Sam Altman 的总计划，还是送给 Anthropic 的礼物？Palantir 与 Shopify 业绩大胜

My big aha is that it's like dealing with a deranged madman trying to estimate what the street will do. I spend no time on this. It's utterly unknowable.

You don't need half your company, and Palantir and Shopify are proving it. You don't need half your company. Let's look at Shopify for a minute. At its peak, in 2022, Shopify had 11,600 employees. Since then, revenue has grown 91%—pretty impressive for a company at $11B in revenue—and employees have gone down from 11,600 to 8,100. They've gone down while revenue is up 91%.

He's ruthless. Zuck's ruthless. Karp's ruthless. And if you think you're going to win in B2B, if you're not ruthless, you're going to lose.

Guys, I am so excited for this. We have an amazing schedule in place today. I want to start with GPT-5, as it's the top story of the week. Consensus is that it's slightly underwhelming. Before I lead the witness, I'd love to hear how you responded to it. Do you agree it was underwhelming, and do you see it differently?

My first experience was certainly underwhelming when it said we had the greatest market crash since the Tulip Mania. But listen, that's what I think. When I look at folks smarter than me, if Aaron Levie is running this through Box and saying redlining, document comparison, and term extraction are materially better, maybe that doesn't make those of us who are using it for therapy excited. I get that.

But if he thinks it's materially better, and more importantly, everyone is talking their game—if it's materially better at coding and competes with Anthropic, that's $6B of revenue that they lost. Right? I get it. It does feel like it's a worse therapist at the moment, doesn't it?

I think underwhelming is great. Let me tell you what I mean by that. Underwhelming kind of took a little bit of the air out of the techno-optimists: we're on our way to AGI, it's all going to revolutionize everything, all that kind of noise. You definitely felt a little deflated here, right? Which is great, because we're now at the “it's a really great piece of software for doing business; let's make it better” stage of life, right?

It feels to me—I remember the first Windows 95: “Yay, it's amazing,” you got a big launch, and then there's just steady growth. Or even the iPhone, a better example: the early one, great launch, it's going to change everything, and then you settle into 10 or 15 years of incrementally making it better until you plateau, right? From a zero-to-one perspective, that's not as interesting. But we're now in the grind-it-out, make-it-better, build-a-business stage of life, which I think is a more normalized world, right?

Implicit in that is the statement, “I don't buy any of this.” They're going to keep on getting better, exponential takeoff, all that AGI rubbish. I've always assumed it's rubbish. Maybe I'm wrong, but at least right now, the evidence has shifted a little more in favor of perhaps not nearly as quickly as you think.

From a business perspective, there's a lot of interesting shit going on here. Jason nailed the first big one, right? They basically—and it's less from a “the product is good” perspective, though I believe 2 of my colleagues are more avid coders—not as good as the high-end Claude Code models, but a damn sight cheaper, right? You're talking 10x cheaper than the most expensive token load, based on what I submitted—8x to 10x.

On top of that, Cursor now is pushing a GPT-5-based product and has free demos of it. You know that they're pushing that on their user base. From a business perspective, aside from the pie-in-the-sky AGI stuff, this is exactly what Jason said: this is OpenAI going after a big-ass pile of revenue that Anthropic has.

Maybe Anthropic overplayed their hand a little bit by bullying Windsurf. If I'm now Cursor, a week ago—or maybe a month ago—I'm sitting there going, “My gross margins are set by my worst competitor, Anthropic.” And now I'm ecstatic, because the big-ass guy on the block is now another vendor of tokens, significantly cheaper. I'm going to push the hell out of this.

That's a really big business comment. It's not as sexy as AGI making us all unemployed, but if you're trying to build a business and you're Cursor, this is the best damn thing that ever happened, right? We have a competitive product at one-quarter to one-tenth the price, and I'm happy.

If I was Anthropic, I would be in that room rejoicing, and I think everyone I speak to on the Anthropic team is, bluntly, laughing, because the demos were terrible. I think it was entirely underwhelming in terms of the presentation of features. Yeah, model consolidation into one and model routing—and we've now taken that? Seriously, we waited for GPT-5 and you gave me a better model? This is disappointing.

There's a reason for it. Sam Altman, in his own way, is a marketing mastermind like Elon Musk. He knows what he's doing. We can make fun of this or that, but he's everywhere in the entire globe. He's doing Stargate, Farscape[?], everything. He's building everything. He is as perceptive as they get. He's dueling with Elon over who manages the X algorithm, right?

The fact that this would come out with a bit of a thud was known to him and the team, right? So there's a reason they did it. Was it timing pressure? Was it to put pressure on the team? I actually don't know. I haven't seen it on the hundreds of Reddits I'm on.

But this wasn't luck. This is like Figma—us saying Figma was dumb. They didn't understand the IPO pop from last week. Neither is true. The idea that it would come out with a thud is not a shock.

Put me down as not quite buying that, because I think—

I mean, as a comment on—

You can be directionally brilliant and entrepreneurially brilliant, and it's hard not to argue with some of it. But it's always worth remembering that the most popular product they ever shipped—they didn't know they were shipping a good product at the time. ChatGPT, which is 70% or 80% of the revenue dollars when they shipped it, wasn't even considered a major release. They were focused on GPT-4 the following March.

So I think one of the big a-ha's on that is, you see this in venture often: if you're directionally correct and you have big-ass vision and drive and the ability to raise money, sometimes that's enough, and you kind of get the breaks on your side. But I don't think he or anyone has complete clarity on every step and the level of forethought that you're implying, Jason, on that.

But I actually don't think I care. I want to go back to what Harry said, because I don't know the answer. It's interesting you made the comment that you think Anthropic are laughing. They're definitely laughing at the thud, and it's hard not to laugh at the funny, stupid graph that was mathematically wrong. Oh my God, they must have burned someone at the stake after that, literally.

But I do think I'm not sure I'd laugh quite as much, because I do think a renewed push on code and pricing and trying to get that business would be better. If you're Anthropic, you'd be better off being a monopolist with the best product than being in an oligopoly where you still have the best product.

You're right: the consensus is that, token for token, dollar for dollar, the Anthropic products do more and are more efficient. But this gets into a very interesting discussion. If the pricing of the other product is low and good enough for companies really wrestling with gross margin, you're going to have some kind of, “Use the cheap shit where you can and use the dear stuff where you have to,” right?

No one ever said, “I'm really delighted my bigger competitor entered my market or product 5x cheaper than us.” I don't buy it at that level, Harry, to be really direct.

Well, but I think that's a very moment-in-time perspective. Everyone is aware that if you're looking at the cost of tokens today, we're seeing them reduce so dramatically within a 12-month period. I don't think Anthropic are looking at that thinking there's a permanent chasm in pricing and we're going to be unable to match it.

I think they're looking and going, “Fine, they might have a slight efficiency on us today, but if our models are better, we'll win.” Yes, you'll win, but it just won't be quite as easy, is my point. We've gone from the pie in the sky for everybody—it's going to be amazing, flowers and sunshine—to, okay, it's a slug-it-out war here. This is the advantage we've got; this is the advantage they've got.

