Coatue 的 Laffont 兄弟：AI、公募与 VC 市场、宏观、美国债务、加密、IPO 及更多 | BG2

Sometimes you make some venture bets and they don't work, and then you're like, "I just invested in the wrong trend." In fact, sometimes you invested in the wrong company, but it was the right trend, and those bad investments cloud your judgment.

Bill, we're back. I think it's the 10th anniversary. Congratulations, Philippe and Thomas. Of course, we're at Coatue's East Meets West down here in Los Angeles. I think it's an event that founders—and certainly you and I—look forward to every year.

As I said to you both, it's hard to put together something that has this much durability and this much impact. You do this incredible overview on public markets, venture markets, and technology that I think you publish online today, and everybody should go out, download it, and take a look at it. We've been at this now for a couple of decades. Having built something like this is really cool, so I just wanted to say thank you and congratulations on the 10th anniversary.

Bill and I thought, why don't we just go through it? You had this slide today, and we got to sit through and listen to you guys commentate about some of these slides. We wanted to share it with everybody else. We're also excited to make our debut as a podcast duo—the world premiere. We've done them individually, but not as—

Oh, are we announcing our new podcast?

Yeah, exactly. We've got BG2, and now we've got LB2: Laffont Brothers 2. Let's go, let's go. So we're squared squared.

By the way, I would just add that I think the conference is an amazing gift to the industry and to the founders who get to come. It hearkens back to when I was really young in this industry, when at the Agenda conference everyone would stay for the whole thing, so your opportunity to network was much higher. A lot of conferences today involve people flying in and flying out, but here you've got some amazing people who are around for the entire thing. It's just incredible.

Well, let's dive in. You have a big budget for smoothies. Your smoothie budget really keeps people in touch.

By the way, for people who are listening, this deck—the Coatue team put it on their website just a few hours ago. If you want to download it and have it as we go through this, it might be helpful.

1. The AI Super Cycle

Yeah, you should. Philippe, let's just start off. It seems like you and I spend most of our time talking when things get bad in the world, and yet this is probably the most optimistic I've heard you on this stage in the 10 years you've been doing this.

You talked us through slide 4, the AI supercycle slide, and slide 6, "When Will AI Reach 75% of Total U.S. Market Cap?" I thought that was incredible and incredibly provocative—how you compared it to industrials and transport—because everybody's saying it's so big already that it can't get any bigger. Just kick us off by contextualizing your level of optimism and this slide. Can it really be 75% of total cap?

Yeah. So listen, every time I'm optimistic, I'm worried this is it—this is the peak. Now that Thomas and I are doing this podcast together, we're guaranteed to be doomed. But I think that, at the end of the day, that's how everybody thinks, first of all, so it's never priced in. Everybody's worried that it's always the peak, and yet, despite that, things tend to work out.

I think today we've learned from these founders that AI is probably the defining and biggest tech trend that we're going to see. I showed you the different waves. There have only been a few waves over the last 70 years or so, going back to mainframes. One person made the point that for networking, we needed PCs; for the internet, we needed networked PCs; for SaaS, we needed what happened before; and AI is also built on what came before. One of the reasons these trends get bigger is that they're built on top of each other.

2. Private Markets, IPO’s, M&A’s

That's one. The second part is that we've tried to do—and, Bill, you've been great at it, and Brad, you've done it too—let's always try to look back at the past. I find that this concept that, even though we're talking about new trends, they've been new trends since the canals and whale oil and things like that.

You look at the 1800s, and we started having a real finance and real estate industry. Then, probably at some point, especially after the Second World War, we had a real manufacturing industry. We've also had a market dominated by energy, and right now it's about 50% tech.

We had the CEO of the largest power company—sort of utility—with us today. We had the CEO of the largest equipment maker for utilities today. You're wondering not just whether AI is going to become bigger and TMT is going to become bigger, but whether there are some sectors that we should reclassify as TMT or utilities now, like the next semicap. What's the difference between your nuclear energy plant and a semicap company? They're both there at the beginning to help you create something that delivers a tech product.

Put another way, technology, when we got started, was 5% of global GDP. Today it's 15% of global GDP, and when we're sitting here in 10 years, I think you're saying confidently that, while there'll be a lot of noise and a lot of volatility, it'll be more than 15% of global GDP.

You guys talk again about the new class of AI entrants. The Magnificent 7 has actually underperformed this year, but we have AI power, AI-related software, and AI semiconductors that are up on the year. You guys have diversified out, Philippe, into some of these other categories. Is that the case—that everybody got crowded into the Magnificent 7, and now you see all of these other companies accelerating this year that are starting to get some of the benefits? Maybe, Thomas, you should take it, and also contrast it to what's going on a bit on the private side.

There was a time when the Magnificent 7 was a real source of excitement, and now it's changed a bit.

Yeah. So it was interesting seeing that, on average, the Magnificent 7 was basically flat year over year, and yet there was tremendous value accretion to the top AI companies, whether it's OpenAI, Anthropic, or all the companies that follow.

To me, my other takeaway, looking at this—and I was thinking about CoreWeave, which recently went public and which you guys are big shareholders in—we are big fans of the management team. I think there's a lot of skepticism around that business and that business model, but at the end of the day, being an AI pure play, there are very few in the public market.

Right. Right.

So I look at this list, and there are amazing companies on it, but a lot of them might have legacy businesses or other—

Right.

I think of Google as an example. It certainly has a lot of good AI, but it also has some disruption threats. Seeing new entrants like CoreWeave that are a pure play on the trend has been a really positive development as well.

3. Stablecoin & Cryptocurrency

Another thing—today is an appropriate day to talk about this—the stablecoin legislation passed today, which is a major step forward for the regulatory framework around U.S. finance. We're going to want to talk about this later, but it was a major step forward.

