experiment with some things that don't provide a return? Of course. I actually sent you guys a slide on this. I think it's pretty fascinating. I don't know if this is independent research that we did on this question of token maxing and what we did is we went to 300 enterprises, right, and we just asked them, are you starting to optimize your spend and if so how much do you expect that you're going to spend year-over-year over the course of the next 12 months. And if you leave that chart up what you will see guys is that in the first category these are all people who are actively optimizing but they still expect to grow revenues at over 50% over the next 12 months. The second category are people who are planning to optimize and even if they're planning to optimize, they say they're going to grow revenue at 90%. And so on. So here's my point on this, right? And this is across 300 firms who use a multiple of AI solutions. This is what they expect of their AI API token usage. So what does that tell us? It tells us that of course people optimize along the way but we are so early in the adoption curve, right? They're barely using coding today. They're just getting on the coding train and they haven't really even started on using AI for knowledge work more generally. So we're low in the penetration of coding as a use case. We're almost nowhere in the penetration of knowledge work more generally as a use case. And then remember this, there are very few enterprises, you know, globally that are even using AI. So, we're really early in the curve of the people who are actually using AI. So, I'm somewhere in the middle. Of course, I believe that optimization will continue. But my hunch is that Anthropic and OpenAI, these companies will continue to grow right through the optimization because the growth curve on penetration of both enterprise and use case is so steep. But, you know, we'll see.

Well, I popped on CNBC for a second yesterday and was talking about Snowflake as an example. I mean, the stock was up what, 35% yesterday. Now, of course, just to be fair, it's only up 10% for the year. Compared to a company like Micron up 200% for the year, a company like ARM up 200% for the year, but they did bounce back. And what I think we're starting to see is the bifurcation. There are companies that are in the token flow. So all these software companies we just lump together, right? We treat them as though they're all equal. But there are certain software companies, Databricks, Snowflake, and ClickHouse, all of which we're investors in. It's very clear to me they're in the token flow. As you consume more tokens, the amount of your database queries goes up. I see it at Altimeter, right? In fact, our database queries are growing faster than our token usage to give you a sense. And this is, I think, so now they've proven they're in the token flow. So they're starting to get some love from an AI multiple perspective. That's very different than a company I think like Salesforce. And I love Marc Benioff and, you know, if anybody can get in the token flow it'll be him. But the reality is the front-facing solutions that they offer are more competitive with the models than something like Snowflake. Snowflake's the enabler of the models, whereas I think that Salesforce competes a bit more. So, it's going to be more challenging. But I also, I've heard so much about this SaaS apocalypse and listen, I did a pod with Satya, I don't know, 18 months ago where he caused a stir by saying software is a thin user interface on top of a CRUD database. And Benioff and everybody freaked out. They're like, what are you saying, it's way more than that, right? And then Bill and I did a pod, is software dead? So it's not like this is new but then everybody started freaking out in December, all these multiples reset. But the question is what did they reset to, okay? And this is what I want to focus on here. So if you show this slide that I prepared for you fine esteemed gentlemen, you know what this slide shows is that the multiple correction just took software from a place where they were way more expensive than the market multiple and brought them into the category of the market multiple. Right. So now they're trading at about 22, 23 times real SBC included GAAP earnings. That's about where the market is trading. So now just follow me on this. Software is trading, most of these software names are trading at a higher multiple than Nvidia. Right. Nvidia is trading about 13 times earnings for 70% growth for the thing that is the most essential thing in AI and they're at twice the multiple. So like when I hear everybody crying that hey these multiples aren't fair, it looks to me like the multiples reset from an above market multiple where everybody thought the software revenues and earnings were impenetrable to now they're saying well I don't know some of this maybe three, four, five years out will be replaced. So, we're going to raise the discount rate. We're going to lower the multiple. They've only lowered it to the market multiple. Let me just suggest that there's a possibility these trade well below the market multiple. Right. I'm not wishing for that, but I'm just saying there's a distribution of potential outcomes here. If you get on the AI train, if you get in the token flow, you're going to get an above market multiple. If you don't, if you slow down and it looks like every time that computational intelligence improves your business gets worse, then I promise you they will trade below the market multiple and there's more room to the downside. So for us as investors, you know, Warren Buffett has this old metaphor, you know, there's the easy basket, there's the hard basket, or the yes basket, the no basket, and the too hard basket. For me, software today is generally in the too hard basket.