I just think we've gone to dealing with business fundamentals. Are you better off with the slightly better, 4x more expensive product? That kind of discussion. Roy, what was your second point? I interrupted you.

No problem. There are 2 or 3 things you have to take seriously. I've spent less time on this, so I'm talking, but they shipped the open-source models earlier this week—oh, not last week, of course, now. It was interesting after what was 2 or 3 years—2 years plus—of no open source.

I don't have as good a handle on the motivation there. Is it to mess with Elon because he's saying they're not open source? Is it because they genuinely feel that's a feeder for their wider models, which someone articulated? But at the very least—and my point is—it's another interesting commercial move.

So, again, in the last one, you were sneering at it, but moving away from all those models to the single model selector.

Yet it didn't work at OpenAI, right? But again, it's the kind of thing you do when you're done with imagining being the next Robert Oppenheimer and now you're focused on what the product manager tells you will get 5% extra conversion among users, right? Someone in product marketing said, “Dude, we have to stop with these 6 different model names that make our users feel stupid and just make it simpler to use.” All those things are okay. It's time to get business-savvy, not just AI-is-coming savvy.

So I thought, yeah, that's what grinding it out looks like. That's what every iPhone release after the first 1 or 2 looked like. Similar with every Microsoft Windows release after the first 2 or 3. Once you had tiling in Windows, the next 15 years were just grinding it out. Same thing with my iPhone. Maybe it's the same thing here. We've got the idea: it's a chatbot with AI, and now we're just going to be grinding here.

I think GPT-5 didn't achieve what Sam wanted it to achieve by July 31st. And you've got to manage the team, even at the highest-profile company on planet Earth. Everyone that's been a founder that's shipped software has had to make a decision: Do I give the team another month, another quarter? Anytime you do a release and it gets too complicated, your team always wants more time. They always say, “I need another week, I need another month.” At some point, as a leader, you've got to ship it.

Earlier on this show, we had a countdown on when GPT-5 would ship, right? It was all over the place. I think Sam said, “August is time, boys and girls,” and if it's going to be underwhelming, I'm going to make the call that we're not going to wait for Nirvana. I don't have a firm date for when it was supposed to be, and I'm just going to box it up and ship it. We're going to make some basic stuff better for Aaron Levie, we're going to cut some costs, and that's it. This is maybe 4.9, right, or 5.00. But this is just my theory: there's so much attrition in this industry, and so many resources flowing back and forth to Meta, to here, to there, that I think he just decided, “Ship it.”

I also think aligning to a $500 billion fundraise is actually better—to get it out, to continue to press on—than to have continuous waiting.

Yeah. It's like the Replit $3 billion round. Just get it done. Rory, would you do OpenAI at $500 billion? And do you feel more or less confident post-GPT-5 about OpenAI?

I'm going to say more confident, and I'll tell you why. It goes back to what I said earlier, right? I think the grandiosity was totally necessary at the start, because if you'd walked into people and said, “We need to raise $30 billion or $40 billion, and at the end of it we're going to ship this chat app that, trust me, people will use as a replacement for search,” that might not work as well as, “We're going to change the known universe. We're going to exponentially eliminate all labor.” You've got to be selling dreams when you have to raise $30 billion on nothing. You're selling the future of humanity.

We're now at the “I've got to make a business, converge roughly on profitability over the next 3 to 5 years” stage. So just go run the business. It's a combination of just ship, run the business, make it better—the kind of next step. Yeah, I feel like—and the interesting thing is, I'm going to say this without any need for AGI or any of that rubbish, right? I think both of these companies can be worth plus or minus $500 billion to $1 trillion.

I think the opportunity is just so obvious that everyone's going to be paying $20 a month, and people are going to be paying more. You get an interesting discussion about how big the high end will be, which matters more for Anthropic than these guys, but they have the mass market: everybody paying $20 a month. That's a big-ass company, and the runway is such that you probably can see a 2–3x just on that. Probably some noise along the way; you might be ahead of yourself, but yeah, I feel their destiny is now locked in for them, which is to be the go-to place for information and, you know, for pretty much everyone on the planet.

I always think that AI is a bit like crypto, which is like, “Hey, don't fucking worry about all the underlying assets. You'll make money if you just invest in the core, which is Bitcoin.” And I'm like, the core unwavering asset here is NVIDIA. Do we think OpenAI or Anthropic or Grok, or you name it, are going to win? I don't freaking know. With each month, the benchmarks change. Humanity's Last Exam changes. The next humanity's test will come. NVIDIA—they all just need a lot of NVIDIA.

What you're saying is true in some ways: sometimes, you just keep the main thing the main thing. I often make that mistake because you try and—what is—this is going to sound really cynical, but early on you listen to all the promotion and all the crazy stories, and I'm very much a rationalist. I try and take them seriously and deconstruct them. I realize, “Oh, these crazy stories are bullshit; therefore, the thing itself is bullshit,” and that's wrong, right?

All the stories in every single thing that people said they'd use Bitcoin for in 2013, except maybe stablecoin payments, have been bullshit. But the thing itself totally worked because it's got this store-of-value idea—the asset independent from everything else. Bitcoin worked. Everything else—ETH to some extent, everybody else, all those little weird coins—didn't matter. Same thing here. The main thing is the main thing. You want to make the chips; you want to be the company that's, in OpenAI's case, pretty much selling to every consumer. Looks like its market share is pretty hard to change at this stage.

So you start multiplying: you'll have a paid tier of $20 times a lot of people. You'll at some point have an advertising tier, and plus or minus, you're Google with a subscription business. There you are: it's $1 trillion to $2 trillion. Thanks for coming. Keep it simple. Don't overthink it. They get hard. Maybe I'm not saying it clearly. I am often guilty of overthinking to the point where I've now learned to watch myself when I overthink. Sometimes—

See, I'm—

Yeah.

I'm very lucky. I don't have the luxury—or the ability—to overthink, Rory. So I'm always a simplistic thinker, which tends to do me quite well. You know this kind of internet meme graph where you have, on the left, a total fucking idiot; in the middle, someone overthinking it; and on the right, not thinking at all, just like the other guy, and you win.

Listen, another element that's just crazy is Perplexity trying to buy Chrome for $34.5 billion. I suppose Aravind said before, “Legit, it's in their court.” How on earth is—

Where is he getting the capital, just to confirm?

Well, that was my question to you.

From Elon. Elon had his capital secured a while back, didn't he?

How is this possible, guys? What did you make of this when you read it?

One, you can get tactical and say, “Where would the money come from?” But now, stepping back, why would Google sell Chrome? Just big-picture points, to cover it for people: why would Google sell Chrome? Answer: they don't want to in a million years. The Department of Justice says they have to, and if they can't appeal it at some point, maybe they have to. They don't want to sell this thing, right?

The second question that's interesting is, what's the Chrome business worth? And the totally interesting thing about that is it's so dependent on who buys it, right? Because the asset itself—Mozilla has around $700 million, I think, of revenue, right? Reasonable market share. Yeah, you don't get paid for owning a browser, to state the obvious, right?