Philippe, you were funny today on stage talking about Bitcoin. You said it's this category that's broken out. You lose sleep over it every single night because you're still not invested in it from an institutional perspective, like a lot of us. And yet, you showed this slide 18 where you said maybe the volatility of Bitcoin is coming down, which might put it more into an institutional asset class.

Talk to us a little bit about how you guys think about crypto, maybe at the private-market level. We all have post-traumatic stress from the 2022 period, I think, of venture investing in crypto. Is that changing? Is it now in 2025?

I do think it's really interesting to think of Bitcoin as a company for the sake of our investing universe. We do think the relative market caps become really interesting. As you see in some of our decks, especially at the end, the first thing is awareness. We need to include the large ones as we think about how they're valued versus other things. And so, how do you think about valuing it?

Listen, just touching back on your point, we're looking at Bitcoin. The market cap—the net worth—of the world is about $450 trillion to $500 trillion. Equities, I think, are about $120 trillion. Real estate's probably another $100 trillion to $150 trillion. Then there's the value that people have in their homes. Gold is about $15 trillion to $20 trillion, above and under the ground.

Then we're like, Bitcoin at $2 trillion. I'm like, "God." Bitcoin represents $2 trillion out of $500 trillion of the net worth of the world—or $400 trillion, whatever; it moves a little bit. Could it be $4 trillion? Could it be $5 trillion?

The largest company, Microsoft, is about $3.5 trillion today. Let's say Microsoft doubles in 10 years. It would only be growing at 7% per year. Microsoft will be a $7 trillion company in 10 years. Could Bitcoin be $5 trillion or $6 trillion? It's a real asset class. On top of that, it's very volatile. On top of that, there are a lot of retail people who own it, and it almost feels like sometimes the institutional investor is wrong and the retail investor is right. Sometimes it's the opposite.

Retail gets caught in a little bit of a meme stock, and it comes back down. I don't think we can afford to ignore it anymore. So it doesn't mean we don't really know exactly when and how to own it.

Your other point that's really interesting is that sometimes you make some venture bets and they don't work, and then you're like, “I just invested in the wrong trend.” In fact, sometimes you invested in the wrong company, but it is the right trend. And those bad investments cloud your judgment.

And there's Bitcoin, there's stablecoins, which we should talk about. They're growing incredibly right now. And then there's all these altcoins. You could say, okay, well, I don't like the altcoins and the meme coins. I don't necessarily like the collectible aspects of things, but I like stablecoins and Bitcoin.

So for us, it's more a process where we just need to become better, be willing to change our mind, and stay open to the future.

Those are a lot of the conversations.

I agree. One of my biggest lessons looking at private-market investors versus public-market investors is that the appetite of institutions in the public market for assets that are perceived to have significant downside—i.e., like 70% or 80%—that are marked to market, I have found, is just really low.

Right? Investors on the public side just don't want to take that kind of risk. Whereas on the private side, you are willing to take that risk because you may have 20 of those. They're not marked to market, and you're like, “Look, maybe 5 go to zero, but my other 10 go.”

So I do wonder how institutions versus retail may be willing to take that risk. I wonder how institutions will think about an asset like that.

Let me telescope out for a second, and I want to get Bill's opinion on this as well. I think all of us, now a couple of decades into this, would say one of the most powerful things about this conversation is mental flexibility.

Absolutely. When you're maybe a little bit younger in the business, you're more dogmatic. You develop an opinion and defend it to the hilt, right? And if you're wrong, it can be extraordinarily costly. I think crypto was that way for a lot of people. They carved out these positions; they were like, “This is a fad,” and then they're proven right at a moment in time because it'll have a 50% drawdown. So rather than reevaluating their priors, they lock into that position.

Bill, how have you thought about this? I find venture particularly tribal about this. Look, you're locked in; you can't sell. I think this is something that you guys develop more of an instinct for in the public markets than in the private markets, because you're in forever with the private companies. You can learn lessons along the way, but your windows are really long, right? Whereas if you're in public stocks, where you can change your mind and make a decision right away, that's very different.

Yeah. I mean, you referenced what sounds like Druckenmiller today. You said you think this may be the most valuable attribute of the great investors. When he told me, “I've made 120% of my money on obvious ideas, and I've lost 20% elsewhere.”

Then you start thinking of Bitcoin and a company being like the 5th-largest company in the world. It's a bit odd what I'm saying, I recognize it, because you could also say, “Well, should we consider gold as the largest company in the world because it's worth $20 trillion?” Not necessarily.

But I do think forcing yourself to think differently and at least being at peace—“Okay, I thought differently. I came to the same conclusion”—and being able to do that is important.

[Speaker?]

Now, as to us being flexible, the fact that you think that French people are highly flexible people, I'm very thankful for that. I'm not sure it's true, but we'll take it.

Two things that are new about crypto should lead anyone to reevaluate. The government's gone from being kind of antagonistic to supportive. That's a big shift, because regulatory risk was a big question for all this stuff. And then the stablecoin—based on what people are talking about—this is a high-utility use case for people; companies are using it as part of their workflow process. That's a new dimension as well.

One additional point on that that I think is interesting is, when you talk about the US dollar, the view is always, “Well, what's the alternative? I'm not going to go into Europe. Do I want to go into Europe? Probably not.” It's kind of interesting: What if the alternative is actually Bitcoin?

That's something I've been spending time on. We talked a lot today about the dollar and interest rates and what's going to happen, so it'll be interesting to see whether that becomes a legitimate alternative to the overspending of governments.