I think it's bad for everybody. It's bad for everybody but most importantly it would be horrific for America. Yeah. Unless we be overconfident in Silicon Valley. Let's remember that activists, a small group of activists shut down supersonic technology and a small group of activists shut down all nuclear clean energy in this country. Okay, we have 100 fission reactors being built in China. We have one in the United States. It's a disaster that happened. And so we can't take for granted that the kooks who are calling for data center moratoriums, right? Which just think about this for a second. All of our GDP growth is coming from the fact that we are building data centers and driving AI and driving productivity improvements in the economy. A data center moratorium would thrust us straight into a recession and high unemployment. Secondly, it would cede the entire global game to China. Like overnight we would lose to China in the global AI race, which is not just about AI. It's about economic security. It's about jobs and it's about national security. So it literally is insane that we would do this. I can't even believe there are people talking about it. However, why are they talking about it? Because people are concerned. Local communities are concerned. I just got back from celebrating my mother's 90th in rural Indiana over the Memorial Day weekend. And what I'll tell you, she's incredible. She is so incredible. But you think about a place like Mishawaka, Indiana. Where, you know, they're building a data center. I mean, folks here, they're worried about their jobs. They're worried about their kids having jobs and then they're told by these crazy activists who show up in their town they're not going to have any water and their electricity bills are going to go up. So, can we blame these people for being a little agitated about what's going on? So I'm actually working on an initiative I'm not prepared to announce today but with like everybody in the value chain, all of the cloud companies, all of the NVIDIA and AMDs and offtakers etc and the White House that would deliver a very tangible and profound dividend to the communities that we're building.

Well, I'm in the mix. I'm happy to do my part. There are extraordinary people around the table. But here's the thing. We have to build the socio-political bridge for the next three years, right? In three years, it's going to be obvious, I think, the abundance and the benefits that AI is driving for us as consumers. Everybody's going to have their own personal assistant in their pocket, right? For next to nothing. Think about that. Can do your calendar, can order your food, can get you a new black t-shirt, send mom a birthday present, all the things. And every enterprise is going to have things that uplevel us all as humans. So I am firmly just like John Maynard Keynes was at the start of the industrial revolution, I am firmly in camp optimism about technological progress, but I'm also not head in the sand about the disruption and the concern people have for the next three years. So, we have to give them tangible benefits that get us over that bridge. I think we're going to do it. I'm feeling pretty optimistic about it, but you're right to bring it up and you're right to be concerned about it. We cannot take it for granted.

For sure. I mean, listen, I think about when I got into the game, guys, 1999, 2000, and we had about 35 million people connected to broadband internet, right? We all saw what Amazon was going to be, but where we got over our skis, right, is we thought it would come a lot faster and we forgot that there were only 35 million people connected to broadband internet. Today, we have four billion, three, four billion. Like, the rate of diffusion and the magnitude of diffusion is radically, radically different. And think about this. We have a natural constraint on how fast we can go because we only have so many memory wafers in the world. We only have so many logic wafers in the world. We only have so much powered shell in the world. That means we can only produce so many tokens. Okay? And it's almost as though in 1999, 2000 we could only lay so much fiber. I've said this a thousand times. When we were putting down the fiber in 2000, we called it dark fiber for a reason. There was nobody using it and we knew there was nobody using it when we put it in the ground. There's not a dark GPU in the world today. Okay, there's not a dark token in the world today. So I think it's a very different thing. I think it's a healthy thing. We have this wall of worry. We can't build that much supply. And I would say if I look at every company, what did they report on their earnings calls? Google was token constrained. They said if we had more tokens we'd be able to generate more revenue. Same for Amazon, same for Microsoft, same from OpenAI, same from Anthropic. The world demands more intelligence. Intelligence can only be produced with tokens. And we have physical limits to the amount of tokens we're going to be able to bring online. So yes, we will have these waves, but I think the rate, the parabolic rate at which these new models are going to produce intelligence, I think we're going to be blown away over the course of the next nine months. You talk independently to Michael Truell and the guys at Cursor and now taking over xAI or you talk to the guys at Anthropic or OpenAI and they kind of look you in the eyes with that Oppenheimer look and they're like, we're here. Right. Like, we are going to, think about this. OpenAI and Anthropic combined to start the year had three gigawatts of compute. Three combined. They're going to end the year closer to 10 and end next year closer to 20. Yeah, we're making algorithmic improvements. We're making massive steps up the scaling wall because the amount of compute we're going to have available to us. Think about, you know, Microsoft and Microsoft harder that Cursor is going to now be able to train a frontier level model on. So, we've got incredible competition in America. We got the right amount of compute coming along. I don't worry about the bubble as much even though I know that there will invariably be some months that revenue doesn't grow as fast. I'm really worried about making sure that America stays foot on the accelerator, competing globally, and winning the AI race. Like, this is going to lead to a moment of abundance for our economy. And it's only through great national wealth that we can raise the floor for everybody else.