And the reason the DOJ is going on this is, you know, as well as owning Chrome—which was one of Sundar's crowning achievements that got him the big job—Google also pays Apple for search placement on Safari, $20 billion a year. The Department of Justice is looking at those 2 things together and saying, “Oh, paying to be in the browser is bad; owning the browser is bad. It allows you to continue your search monopoly.”

Right. But the question is, owning—so from Apple's perspective, owning Safari on the iPhone, where it's kind of a pain in the ass to change, is clearly worth $20 billion of profit a year, right? Is owning Chrome—how much is owning Chrome worth, right? In other words, how locked in—who would buy it? What would they sell through it? And how locked in would the service be?

Ironically, the person for whom Chrome would be the most valuable, if it wasn't owned by Google, would be Google. In other words, if someone else owned Chrome, the highest person to monetize Chrome is Google because they have a search business. So they would probably say, “Harry, if you bought Chrome, you have no revenue source, but we will happily pay you $20 billion a year for all your search, and you'll be, you know, the richest single-person company ever.”

Right? So that's a kind of weird asset where intrinsically, in itself, it has no value, but it's kind of a gateway to some product that does, which now gets to Perplexity. The thought process there is they're already building a browser. If they had Chrome, they could basically be the AI backend instead of having to build their business browser by browser. They would basically jam Perplexity into every Chrome user, and every person who's using Chrome today, including me, would have some version of Perplexity as a default AI engine, right? Great work if you can get it, right?

And I totally get it now. Is that worth $38 billion? Is it worth more? Is it worth less? Do they have the money? All those things are unclear and may not matter yet. To Jason’s point, is it real? But would an AI engine competitor to ChatGPT absolutely kill to own the Chrome user base? Absolutely. So that’s a long-winded answer, but that’s kind of how you get there.

You know, for what it’s worth, do you think it’s likely in any way?

I assume Google will litigate this to the end of human time. Which actually gets to step back and say something else: I’ve just got to get it off my chest, right? The Department of Justice gets around to killing American companies and technology just when they’ve become irrelevant. Now, irrelevant is the wrong word—just when they’re at risk.

The idea is, Google—think about it. The whole idea here is that Google is so powerful and almighty that we’ve got to smash its search engine monopoly with legal remedies, while all of us here are talking about whether Google could be screwed because OpenAI’s ChatGPT could be a better product. It’s literally not kicking Google in the nuts when they should have, right? When they were evil bastards for the last 15 years, grinding everyone down like Yelp.

And then finally, when their backs are to the wall, now we’re going to kick them, just too late, when in fact they’re now at risk. I forgot to mention that, at a zoomed-out level, all this is freaking stupid. It’s the Department of Justice, just because they take so long to make decisions. It’s like they only started kicking Microsoft when Google was in the ascendancy and Microsoft was on the defensive. It’s the same thing here.

Leave poor Google alone. They’re just a simple, humble trillion-dollar company trying to survive. So dumb as rocks. Sorry.

Maybe just 2 things. One, sometimes I’m a little slow on marketing, but there’s a huge need in AI for constant marketing. Constant marketing. And there’s a huge need to be one of the top 2 players, right? Everyone’s doing a lot of marketing. Lovable, Project Lovable—why are you doing these hackathons? Why is Bolt doing a hackathon in every city? Why is Sam Altman everywhere? Why are Anthropic’s founders everywhere?

I’m not even sure, even with Nvidia, Jensen’s everywhere, right? And he probably doesn’t need to be. But I think it’s important for Perplexity to be in that conversation, because we can think of a lot of tools that aren’t in the conversation. We talked the last couple of weeks about Cognition buying Windsurf, right? At least it was in the conversation. I mean, 2 of my portfolio companies use Devin, their AI tool, and they love it for niche use cases, but we never talk about it.

I don’t think it’s cynical. I mean, it’s a little bit cynical. But even if he had $38 billion, I’m guessing another suitor might come out of the woodwork. I think Satya would raise his hand and say, “I’ve got $39 billion.” There’ll be a few other folks.

To grow at the pace the AI leaders are growing, at this incredible pace, you do need to fuel the marketing engine. You need to fuel it aggressively, and you’re not going to do it on AdWords, a few sponsored blog posts, and SEO. You absolutely have to.

I actually interviewed Aravind at an event, and I said, “To what extent do you feel like a politician or a state leader, where you actually don’t do anything in the machine, but you would just entirely speak for the machine?” And he said, “That is completely my job. And that is the job of Dario. That’s the job of Sam. That’s the job of me. We are on the machine, shouting.”

And then you also look at the bluntly poor performers—and I hate to name people, but this is the show that we have—like Cohere and Mistral. You also correlate that with who does next to no marketing or speaking publicly. Cohere and Mistral—I have no idea whether their models are good, not good, or whatever, but they are both incredibly quiet and don’t do press, and they are also the ones which people have deemed the losers.

You do have to do it. Virality alone isn’t going to do it. The other captain-obvious point is, where does Google stand today? Is search under massive existential threat? Yes, but search revenue is up. It’s confusing, right?

I think the captain-obvious theme is that, no matter what happens, Chrome ends up being this accidental gem. Chrome was built very strategically in the old days, in the desktop-focused days, to counter IE, which it destroyed, right? And then it’s open source, so Safari and Opera and everything are built on Chromium. They gave away the open-source version, they destroyed everything, and then it kind of plateaued, like a lot of software did.

We had plug-ins, we had Chrome, we took it for granted. Now it’s the crown jewel in the age of AI. It’s the crown jewel again. It’s the return of the browser. It’s not just the browser wars; it’s the return of the browser being this core piece of software that we forgot about for the better part of a decade.

First of all, agreed. And I can’t resist that “Chrome was built in the day” statement, but moving right along. You’re right, and it’s only because, for 10 years, Google won in the sense that no one else could monetize traffic anything like the dollars they could. So why bother building a browser?

I mean, you could have these niche browsers, but you couldn’t build a humongous business, because the only thing you could do with that traffic was monetize it, and the only place to monetize it at scale was Google. The interesting thing now, to your point, Jason, is that you can envisage a world where someone monetizes Chrome traffic via an ad-supported product.

But there’s also ChatGPT. It’s an astonishing consumer subscription product. What is it—700 million free users, 20 or 30 million paying consumer subscribers? I think those are the numbers. So that’s a very profitable business where the rate-limiting constraint looks like your ability to sign people up.

That’s why Perplexity is smart. If you could get a billion people coming through your door, and if a billion of them were free users and you had the same conversion rate as ChatGPT does—which is a leap, obviously; I doubt you would—but any kind of conversion, what’s it, 20 over 700? That’s about—yeah, it’s about 2% plus, a little under 2%, which is what you see with them.

If you get 1% or 2% converting into paid users, it’s a compelling business. So suddenly Chrome matters not because it got better, not because the browser is uniquely different, but because you can plug it into a machine now that can collect checks.