When you have a stablecoin, how long is it before a new regulation goes through that allows a stablecoin to pay interest? It's sort of odd: Stablecoins can offer rewards but can't pay interest. And when you have a stablecoin with interest, how long is it before the government creates a 1-year stablecoin, a 5-year, a 10-year, and a 30-year stablecoin, which will allow every single person around the world to invest in the USA?

The government is going to have an incentive not to have these bonds sold through these weird dealers and this and that. The government should go direct to the consumer, just like companies do. So I bet you that in the not-too-distant future, people will be able to automatically invest in bonds.

That's yet another example, on top of what you were saying, Tom, about Bitcoin and stuff. Anyway, we need to switch topic; otherwise, I'm going to really pull the few hairs that you and I have left.

4. Consumer AI & Impact on $GOOG

Okay, back to AI. One of the topics was consumer AI. You guys had an incredible audience here. You had Andy Jassy here talking at lunch, and Kevin Weil from OpenAI. One of the most incredible pieces of data that you guys shared was looking at the impact that ChatGPT, which is now scaling to about 1 billion users, is having on Google.

You did this by conjoining a couple of pieces of data that you guys had. So, Thomas, do you want to talk to us a little bit? This is slides 22, 24, and 26.

I'll pass it to Philippe for this chart, but I think Bill and I were chatting about this earlier. Anecdotally, it certainly seems to be the case, right? The more people I talk to, the more I ask them, “Do you feel like your Google search has been impacted by ChatGPT?” And resoundingly, almost everybody at this point now agrees that's the case.

We can argue whether the queries are commercial or not. I think the queries are getting more commercial every day, but without a doubt, it's having that impact. We could not prove it numerically. It felt intuitively true. Obviously, Google is telling you it isn't.

So I think we went about seeing whether there was a numerical assumption that we could make that would kind of prove this out. And I think you should introduce the work.

And by the way, as we all know, all these platforms have some businesses that get threatened and other businesses that benefit. Google could still be an amazing company by just saying, “Listen, maybe search is under threat, but YouTube, with all this new AI content, is going to explode and potentially threaten Netflix, and maybe what sounded like Waymo is going to also do incredibly well.”

Our judgment is more around exactly what we talked about there. I mean, the assets of the Android phone and Gmail and Google Docs—that's a nice set of complementary pieces. They have so many great assets. It'll be very interesting to see how it plays.

For me, if I were CEO of Google, that would be way above my pay scale. I had no idea how to put it all together, but God, is it just fun to be alive and see what's going to happen. What's Amazon going to do? What's Google going to do? What are all these guys going to do?

So what we try to do here, as part of the data science that we do, is process probably 100 million credit card receipts a day. We have a very fine view of what the US consumer does. We have another data set where we know what consumers do based on their email receipts.

The trick was to try to join those 2 data sets. And in general, in data science, my only lesson learned is that data is useless unless you can join data sets that don't speak to each other. That is the unlock. And so we did that.

What you see on that chart is that absent ChatGPT, Google page views for a particular user were maybe growing 4% per year. Then we get a subscription to ChatGPT, and now we're like, “Ah, this guy's paying $20 a month.” Then we track what happens to the usage once he started paying $20 a month to OpenAI, and you can see peak to trough it's down 8% year-over-year. Peak to trough, it's, let's say, down 11%.

So clearly page views are going down, and that's over almost 2 years, right? So it's not like it's immediate. But one thing we've learned—and Thomas and I repeat that to each other all the time—is that these major shifts just start one little step at a time, and that one little step becomes a gigantic move quickly. So you can't underestimate these small moves.

I think this confirms something that we all know anecdotally as we're talking about it.

Well, and I think that even slide 24, when we were talking 2 years ago about ChatGPT, we knew it was off to a good start, but the question was, what's going to happen when Meta gets its game going? What's going to happen when Google launches Gemini? What's going to happen when Elon launches Grok? What's going to happen when Claude gets better?

We all thought that when they got into the game, this line would start to flatten out.

But the fact of the matter is ChatGPT has been radically more resilient, and engagement has increased much faster than I think any of us would have thought with that level of competition. What’s interesting is that this is true in the U.S. and internationally. It’s true whether you look at downloads or engagement. It has blips here and there—the DeepSeek moment and others—but the resiliency, to me, does remind me a little bit of when Uber got started. They established that market share, and it was incredibly difficult to disrupt.

For listeners who don’t have the slides, we’re looking at ChatGPT adoption against Twitter, Instagram, Facebook, and TikTok. It’s just straight up and way ahead of those. By the way, those apps had inherent virality, as you know. I mean, you’re kind of the expert on that. ChatGPT doesn’t; it has no virality to it. It’s just value to the consumer, and that’s driving adoption.

Although I would say that we’re starting to see network effects on the data side. We’re starting to see switching costs with persistent memory, as you and I have talked. What’s amazing is that you have this level of adoption even before those things begin to kick in.

But it confirms what we kind of know to be true. We saw this with Google, right? We saw this with Facebook. And now we’re seeing it again with Kevin Weil, who was on stage after you guys talked. He made an interesting comment. It’s tautological, but it still resonated with me: “Look, this product’s going to get better.”

So you have all of this adoption with a product that’s not even 3 years old. We could show—or maybe we did show—the stat about usage in terms of minutes: more weekly users, more daily users, and then more time per day, which is a lot. That’s also consistent with all of our personal lives.

5. GPU Allocation v Cloud Revenue Market Share

We’re going to keep forging ahead here. We’re going to get crunched on time. Slide 27, Bill, is a slide I know you wanted to talk about when we talk about these new hyperscalers. What you did here is map cloud revenue market share against the share of NVIDIA GPUs. So, Bill, why don’t you—and I’ll just describe this so people listening can follow along—and then we’ll ask you guys to talk about your takeaways from it?