You know what? We’re not talking about AI for normies. And this sounds incredibly condescending and patronizing, but AI for normal people—and normal people are kind of the slow laggards to adopt. And what I mean by that is—

For the record, “normal people” is not patronizing, but “slow laggards” is. So you might want to work on that. I’m willing to be normal. I prefer not to be slow, but keep going, Harry. Be a joke.

Fuck it. I’ll keep going with this. A great example of that is Threads, something that we all deem in our circles to be, like, “Who the fuck uses Threads?” Threads is actually a massive success, a massive hit.

The reason I’m going there is because, yes, we all think ChatGPT is great and a massive hit, and 700 million is amazing. And yes, it is unbelievably incredible and insane. I’m not doing a prior guest a disservice by saying it’s not, but Gemini is great, actually, and it’s getting better and better and better. They will plug it into the core engine and the core distribution channel, and actually, for AI for normies, there’s a real chance that they don’t lose any market share.

You’re right: for a good slug of the population, getting an AI answer—there are going to be times when an AI answer meets your need on Google and you’re done. There’s going to be—

Yeah, Gemini as a model is actually apparently pretty damn good, even for coding. Someone was walking me through it. So, yeah, they’re going to have market share. They’re not going to roll over and die, and the search revenues will go up.

It’s simply that you’ve gone from a situation where you’re the only game in town to a situation where these other guys are taking a significant slug. It’s always better to be a monopolist than an oligopolist, but it’s still pretty okay to be an oligopoly.

I do want to move to an insanely hot deal, guys, that really in Europe was a talking point for everyone, and it’s n8n. It was done at $3 billion. Reportedly, it was at $40 million ARR, ending the year at $80 million. Accel led the round. Massive.

The last round was not too long ago—$300 million or so, give or take. It’s not a new company. It’s a 2019 company. How did you think about this?

With envy. Let’s start with that, because in fact we had talked to them in about 2022 or 2023. You’re right: what we clearly underestimated was this whole workflow automation category. It’s a category pre-AI. We looked at a lot of players. It’s a noisy one.

You have everything from RPA to process mining to low-code/no-code, and this big, indigestible mass of things where every solution blurs into the next one. It was just hard to know what was what, and therefore hard to build a big, compelling business, especially with lots of competition.

Clearly, they did a brilliant job of coattailing onto AI, because when you go from automating workflow for people in a very deterministic way to actually getting more of the work done using AI, the value proposition of your software goes way up. Instead of saying, “We’re going to automate a little bit of shit,” now we’re going to literally do the work, and you can get rid of all these people who were doing the boring work.

They clearly did an excellent job of that, and in a space of 6 or 9 months just massively accelerated. Well done to them. It just shows you’ve got to stay on top of these things. Good on the guys. Highland Europe, I think, did the last round. Good on them.

As I’ve gone into vibe coding, I’ve reused Zapier so much at the low end because I’ve got to hook stuff up without wanting to code it. If n8n is a more developer-oriented version of that, I’m having a Zapier renaissance.

Today, with the explosion of applications we’re building and the explosion of things we want to connect with AI, it’s an order of magnitude bigger. It’s an order of magnitude bigger. Maybe, Rory, when you met them, you couldn’t predict the accelerant that would happen. It was hard to predict, and it’s probably in the last 7 months, right, from the revenue. And then, boom.

That’s the aha. There are companies that you’ve tracked for a while that you mentally might be writing off. What you can do is either decide it’s not knowable, bet randomly, or try to track them just based on metrics.

The most logical thing is to sit down and say, “What areas will the model significantly impact in the next 6, 12, or 18 months, and what are the companies that could benefit from that?” The second criterion—and you’ve talked about this before, Jason—is that it’s all very well to say there were probably 10 related workflow automation-type companies. Which one of them will get the prize?

I tell you, it always comes down to the founder who gets it the most, gets it the quickest, and just puts everyone in the room and says, “No one’s leaving until we’re shipping an LLM-enabled version of this, and we’re going to get it in front of 20 customers by Friday.” In retrospect, that was the aha here.

They clearly had those elements, and if you tracked it, you’d have made a lot of money here.

The thing that’s really stuck out to me is Index. Index are making everyone feel like shit right now. In particular, in Europe, Sequoia and Accel have to win; otherwise, the gap widens more and more and more. They won Lovable at any price, and now they’re winning n8n at, I think, any price. I think it’s a well-priced deal.

But they did that at Facebook back in the day, when they were out of the game. They bid—I think, what did they pay for Facebook, Rory? Maybe $200 million or something, $300 million to $500 million pre?

$500 million pre.

They were insane. People said Accel was washed up, and they had to bid up this fledgling social network.

I’m actually praising them. I’m not saying—

I think you’re saying the same thing. And look, if you pick right, no price is wrong.

Yeah.

Absolutely.

But I do think there is a retaliatory element from Accel of, “Oh, shit, we have to.”

You know your European market better than me, and maybe because it’s smaller. Humans are human, so that kind of dynamic is a bad way to try and make money. It’s hard because we’re all prone to regret and FOMO, to making decisions and looking back at them. But you’d like to think you just make every decision on the basis of: is it a good bet or not?

I don’t know what you think, Rory, but this is what I was taught. I’m not a growth investor. If you’re writing one of these checks and you believe it’s a winner and see a clear path to 5x, and you can deploy enough capital, you do the deal—not necessarily at a certain moment in time. If they believe this, I mean, 5x with dilution is a lot. It’s got to be worth a $20 billion company, right? Lovable’s got to be $20 billion. They’ve all got to be $20 billion.

But if you genuinely believe it and you overpay, it keeps you in the game, and it’s for your brand. It’s not your whole fund, and it might not be worth the bet if you genuinely see a 5x. The growth rounds don’t all have to be 10x or 50x, right?

That is definitely true.

It’s why you overpay to get into a good deal where you’re not in a space, but it’s not insanity if you see your way to 5x. It just might not be a 5x fund returner if that investment makes 5x in a growth fund.

Well, there are actually 2 things to unpack there. One is the 5x versus fund return. You’re exactly right. Growth investors typically, when they’re looking at something, can say 3x to 5x. The interesting thing is, they’ll say—and most of the time, it’s hard to go beyond that. Then, by virtue of the power law, 1 in every 10 does it. It’s not dissimilar, in a very fractal-y way, to seed, but when they’re much bigger, it’s harder to envisage, right?

But the truth—I remember a Facebook investor, an early Facebook investor, said, “You know, they’d bid on another deal, they’d lost it, and they said they could see a 3x to 5x from here,” and obviously they made 100 times the money. So you basically just operate on the power law. Some of them turn into amazing outcomes, but you want to be able to say your base case is a 3x to 5x. That’s absolutely the business they’re in.

You went from there, halfway through the paragraph, into some kind of “we should do it for the brand” or “we need to be relevant.” At that point, what you’re really saying is, “I don’t see a 3x to 5x, but I’m just going to convince myself to do it because I’ve got to be relevant.”