The Coatue team mapped out cloud revenue market share, and you have Oracle at 5%, Amazon at 44% because of the success of AWS, Google at 19%, and Microsoft at 30%. Then you show, right next to it, the share of NVIDIA GPU allocation. Microsoft and Google are roughly equivalent: 30% and 20% of GPU allocation, respectively, which is close to their cloud revenue market shares. Amazon notably has 44% of cloud revenue market share but only 20% of NVIDIA GPU allocation. Oracle jumps from 5% to 19%, and CoreWeave comes out of nowhere to be 11%. So tell us why you guys put this together and what your big takeaways are.

For me, as an analyzer of companies, this might be my favorite slide because it shows the competitive dynamics at work and whose strategy will win out. I look at this, and one obvious takeaway is that Amazon has half the share of GPUs as its share of AWS. That could mean one of 2 things: either AWS is behind in AI, or it’s pursuing a different hardware strategy than its competitors, which Andy spoke specifically about. So it could be one or a combination thereof.

Number 2, it shows the reinvention of Oracle: left for dead in the 2000s, left for dead in the SaaS era, left for dead in the AI era, and now coming back. I also give CoreWeave a tremendous amount of credit for just entering the market as a pure play. It had difficulty raising capital. None of us ever believed it—there’s no IP; you’re just buying GPUs and reselling them. Just by being in the market and being focused, CoreWeave started to build that relationship with NVIDIA and now is punching way above its weight.

By the way, the third theory could just be that NVIDIA would prefer not to have a dominant customer. They wouldn’t want this to be the case, though it hasn’t seemed to impact Microsoft and Google.

Do you want to add anything?

Yeah, I mean, listen, the one thing on that chart is that it’s damn hard to get the numbers right. We have to explain to viewers that we could be off by 5% or 6%, up or down. But I think where we’re not off is the concept that some players are getting more GPU chips than others. The question then is: Are NVIDIA GPUs a predictor of future cloud revenues?

I think the answer is yes, and we haven’t even included Stargate, which is going to start coming up here, right? What if Anthropic also becomes its own hyperscaler? You could have a world with more like a dozen hyperscalers than the 2 or 3 that we have today. Then you’re going to have the sovereigns, for sure. So you’re going to have some telecom operators, more traditional operators in Europe, and so on. There will be more, right?

But I think what’s definitely going on now is there’s a battle between people who want to standardize on NVIDIA, pay the NVIDIA rent, and get the supply, versus people who also think, “Hey, I’m bringing a lot of software, I already have a lot of the data, and I can afford a different strategy.”

In the internet era, almost every startup started with Oracle and Sun, and 5 years later they weren’t on it. So there is some precedent.

There is. And I also think the other thing that surprised me—although I even have a hard time believing that those are the numbers—is that I thought Google was more skewed to TPUs than NVIDIA. So there are some people who are going exclusively with one chip, and there are some people who are going to go in a hybrid way. Google has both NVIDIA and TPUs.

I think Amazon is also choosing a path of, “Hey, we’re still making a ginormous bet on NVIDIA, but we also would like to have our own bet.” I wouldn’t be surprised if maybe someday Anthropic or maybe even OpenAI said, “Maybe we should design our own chips.”

Then, frankly, you might have some very expensive model with enormous reasoning that runs on NVIDIA, and maybe a super-cheap model just for some very local application that could run on a custom chip. So I think a lot of it is going to morph and change over time.

But at least what’s fun here is, let’s go revisit this chart in 5 or 7 years and be like, “Okay, different people play chess in different ways.” What’s happening?

And I think, to me, the thing that stands out most about this slide—again, slide 27—is Microsoft. You talked about the explosion in terms of token production. We might be at 100 trillion tokens a month already out of Microsoft. Microsoft is OpenAI, so you’ve got ChatGPT at 100 trillion. I know you knew that.

What’s really driving inference and this token explosion? Consumers, first and foremost. Google’s got Gemini, right? Microsoft, derivatively, is supporting ChatGPT, as are Oracle and CoreWeave on the slide. Amazon doesn’t really have a big consumer application, so their need for those GPUs may also be a little bit lower. Correct?

6. AI Impact on Macro + Debt

I want to jump ahead a few sections because I want to get to the private side, the venture side of this, but I want to end the public side with the macro backdrop. Philippe, you’re one of the best. We’ve been at this a long time. We know that we invest in companies that are doing extraordinarily well, and we look at fundamentals, but you can’t ignore the macros.

Dan Loeb says, “If you don’t do macro, macro does you.” We found that out the hard way too many times in our careers. But if you obsess about it, it can also be your undoing. One of the things I thought was so interesting is that we’re at this moment in time where we’ve heard from Elon and the guys on the All-In Podcast, David Friedberg, and others who are saying, “We’re in this debt spiral. There’s no way out of the debt spiral.” And yet, if you look at the 10-year, the 10-year is still at 4.43% or 4.44%, right?

Despite the calls that it was going to be at 6.5% or 7%, we haven’t gotten anywhere close. We’ve been in a band between 3.5% and basically 4.8% now for 2 years. And you presented an argument on slide 45 about the productivity cycle that may come out of AI, that may drive faster growth in the economy, much like we saw in the ’90s with the internet. That could, in fact, lead to lower inflation and lower rates on a permanent basis, kind of this backdrop that would bring the deficit to GDP below 4%.

I know you guys work closely with Larry Summers and others. So, as you think about it, how important is believing this to be true in our overall public investing today?

So, your original question, Brad, is: Should we be worried that we all think AI is a big deal? The counter to that is to say, “Okay, what if we’re right on AI but we’re wrong on something else?” And we actually analyze 3 things.