The sad thing is, sometimes that might be the right thing to do. I recoil against it, perhaps wrongly, to be perfectly honest. Maybe there are times when you should just do that, but when you’re writing it, it feels like—you know, I remember when I was a VC, they taught you in Catholic school about sin, and they’d say you start on little things and then it will just get worse and worse. Once you start doing deals not for return, does it just get worse and worse forever?

So I don’t love the “I won’t make my return, but it’s good for marketing” school of investing. But I also recognize that sometimes you can say, “I can squint and get a 3x, and it’ll be good.” And hey, sometimes you do what you do.

Do you know the funny thing? We have a YOLO segment of our fund, which is a tiny pool of cash, whatever, $2 million or $3 million. Basically, we have amazing access because of the shows, and people will give us positions at high prices, but we get into amazing names, and it’s good for marketing and brand. It’s 7x as a pool right now, which is crazy, but it makes sense.

When we did it, we were like, “This is total brand.” We never expected—

No, and I think that—

Yeah, but then as you get later into the fund, you’re like, “Well, so what? I got 12x.” You’ll make it, I believe it, right? But it returns 5% of the fund. That’s the problem.

And I think the other problem with that is the Howard Marks quote, which is—you know, I’ve used it before—that circumstances change and people don’t. What works in one market doesn’t work in another, right? That same strategy, when you look back in 2021, felt really silly. In 2022 and 2023, it felt like, well, that wasn’t the right strategy; be more careful. And then, obviously, whenever you’re accompanying a boom and you’re leaning in, leaning in is the right strategy, and it’ll be the right strategy until pricing turns.

You know what? It has this super-narrow niche. One small advantage of that strategy, though, thinking back, because I accidentally did some of it pre-AI, is that it can get you early DPI in a fund. I accidentally did some of that, thinking back, and from my 2017 fund, I got about 20% of it back early from these early exits that had a nominally large value.

It doesn’t matter at the end of the day so much, but you’re going to hold some of your early stuff—some of it may take 15, 20, 35, 40 years to get liquid these days. So it’s nice if you can get that early DPI, even if it doesn’t matter too much in the long run.

Well, Jason, listen. You and I are both fortunate to have Horsley Bridge, who I think are one of the best of the best, and I love them dearly. But they’ve taught me, really, that venture is a very challenging asset class unless you take advantage of very small windows of hyper-liquidity.

Where I think about this, I’m absolutely aligned with you: I’m leaning in, being very cognizant of when to lean out, and being very aware that this will fall off a cliff, but I need to time it well.

Yeah, I buy that.

I agree with that.

Yeah. But if you want to be a little cynical about venture, I do think, depending on where you are in your career and in your earlier funds, it is like—everyone on X talks about DPI, DPI, but you don’t want to sell your winners early. We’ve had that conversation too many times.

And Horsley, you know, I had a conversation with Horsley. I had a winner; I had a chance to sell for $1 billion, and they were like, “Don’t sell it. Don’t sell it at all. It’s not enough of a return,” right?

But it is nice if you go out to raise your next fund and you’ve got some DPI, right? So if Harry’s little sliver is at 10x—and again, it’s 10% or 20% of the fund—and you go out and it shows that this current fund is making progress versus all markups and no cash, it is cynical, right?

But fundraising is just like founder fundraising, for folks—for founders who are watching and listening. You’ve got to put some points on the board, right? So it’s not bad to return some of your fund at a nominally very high IRR and multiple, even if the absolute number, if you squint, you’re like, “You know, that was only 11% of all that; together, that’s only 11% of the fund, let alone a 1x of the fund.” I think it helps.

It also does amazing things for relationship-building with founders. When I look at Mercor, when I look at Perplexity—honestly, Aravind and I chat a lot—would we chat a lot if I hadn’t put a tiny check into the company? I don’t think so. Not as much. There’s no tie that’s enduring.

So I think it actually does outweigh. Moving on, you guys are going to be exceptional on this, and it was a nuts week in terms of earnings in many respects. I want to start on Datadog, because wow: best net-new ARR quarter in company history, $260 million ARR in the quarter. Initial reaction: pretty positive, and the stock is down 10% today and lower than before results. I just couldn’t get my head around this, guys. Honestly, it seemed great, and the market puked.

One of the things that, when you talk to CEOs of public companies—and you should do this, other than the obvious major beat, major miss, and we’ll talk about one in a second—for all the middling things, if you talk to a lot of CEOs... I remember we used to play this fun game on some of the boards where you’d know your numbers, and you wouldn’t know the market’s reaction to those numbers, and you’d try to speculate. At least half the time, you’re wrong.

It’s actually a very fun game to play on a public board. You have the earnings call internally, the internal thing. You look at the numbers and you say, “Okay, the stock’s at $32. When we announce these numbers, up or down and by how much?” The error rate is massive.

In other words, perfect information doesn't tell you what the market will do. It's incredibly hard. Other than the obvious—when you miss by 20% or beat by 30%—it's pretty obvious what happens. But in the middle, it's very hit-or-miss, which is the odd thing. I just offer that as a kind of preliminary comment.

I remember watching my CEOs internalize that, so I've actually stopped worrying about it. I've really internalized it because, remember, you know what you did relative to your budget. So you know how you feel as a board member or CEO running a company.

They're comparing your numbers to 2 things. One, their internal estimate of what they thought you were going to do, which you have no visibility into. And then, even more zanily, to use Keynes's quote, what they think everyone else thought you were going to do, to try and figure out what happens. If you look at Datadog, the stock went up in the aftermarket, right? In other words, people looked at it and said, “Oh, this is good. Stock up.” The next day, it's like, “Oh, it's not good. Go down.”

My big aha is that it's like dealing with a deranged madman, trying to estimate what the Street will do. I spend no time on this. It's utterly unknowable.

So there are folks that are directly benefiting from the AI boom. They're selling AI products, like n8n and others. Then there are folks that are benefiting because AI is exploding. Apparently, OpenAI pays Datadog $240 million a year.

I'll give you a very small example. Harry and I are both investors in a company called RevenueCat, which powers 40% of all mobile subscriptions. Its usage has already doubled this year from where it ended last year. Even though it's not an AI company—neither of these are AI companies—RevenueCat is more extreme, but all the AI guys are using its product.

So it's very exciting, but it also has concentration risk because there are only so many of these large players, right? There's pricing pressure, and OpenAI has already said they're going to renegotiate down the Datadog deal, as they should. There are folks that are exploding not because they're AI, but because of the AI economy. I'm not sure that's why it's up, but it's one of the reasons it's up.

Obviously, that's a lot of incremental revenue for Datadog. A $240 million-a-year customer is high, but the incremental revenue and spend is from AI, right? If you're a B2B company, you've got to go get AI revenue, because you're not getting it from John Deere and the rest. They're spending the same—in fact, probably a little less.

No, you're exactly right, because that's the meta point, zooming out from this. You've seen the data: almost the entire GDP growth is AI capex. So if you can co-attach, even if you're not AI capex, if you're not Nvidia, if you're not directly in AI, and you can just co-attach to the money, you're going to get a pop.

The spend is happening there. That's where all the capex is going. I always think of what the bill of materials is to build AI. If I wake up tomorrow morning and say, “I want to build safe superintelligence. I want to build one of those LLM models,” you just make a list of the things you need, right?

It starts with the basics: you need chips, you need a data center, you need power, and you need some coders. Then you're right, it's all the little things. I need to keep this up and running. I probably need Datadog. I need all the other software tools to make it happen.

If you just co-attach to that, it's the biggest single thing. It's showing up now. The funny thing is, the productivity is not showing up. The revenue is not showing up at the app level, but the capex is showing up in the GDP numbers. It's so big, right?

You're exactly right. If you can get your piece of that, I'd prefer to have it than not. People are saying, “Oh, it's a bad $120 million concentrated-risk customer.” It's a downside, but it's better than not having that $120 million customer.

But I'm sorry—is this a gift or a curse? Forgive me for my naivety, but the concentrated risk of $260 million—if OpenAI continues its trajectory, it'll be $520 million in a year's time.

Yeah, whatever. It's a gift. Yes, it's a gift. We agree. I think both of us are saying the same thing, Harry, which is that the people who are worrying about it are getting into the second-order derivative thing. The people trying to value the stock aren't just saying, “We signed a $100 million ARR deal with the most exciting company on the planet. We killed it this quarter. It's freaking amazing. End of conversation.”

No one is going to say, “Oh, I'm really worried that they might reduce the price in 2 years. That's bad.” They'd be like, “Dude, we're so happy,” right? As a board member, I would be too.

Wall Street, as I say, has this different role. It's not just guessing what you are now, but guessing how you're going to be compared to what they thought you were going to be. They're paid to second-guess themselves. My best advice is to ignore it.

Maybe in 2 years' time you'll be dealing with a renegotiation. But as you point out, maybe you'll be dealing with the fact that OpenAI is growing so quickly that the contract has doubled, and they don't have the engineers to waste. I'm kind of envying you, and you just take the check.

Are we just seeing big-ish tech win? Should you just invest in big-ish tech? AppLovin has crushed it, HubSpot has gained, Shopify is ripping, and Palantir is obviously ripping. These aren't quite the Mag 7. They're great companies. Is it not just big-ish tech? The whole AI wave is mega.

That's true. I'd say there are maybe 2 things going on. One is, yes, the AI wave is mega. I think you definitely have Palantir, and probably Datadog, benefiting from that—not so much HubSpot and Shopify.

Then the second thing is, rather than this whole “Oh my God, SaaS is dead” thing, I think what you're saying is correct: these markets are big and huge, and if you're the winner, you're public, you've got scale, and you execute well, you can add a little AI, defend your position, grow nicely, and kick off a ton of cash.

I wouldn't like to be a little startup trying to enter the HubSpot space or the Shopify space. The death of SaaS has been overdone for the public-market winners, which is a very different thing from saying you'd want to invest in “the next Shopify” at $1 million pre, at $1 million in revenue.

These guys have clear market-leadership positions. In every case, they have strong founder CEOs, profitability, and the growth reacceleration this quarter has been pretty real across the board. As I said, I'd put HubSpot in the category of using AI, but not benefiting directly as much.

Datadog is in the middle because it got a big-ass contract from the biggest company in AI, so it kind of co-attached to that. And then, obviously, Palantir is white-hot in terms of its AI story. It has brilliantly become the way large corporate America implements AI at scale. It's a beautiful position.

Can you guys help me on Palantir? I always look at it and think, “Amazing company, but oh my God, look at it. It's so overpriced.” Every time I do that, I'm proved wrong and it goes again. How do you think about that? And legitimately, how would you advise me? I love these shows because I learn from you both.

I mean, the growth is freaking breathtaking, right? It goes from 12% growth at about $2 billion in revenue in 2023 to almost 45% growth at $4 billion in ARR. Goodness. Forget about the multiple, right? The secular fuel is huge. The contracts are getting bigger, with a record number of $5 million-and-up contracts.

The fact that they are the AI solution for both commercial and military markets is remarkable. If I didn't even know how to spell Palantir or what it did—which I think was true of most people until 12 months ago—when you see it go from 12% growth for a public company at billions in revenue in 2023 to almost 50% today, we've never—I mean, maybe Rory can come up with an example—I don't think in enterprise software we've ever seen that level of reacceleration, ever.

You can either say it's going to decay, like all curves do, or you can say, “Good God, this is my ten-bagger.” Everyone in the public markets is talking about ten-baggers. That's what my social media is full of: “What's your ten-bagger, Rory? What's your ten-bagger?”

Yeah, no, I look across 20 years. Only 1 in 3 companies reaccelerates for 1 year, and about 1 in 9 or 1 in 10 reaccelerates for 2 years. So that's just the data.

That's daunting.

These guys have reaccelerated significantly at scale for 2-plus years already. So I can see how you can build a model that goes from 12% to 23% to 45%. You're right—the next question is, do you fill in 55%, or do you fill in decay at that point in time?

The problem is that any exponential curve going up can justify almost any valuation. My gut, Harry, is that your instincts are right: it's an amazing company smack in the middle of 2 huge trends. We'll come back to that in a second. At the same time, 120 times revenue, plus or minus, is probably not sustainable. I saw the data.

It looks like it was a good statistic. Someone says 5 years of 40% to 50% growth, and then they'll be valued about the same in terms of a revenue multiple as Google is today. So the interesting thing—and you can have 2 responses to that fact—is part of me says, “Oh my God, 5 years where you’ve got to pull it off. That’s a lot.” And then you’ve got to go, “It’s not crazy.”

I mean, you look at it, and I would be terrified of buying it at that price. But the interesting thing about that calculation was that you go, “Oh, yeah. It’s the beauty of compounding, right?” Five years times 1.5, and you just end up in an amazing place. Now, I haven’t checked that math fully, right? So it does feel incredibly lofty at scale.

There’s that great Scott McNealy quote about trading at 10 times revenues, where he walks through how absurd it all is when Sun was trading at 10 times revenues in 1999. Every once in a while, you should reread that quote, because he was totally correct and it ended in tears.

Intuitively, 122 times revenues is not a sustainable place, but they’ve got 3 huge trends on their side. They’re the AI solution for large corporates, they are the AI solution for defense, which is having a boom, and they’ve got the administration on their side.

Just this last quarter, they closed $843 million in US commercial bookings, up 222%. So help me with the math: 222% on $843 million. I mean, this isn’t quite Lovable growth, but this is pretty good.

222% at $843 million in their commercial division.

You’ve got to again step back and give credit to the founders—and obviously Thiel and co. in 2003–04—with a sense of mission around 9/11, who built this stuff and then moved into commercial. It was slow for a while. Some of the early commercial customers weren’t wildly successful, but they’ve ended up now in a place where, when you think about big companies, there are relatively few places where big companies want to do big things with a project like this.