We analyzed, “Are markets expensive?” And the answer is yes, but markets were expensive in the ’90s during the PC and internet era, and the market did well. So that’s number 1. Number 2, we said, “Well, are tariffs a big deal?” And we said yes, they’re important, and maybe they haven’t gone through inflation yet. But this doesn’t feel to us like—I like to say that the tokens trump tariffs, basically. And so we’re basically left with this deficit.

The first thing I tell people about the deficit is that having DOGE and having people like Elon say that we’re spending too much is useful.

Yes.

And we should repeat that every day.

It doesn't hurt. But what I was wondering is, since it's so obvious that we need more DOGE, need to spend less, and stuff like that, who are the people who every day are buying 30-year bonds at 4.5%? I'm sure your listeners know this, but a 1-year bond at 4.5% stays and gives you 4.5%. A 30-year bond at 4.5%, on its way to 6% or 7%, could cause you to lose 60% or 70% of your money.

Yes. You know, once you have a 30-year multiplier on a change in interest from 4.5% to 6%, right? Our basic instinct was to analyze what happened during the internet and PC days, when we had exceptional productivity gains as the internet and PC really took off in the ’90s, and to say, “Hey, what would happen if we had similar exceptional productivity gains?”

Basically, the answer we were trying to solve is that, in essence, today we’re at 100% debt to GDP, on our way to 140%. We asked what it would take for debt to GDP to stay at 100% or maybe even bend the curve and go down to 80%.

What’s really surprising, I think, if you just show maybe the next slide or so—if we move forward just a little bit—you’ll see that if productivity for the next decade or so was about 2.5% to 3.5% per year, we could achieve substantial reductions in this key ratio of debt to GDP. I’m not saying we’re there, but I’m saying that at least we’ve been able to bookend what productivity would need to be to achieve 80% to 100% debt to GDP instead of 140%.

This is slide 51, just for the people following along, which is, again, an incredibly important point. We know there are people buying bonds every day at 4.5%, so the question is, why are they doing that?

One of the answers may be exactly what you’re saying: What if they’re right?

Exactly. What if they’re right and we’re wrong? In fact, one funny part is that in 1993, debt to GDP was supposed to go from 60% to 80%, according to experts, and it in fact went from 60% to 40%. Experts can be wrong by a lot.

Right. And so I’m not a good enough macro guy, and if tech guys pretend to be good macro guys, it’s the beginning of the end. But at least we have a little bit of analytical thinking around what it would take.

And Bill and Thomas, you guys are much better placed than me in terms of your discussions with all the privates, which I think we’re leading to now, and all these amazing new products. You’re telling me that that’s not going to create massive productivity? I really think it is. Drawing from that, you would end up with GDP growth way more in the 5% plus, maybe even 6%, which, by the way, was the case for many of the years in the ’90s.

By the way, the 6% would represent 4% in real terms, whereas in the past, most recently, we’d more be at 2% or 3%, which is more like 1% in real terms. So, just to wrap up your flight path for the public markets, I think it’s fair to characterize it as tariffs fairly much being under control. Multiples are pretty full, but like they were in the ’90s, they can stay full. The backdrop is okay. The bond market and rates are still in the 4s, and we have this AI supercycle.

Would you characterize your exposures to the public market, Philippe, as in the top third, middle third, or bottom third of your average exposures?

Brad, I knew you would ask me that, and I’m not going to answer that, but nice try.

I tried.

Nice try.

I tried. Okay, let’s shoot over to privates. I think one of the things that was a consistent theme, if you look at slides 60 and 61, is this idea that the private economy—Thomas, right?—we’ve had 3 or 4 years of really nobody getting out of the chutes. These companies have all stayed private. The percentage of unicorns as a percentage of the public markets has gone up, but now we’re starting to see an unlock here, both in terms of M&A and in terms of IPOs.

So talk us through the big themes from slides 60 and 61 today about how AI has reignited deployment and exits are starting to rebound.

Yeah, I’m curious to get Bill’s view here because he probably thinks about this as much as I do, and I’m curious whether he’ll draw the same conclusion. I think by and large we all agree that the environment of 2021 was incredibly unhealthy, both for companies and for LPs. Too much capital going in, not enough coming out—a kind of broken cycle, if you will.

You could see that in so many measures: the amount of dollars going in, no money coming out, historically low IPOs, even worse than post-financial crisis, which is kind of incredible to think about. So on almost any metric you looked at, we were kind of in the danger zone. I would say more or less that’s been true over the past 2 or 3 years.

This is the first year, and this is the crux of the view. I’m curious if you share where the signals are going from red to, I would say, yellow and potentially green. We’re seeing, first of all, a rebound in IPOs. We’re seeing IPOs perform better. We showed basically the performance of the cohorts and how they’ve improved substantially since 2021.

One of the data points that shocked me, looking at this again, is that the 2021 cohort within 1 year of going public was down 40%, and 5 years later is down 50%.

I mean, just pause on that for a second. That was a shocking slide: here we are 5 years after those companies went public, and basically the market has gone vertical.

That’s correct. On a relative basis, they’re probably down 75%.

You’re right. Slide 71.

But I didn’t believe this, Brad, so I actually went to look at every single company on this list. But this does not include SPACs, which is even more extraordinary. This is just traditional IPOs.

So it’s not dollar-weighted?

No, count.

Correct.

Yeah. So there’s a lot of scar tissue there. But I think we have signs that things are improving. We just talked about the IPO market. We’ve now seen some really strong IPOs that have performed well: CoreWeave and Circle.

Hey, Thomas, remind me—what does ZIRP stand for? Sorry to ask such a dumb question.

Zero interest rate policy.

Geez, that’s how little I know.

But we’re also seeing companies like—one of the things that really impressed me about CoreWeave, we had a slide on this. I can’t remember what the number is, but people are starting to understand how public markets think, and I do think they executed incredibly well on the timetable: how they released information and how they explained the business model. This is slide 75 for people at home.

So we have better IPOs that are being rewarded, and another thing that struck me is we looked at the cohort of IPOs, right? By and large, you can see, unsurprisingly, that growth and profitability, yielding a Rule of 40, was kind of the average of the cohort. I thought that was bullish for the ecosystem.

And then finally, you’ve talked about this on the pod before, but the M&A environment is coming back: different types of structures, Zuck’s bold move, right, to pay 100% of the value of a company to only get 49% of the company, buy the team. Urgency is now: “I need you tomorrow, Alex, to help me fix my business.” Right? I thought that was the best description of the Scale deal that I’ve heard.

And maybe just click on that again for a second. For the audience, most people know that Meta has done this interesting structured deal. They’re buying 49% of the company. They’re paying a $30 billion valuation, so they’re paying effectively $15 billion. They’re avoiding regulatory scrutiny. The CEO of Scale is going to help lead efforts at Meta, and all the customers have left, right? And so they’re leaving kind of a shell company behind.

So we don’t know if that avoids regulatory scrutiny.

Exactly. We’re going to find out. Right, right. I don’t know if there’s a breakup fee or not. It’d be interesting to see.

Yes. I don’t know that either.

But I mean, I think one of the things it shines a light on is the speed at which everything is moving. Here we are, and we can all say that Zuckerberg’s in beast mode. Meta is one of the greatest companies on the planet. He’s extraordinarily focused on getting AI talent.

But why do you think he was willing to pay 100% of the value of a company and only get 49%? Is it that the imperative to have talent today is so important because 2 years from now you may be so far behind, given the rate at which AI is moving?

I tend to think it’s related to 2 factors, right? One is the size of the prize. I think he clearly sees that this is the biggest prize in tech in the world, frankly. So I think relative to his $15 billion—to all of us, it’s a massive number—probably in the scale of the multitrillion-dollar opportunity that he sees, he might just think it’s a bet I would make all day long.

You look at it as a percentage of market cap and you say it’s like 1%, right?

Correct. Right. So I think that’s number 1: the scale of the opportunity, no pun intended. And I think number 2 is how quickly the ecosystem is moving.

There’s some data point—I mean, people had this view already that Llama wasn’t quite at the top, but this is somewhat confirmatory of that, that he’s fixing a problem, right? We’ve seen Anthropic. We have data in here that I think it took them about 1 year to get to their first billion in revenue. It took them 3 months to get to the next billion, and then it took them 2 months to get to the next billion after that, right?

So he’s probably seeing how quickly ChatGPT is growing social users, how quickly Anthropic is growing business users through their API, and thinking, “I don’t have 2 years to wait in European regulatory purgatory.”

I have a question for you, Thomas, on the IPO. Simultaneous with seeing more IPOs, which is awesome, there has been a trend for companies to stay private longer. I think the Collisons used to hint “maybe,” and now they’re more kind of “maybe never,” and some investors in the ecosystem are encouraging that behavior. What do you think is different about the people who choose to go out now that the window is, quote, “open”?

I think they each have different reasons. Some may just view it as a financing opportunity: the ability to tap the public market, both on the equity and the debt side, could be simpler, right, as a public company. I think that’s the big piece of it.

Second, look, it could be a brand-defining event for a company, right? For your product and your employees, it gives your customers transparency that you’re well-funded, that you have a fortress balance sheet, and that you can withstand the regulatory scrutiny that comes, as well as the scrutiny from investors. It shows that you have the discipline that comes with everything being public and people looking at your numbers.

I happen to believe that all these companies should go public. I also think, by the way, there’s a democratic element to it, where I think the wealth creation belongs in the public market. You attract different types of investors—not just public-market versus private-market investors, but also retail investors.

What can you learn from the retail investor, either positive or negative, about your business? I think it’s such an important point.

I made this case to everybody at OpenAI. I think they’re the most important company of the era. I think it’s hugely important from a regulatory-scrutiny standpoint and from the democratization of finance. It needs to be a public company.

The idea that we’re going to have companies worth $1 trillion and the only people who get to participate are the people sitting around this table—I just think that’s unhealthy for our capital markets. The fact of the matter is, we call these companies venture-backed companies, but we all know there’s a whole new market that’s evolved here that I call quasi-public. These are companies worth over $5 billion or $10 billion. They would have all been public 10 or 15 years ago. Why? Because the private markets just didn’t have the depth of capital to serve these companies and their voracious capital needs.

This is happening as we speak in private equity, right? Some private equity companies just go from one private-equity owner to another. Then you have continuation funds and big secondary transactions. This is happening in the private-credit market, where now you have huge private credit as an asset class. Not just that—this sort of healthy tension between public and private is important.

I just think that these super-large private companies, if you’re not willing to submit yourself to the sunshine and the ray of light of the public markets, you’re going to get it through a regulatory agency. So pick your poison, and be careful: if you think you can live in the public market purely to live in the shadows, that’s not going to work.

As you become a large company, you’ll be regulated. That’s maybe even more correct. That’s why I really hope these companies will choose to go public. You make the democratization point: retail investors should have access to these companies.

I just think, in general, the concept of mark-to-market isn’t perfect and there’s increased volatility, but every day we learn something, and every day we know it’s the price you can get today. Today, by the way, I thought one of our best speakers made this great point: just because I’m public doesn’t mean I need to change how I run my business.