It’s not the kind of thing you can give to little SaaS companies just starting out. Big companies need to spend on big initiatives with big vendors. The point is, all the other big vendors are old and stodgy, like IBM and Accenture. Now you can have these dudes who’ve been working for the US government, and they’ve got the whole forward-deployed shtick.

I see how it’s working, because in corporate—I went through the last earnings and actually went through a bunch of the use cases in corporate—and it’s all over the place: supply-chain planning, scheduling for an airline. It’s a whole bunch of very different stuff, right? You say, “How can a single SaaS app do that?” But what they have is a platform that’s been around 10 or 15 years. They have another platform they’ve added around enabling LLMs, and then you take the forward-deployed engineers and you can squint and say a lot of it is services. But who the hell cares? The margins are 50%.

The reason they can get it is they can look the CEO or the CFO of a Fortune 100 company in the eye and say, “We’ve done 10 of these. You give us the $10 million, and we’ll get this puppy done,” right? The other 2 competitors are a little SaaS startup that says, “We have a $100,000 product. It’s way more efficient than Palantir. It’ll work, but it’ll need a lot of work.” And they’re like, “I don’t know about that.”

The other competition is Accenture saying, “We’re going to build it from scratch.” They’re just in that sweet spot where they can make the pain go. You can get your AI initiative as the CEO of a large corporate company. You give them $10 million, you’ve got your AI initiative, and they’ll probably get it done. It’s a golden place for the next couple of years.

You know what the other crazy thing about Palantir is? Cut me off, Harry, if you want to move on, but I thought this was fascinating with Alex Karp. They’re a Rule of 94 company today, right? Pretty good, right? He’s on fire. Just watch the body language. He knows they’re going to crush it, right?

But he said that when they’re 10x bigger, at $40 billion, they’ll have 10% fewer employees than today, and they’re already on that trend. I think there’s a meta point. I don’t know, maybe it needs a few more people, but I think this has been well thought through, and this is a hint of the future in Palantir, right? This efficiency level.

Microsoft has already reached peak employee count, right? Google’s already reached peak employee count. But Alex Karp going on record saying that at $40 billion in revenue—10x bigger—they will have 10% fewer employees. I think it’s the journey we’re all kind of on. He’s out there because he doesn’t give a rat’s ass what anyone thinks, in his farm in Vermont and sandals, crushing the numbers, right? To me, this is the future of B2B companies: trying to get to massive, massive scale with few humans.

I wouldn’t pencil it into my model, but it doesn’t matter because, as you point out, with 50% operating margins today, don’t bother getting more efficient—just scale.

Well, he is getting—he’s not doing it, obviously, directly for the margins, right? But it’s how to structure the company. This may be where the future is. He’s just so far ahead; he can make these visions and bets, right, and not be hiding in 7% growth, as some others are. I just thought that that’s the future.

I think this age of these bloated companies is over. It’s the same as what you talked about with Shopify. Shopify is at $1.3 million in revenue per employee now, and it’s reaccelerated and reaccelerated. I don’t think it has—we could talk about Shopify if you want—but the efficiency at Shopify is not at Palantir levels; it’s breathtaking efficiency. Tobi’s ruthless on this, and founders should be too. Founders should be ruthless on this.

You don’t need half your company, and Palantir and Shopify are proving it. You don’t need half your company. You literally don’t need half the people working at your company. You don’t need it today.

I’m glad that half isn’t listening to this show.

Work hard. Work harder. Learn your product. Ask yourself: are you actually valuable to your company? Be honest. It’s not just whether you’re a people person. What are you doing at your company where you’re irreplaceable?

It’s not that you’re going to get fired or anything. You might be moved out. It’s just that your future is uncertain if you’re not irreplaceable. It’s uncertain. You’re not getting fired, Harry. You’re just getting moved out, you know.

Oh, there we go. It might be slightly hard to move me out, Rory. I feel a slight protective layer around me.

Yeah, the team should be uncertain. You may not need them for investing or content production, and in 2 years, you may not need anyone else helping you to invest.

I was more giving Jason a hard time.

This is an interesting question. Do you want your team to feel uncertain? Do I want my team to be scared that they might lose their job? On the one hand, it forces them to find the irreplaceability of their roles, and it also can make some people uncomfortable—horribly uncomfortable, scared. It’s not a nice thing.

My learning is it doesn’t matter anymore. There’s so much uncertainty out there. We coddled people since 2020—since the middle of 2020—and then we coddled them like there was no tomorrow until early 2022. Take 3 jobs, work 2 days from home, life is easy. Take care of yourself, and if you can come to work, come to work, but take care of yourself.

Shopify had twice as many employees at the peak as it has today. Shopify—forget about them—twice as many. Tobi was telling everybody to relax in late 2020: “Take it easy, guys. The commerce will come.” Now he’s freaking ruthless, isn’t he? He’s ruthless. Zuck’s ruthless. Karp’s ruthless.

And if you think you’re going to win in B2B, if you’re not ruthless, you’re going to lose. So I don’t think you should scare people. I don’t think you should tell people, “Hey, if you don’t step up, your job’s over.” I actually think that’s a dated approach. Just step up and they’ll quit.

I’ll give you a small example, if you want. It’s interesting. They’ll just quit, right? There’s a tool, for example, that we use. We use 10 AIs now at Aster, up from 0 at the start of the year. One is a niche tool called Momentum.io. It’s a cool startup. Rory, you should do the next round if you can, maybe.

It basically mashes up everything—Gong, Granola, everything—so that you have real-time insights into everything your sales team is doing. Is that product good?

It’s a good product. It’s not perfect, but here’s my point. Here’s what’s interesting. It’s not that any individual component is all that interesting. It’s how it synthesizes everything elegantly through AI.

But I’ll tell you, every company—and I got turned on to it through Kyle Norton at Owner, and I’ve brought it into other companies—every single time it’s been brought in, someone on the sales team has quit the first day. They quit every single time, the first day, including on our little SaaS team. Someone quit the day we brought it in. He quit that afternoon.

Because the gig was up. The gig was up. That’s why, Harry, I don’t know if it matters whether you prepare the team, because there’s going to be so much AI around us, the gig’s going to be up one way or another. If you’re not productive—if you were too busy to get the podcast out this week, if Harry had to do it himself, if you forgot to do the TikToks—the gig’s going to be up across tech.

So you don't have to cuddle people or scare them. I think they're just going to get a report every day that says, “You didn't push out enough code to Shopify; you're going to get pushed out.”

I'm going to go back to the scared thing because I'm thinking about it a lot, right, as you mentioned that. It's always funny how we start off with a prepared script, and then you take us so off script by the end. I haven't spent any time thinking about this until you asked, but I think there are a couple of different categories.

I don't think it's great for society if everybody's scared all the time, right? At one level, there are jobs where it's okay. Once upon a time, I ran a manufacturing company, and at that level, blue-collar jobs are pretty interchangeable. If it doesn't work out with employer A, you can go down the street to employer B.

I think it's good that most people don't live in terror of not being able to put food on the table. There are a whole bunch of jobs where it might work out with this employer, but I can get roughly the same wage from another employer. I have to follow the rules and work hard, but I'm not sitting there every day terrified, right?