7. Golden Age of Margin Expansion

Well, maybe we talk about AppLovin on slides 91 and 92. Slide 91 asks whether Microsoft has reached peak employees, and slide 92 was about how AppLovin has gone AI-first and had massive margin expansion, or revenue per employee. I tweeted about this the other day. I call it the golden age of margin expansion, right?

If you look at the Magnificent 7 over the last 3 or 4 years, they’ve grown over 20% compounded, but the number of employees—their opex—is growing at 2%. We’ve never seen this in the history of technology that we’ve covered. So why don’t you talk a little bit about it?

I loved this chart on 91. Previously, we just had the chart without the blue lines, right? For those listening, this chart tracks Microsoft’s employee count. What we realized after we did this chart is, wow, there are actually 3 distinct chapters being told here.

Chapter 1 is the ZIRP era. It’s COVID; software is everywhere. The only way these companies think they can grow is by hiring more people. So, reflexively: big opportunity, I’ve got to hire more. Which, by the way, made sense, because if you grow by producing more code, you need more people for more code. I think it was completely logical: we’ve got to hire more. So that’s the ZIRP era.

Then, ironically, just as GitHub Copilot comes in, you get the get-fit era. It’s like, “Hold on, we need to get fit. We’ve gotten too big.” Then you can see headcount stabilizing and coming down in a lot of other companies.

Now we’re entering the AI era, and I do think it’s kind of a provocative question: has Microsoft reached peak employees, and will it never cross that threshold ever again? I had a conversation with the CFO of a major company recently, and they said, “Thought experiment: what if our headcount was down 50% in 3 years?” Those questions have never been asked for companies that are growing and thriving.

I do think what I get excited about as a public-market investor, Philippe, is that it’s not just that we’re seeing a reacceleration in topline growth for all these companies. Every one of these companies, literally from Uber all the way to the largest of the Magnificent 7, is growing its topline without growing its headcount.

But AppLovin has done as good a job as any. Tell everyone about this slide you put together.

This is another one of my favorites, right? What this slide does is track AppLovin, a public company run by a brilliant, in my opinion, generational entrepreneur. It basically looks at 2 things. One, it looks at the company’s revenue, annualized since Q2 2021—that’s the blue line. Second, it looks at the employee count over that same period.

Basically, in 2021: big opportunity, I’ve got to hire tons of employees to try to capture it. What else can I do? Then it realizes, “Oh my God, my company’s gotten too big. I’ve lost control of my culture. We’re not innovating fast enough. There are too many layers of bureaucracy. We’re not set up to capture the opportunity.” It rightsizes the workforce.

At the same time as AI comes in, now the company’s lean and mean, innovates, outcompetes companies like Google and Meta, and doubles the size of the company as the employee count is down over 35%. Think about this: we just showed the slide of ChatGPT going parabolic, Google losing page views. Google has 187,000 employees; OpenAI has 2,700. We’re not going to be a company of 20,000 employees. He didn’t say we’re not going to be a company of 187,000 employees, right? He’s saying we’re going to leverage our models, our agents, our capabilities, which is exactly what Jensen Huang said to us last year. He said, “Brad, I’m going to 3× the company, and our headcount may not grow or may only grow a little bit.” And I said, “How?” He said, “Because I’m going to have agents who report to me. I’m not going to have employees who report to me.”

By the way, I’ll tell you what this made me think of. Back to the AppLovin slide: in 4 years, they doubled the revenue per employee, and now a company with a high growth rate that’s profitable and thriving is lowering headcount because of AI.

It really struck me that there’s a level of confidence in a company’s use of AI if they’re willing to actually reduce headcount. A lot of companies give lip service to using AI, but a willingness to reduce headcount is a different level.

One point Adam would make if he were here—and I think it’s important to state—is that he’s not doing this because he’s a masochist who loves to fire people, right? The reason he did this is he believed that’s the shape the company needed to be in to win and outcompete. I think that’s really important.

It’s not like, “Oh my gosh, all of a sudden I want to be much more efficient, and I think I can create so much more value.” It’s, “I believe this is what the company needs to look like so I can win this market. We need to make decisions faster. We need fewer layers,” right? I think the motivation is really important. This is just an output of that.

The final thing I would say about this, Philippe, is that the thing that should give us confidence about this productivity explosion in the economy is that, at the end of the day, our economic productivity is just a combination of all these companies. If a lot of companies are doing this and you pile them all together, you’re going to get more output for a fixed amount of labor and capital. That’s going to drive economic productivity.

8. Jevon’s Paradox

The last thing to say on that, which is really important, is someone is going to then say, “My God, what’s going to happen to employment?” If we have all these companies that become so efficient, right? Today, someone brought up the concept of Jevans paradox. I’m going to actually use my ChatGPT to study a little bit more over the next week or so.

But it is the concept that sometimes, as you have fewer employees and the cost of employment goes down, the unemployment rate will go down, not up. And I'm really summarizing it in terrible terms, but I think it's really important to say that it's possible that companies need less employment, but more companies get created because it's much easier to create a company. Smaller, vibrant companies get created, and jobs become more interesting.

And so I think there's going to be a big debate around, okay, all this AI, is it going to increase or reduce unemployment? I'm not 100% sure what's going to happen, but if you force me into an answer, I have faith that it might actually create more jobs—more interesting jobs with more responsibility—versus the other way around.

9. Insights for Founders and CEOs

Yeah, we have 2 more slides we want to cover, and I think maybe we're going to end with the best because you guys had a couple of powerful things. The first was slide 98, right? After all of this, covering what's happening in public and what's happening in venture, Thomas, I think you summed it up well, which is, okay, so what does this mean for me? If I'm a founder, if I'm a CEO, what does this mean for me or my company?

Bill, why don't you let me describe what Thomas did, and then Thomas, you can do the analysis from it. He created a quadrant, and on 1 axis he has growth rate above 25% or below 25%. On this axis, you have profitability: either you're cash-flow positive or you're not. So walk us through your recommendation for companies that find themselves in each of these 4 quadrants.

Yeah. And Philippe, chime in too. Look, we're very proud of the work that we put into this deck, but we also want to be mindful that it's a lot of data. We thought about how we could crystallize everything that we see in the market—from all the data and all the smart people that we talk to—in terms of generating useful advice for entrepreneurs, right? And so we came up with this matrix.

If you look at the left side, which is basically growing companies growing in excess of 25%, you might argue this is kind of the easiest bucket. You're growing 25%, but we do think the delta is kind of different. By the way, 1 thing we skipped over: you guys had 2 or 3 slides on the fact that growth has become more scarce in the public market. There's a big delta now in revenue multiples for growth and, obviously, diminishing multiples for lower-growth companies.

We have seen growth be re-rewarded in the public market post-2021. So our advice to entrepreneurs is that if you are growing over 25% and you are profitable, it's time to think about whether you should be public. But that doesn't necessarily mean going public, as you well know. There's a difference between being IPO-ready and going IPO, but we think certainly putting all the steps into place starts to make sense.

If you're burning, then now might be the time to build a fortress balance sheet. We just saw OpenAI raise $40 billion, right? These companies are accumulating massive war chests, so you don't want to lose out. It's time to really build up your strengths.

I think where you're going, Bill, and what you and I also spend a lot of time thinking about, is what about the companies that aren't growing 25%? For Philippe and myself, we take the responsibility of having invested in companies really seriously. We're on the boards of many of the companies in which we are invested, and we don't bail on our entrepreneurs when we make those commitments. So what do we do in those companies, right?

I think each bucket is interesting. The “I'm growing less than 25%, but I'm profitable” bucket is kind of an interesting case study because that's where you might be complacent. You might have said, “Look, I got fit post-2021. You told me to cut my burn. I'm profitable now.” And by the way, the reason I think a lot of companies ended up in these low-growth situations is that they had a ton of capital. We had that massive correction in 2021. Everybody said, “Get to cash-flow break-even.” They all ran that way, but that meant cutting headcount and cutting programs that they might have been doing. You end up in a low-growth situation.

So we thought that this bucket is actually in a potentially generational transformation and architecture shift because of AI. It's time to maybe look at it and say, okay, what can AI do for your business? Is there a new way that you can invest? Is there an M&A opportunity or something interesting? We think now you can afford to be a little bit more on your front foot. You've gotten the business healthy, and you've shown you can be profitable. We have a generational architecture shift, so it's time to see how we can play offense.

Would that even include maybe becoming unprofitable?

Potentially. If you have the signs and you really start to see the growth reaccelerate because of it, potentially. Absolutely. A lot of AI companies are not profitable right now, so if you think you can win and you can benefit, I think that makes sense.

This is probably the one I had the most debate about, both myself and with others: what to do if you're growing less than 25% and you're still burning capital. Obviously, no one chooses to be in this position. Circumstances of the business, whether it's competitive dynamics or something else, have put you in this position, and now the question is what to do. I went through a lot of different iterations here, and the best word I could come up with is: it's time to reinvent. And reinvent could mean a lot of different things.

Let me posit that you might have 2 businesses. Let's say you were at $50 million in revenue, and you might have your $40 million core business not really growing. The unit economics are tough, but maybe you've got an incumbent, maybe it's an on-premise product, and now you've incubated a new SaaS cloud product that's maybe only $1 million or $2 million in ARR, but it's really growing quickly. It's putting the company back on offense, and the team's really excited. It might be time to say, “Hey, let's go all in on this new product,” even though it's much smaller. That's 1 reinvention.

It might be that you have a gem of an asset. It might be trying to open-source something that previously you didn't, right? That's kind of what I mean by reinventing. It's the opportunity of looking at this moment and thinking, what can I do? And also realizing that you as an entrepreneur have an opportunity cost of not doing other things.

So the best word I could come up with is “reinvent.” It's going to mean different things to different people, but we thought now was the time to think about that.

I thought this was amazing. And I will tell you that I think one of the biggest challenges that these companies in this quadrant—and I think there's a lot of them; there may be 1,000 of these out there—have is having survived to this point and having succeeded. Let's say they have revenue of $50 million to $100 million. They feel like they need to protect something, and it puts them on the back foot, not the front foot. It makes them conservative.

And I like your word, “reinvent.” They need to increase risk. Actually, I think one of the problems is they don't internalize the fact that if they stay low-growth at this size, their multiple could go from 5 to 3 to 1 times revenue, right? And they're protecting something that doesn't exist.

So I'll leave you with this last thought. Brad, you and I have talked about this. There's an amazing element of the venture community: they tend to be tribal, and I think there are a lot of benefits to that. But I also think there are a lot of benefits to what I'll call more mercenary thinking, which is more reinventing from the ground up, right?

And I think that, ultimately, the combination of both of those—which tends to be more of a public mindset, again, because we do have the ability to sell—and venture, to us, bringing those 2 strains together in the boardroom can yield, hopefully, some good outcomes.

Awesome. Thomas, thank you for being with us.

Thank you for having us at the event.

Yeah. It's really incredible. The amount of thought that went into this is extraordinary. And I would just say, on behalf of all the founders, those people who partner with you like Altimeter and Benchmark, what I love about this ecosystem is that most people think that we compete like dogs, but the truth of the matter is, you're one of the first people I call—or Philippe—when we're trying to figure something out. And you guys do the same to us, and that's why Bill and I do this pod: because we actually just want to be smarter and get to the right answer.

So we appreciate you having us, and awesome job again. Okay, thank you so much.