I don't want the teacher who teaches my kids in Nashville to be terrified. They should know that if it doesn't work out at this school, they can go somewhere else. That's most jobs. But there are 2 categories. Maybe it's only 1 here. If you're holding down an extremely well-paid to extraordinarily well-paid job in an entrepreneurial company, you should be afraid.

And let's be frank: anyone getting paid $100 million over 4 years at Meta should be extremely well-paid. If you're pulling down those kinds of salaries, or if you're in an entrepreneurial company, you should be afraid. It's very different because—

But ironically, this doesn't affect any of those. This affects the $100,000-a-year marketer. This affects the $100,000-a-year SDR.

Yeah, they're going to be out of jobs, man.

But it's them that it affects, not the $100 million.

And the point—the good thing about the $100,000-a-year person—is that there are other white-collar jobs at $100,000. I'm going to say this: I think I'm a compassionate person. If I thought that person's next-best option was $20,000, I'd feel real empathy for them.

But if you have a $100,000-a-year job that's going to be automated and your next-best job is $95,000 doing something else, that's life. That's American capitalism. Get over it, right? I don't think that's an appropriate level of fear. You should feel the need to hustle there. I don't think that's quote-unquote fear.

Do you understand me? You're putting on a weird face, as if you disagree. You think that?

Well, no. I just think there's a generation of 23-to-30-year-olds who aren't really masters of the craft in any way and don't really know what they want to do, so they're doing SDR or marketing roles. They're about to have a train hit them in the face, and they're going to go, “Oh, shit.”

Well, let me give you a quantitative version of it.

But they'll pick themselves up and keep going because they're young, they've got a good education, and they'll be fine. It'll be a little bit hard.

Now they're checking out. Just to pull the numbers, because Tobi Lütke did do the night-and-day, black-and-white comparison at Shopify, let's look at Shopify for a minute. Shopify's peak employee count was in 2022: 11,600 employees. Since then, revenue has grown 91%. That's pretty impressive for a company at $11 billion in revenue.

Employees have gone down from 11,600 to 8,100. They've gone down while revenue is up 91%. I'm sorry if you're worried about it, because Tobi Lütke was early on this. He went into beast mode, and he destroyed the competition. BigCommerce doesn't exist. WooCommerce doesn't exist anymore. Amazon was never a threat, even though it wasn't a direct competitor.

He went into beast mode, and he just got there earlier than the cracked kids in San Francisco. He got there earlier. Shopify is doing 91% more revenue with 30% fewer employees. If you want to fight that, people would say I was toxic or was hurting their feelings a couple of months ago on LinkedIn. I'm like, “I'm trying to help you. Do you want to be one of the 8,100 at Shopify or not? You've got to decide.”

You can't leave work at 3:00 to get your salad and go work out. It's just not going to work. You're never going to learn the product, and even worse, you're never going to beat the AI that knows your product cold. This is the coming reckoning for these people: AI knows every Shopify feature, and every Shopify merchant knows the product better than the humans.

AI is smarter than most humans already, pre-AGI. It is smarter. Have you ever talked to an SDR who even understands the product they sell? Once in my career, have I talked to a 22-year-old SDR who knows the product better than me? It's worse with AI. These people are gone. Shopify is going to be at 7,000 employees and 200% bigger. You've got to adjust, right? The ship's left the station.

And do you think—let me ask a question here—at scale? First of all, I agree with that. Look, the automation—

The Shopify numbers are stunning. If you think about it, isn't it a 30% employee reduction, 91% revenue growth, and it's just starting? It's just starting. And that is before all the AI mandates went out, right?

It's concurrent, but it started before everyone in tech was saying, “Before you replace someone, find an AI first,” right? It's become the mantra, so it's going to accelerate.

And I'm just, again, yes, agreed. I'm going back to the scared comment because it's funny: I sound like I'm arguing, but I'm actually agreeing. My guess is that person is paid $100,000. Maybe they only get $80,000, but whatever.

The thing you should be scared about is if you're on board something like Shopify. What you've really lost there is the opportunity to participate in 4 years of equity that could have made you $4 million. That's fear. That's where fear is.

Anyone who's in the kind of job where they can make millions of dollars should be scared every day, because you should never take your job for granted, especially if you're in the kind of job where you can make millions of dollars if you do it well. I think that's the level at which you can be scared. Those folks should have been scared, not because they went for the $100,000-a-year job, but because of the equity they—

I get you there. The opportunity cost on the upside of the equity is insanely real. But having come from a family that's lost everything, I worry about the downside. Quite frankly, I don't think there are a ton of employers queuing up for a 23-to-30-year-old graduate from a mid-tier university who's not a specialist in anything, in a world of AI cost-cutting and economic questionability.

No, no one wants to hire them. No one wants to hire them.

They'll have to suck it up and learn to do different things, but to a rounding error, they'll be fine. If you look at graduate unemployment, the anecdotes are there, and it is higher than it's been. But I'm not going to cry for someone in their 20s who has to adapt, right?

It's a very different feeling. Put the kind of fear someone feels at 55, when their job is—fill in the blank—could go and they've got nothing ahead of them for 10 years. That's fear, right? I've seen that kind of fear, too.

It's going to grow.

With all due respect, you're right. I think we're actually hitting the same point again. You're overpaid. You could probably get another job. It won't be as good. It won't have the equity upside.

They will. They will. I mean, they'll be fine.

Listen, if a couple of years ago—until maybe 18 months ago—any seasoned B2B executive I knew needed a job, someone I knew reasonably well, I could get them a job with an email. I could reach out to someone in my portfolio, or even in my extended network—founders I knew and had met on 20VC—and I could get them the job.

Today, the odds are like 10% that I can get them a job. One, the fire isn't there. Two, they're not willing to be pushed hard. Three, they're not willing to come to the office. Four, they don't know the AI tools. I can't get them a job anymore.

You're interesting, but you're jumping. Are you talking about the 50-year-old guy or the 20-year-old?

Yeah, now I'm talking about 35 to 55.

No, those folks—

I can't find them any jobs—the veterans. They're not—

And I totally agree.

Not a single job. I say go to Cisco. I say go to Cisco before Cisco realizes they don't need you.

Guys, the thing I love about these shows is learning from you. I learned from you on Palantir. I don't freaking get how Monday.com can have a good quarter and be down 30%. What the what? Rory may have a more nuanced view on this one.

I know the Shopify data and Palantir. I think Monday.com is actually the one that should worry us more as investors, right? Monday.com was priced to perfection and growing to perfection. I don't think it's this quarter. It's that it came up a little soft relative to where they thought the growth would be, but it's still elite.

So the bar is 50% at $500 million ARR, never missing a quarter, and keeping it going. We throw around all these numbers at Sierra, Lovable, and n8n, but the bar’s high. That’s the only reminder to me: the expectations are so high today for the top performers that the ones with those outlier revenue multiples are held to an incredibly high standard.

Do you want an unfair question, Rory? I’ll give you an unfair question because you love unfair questions. Datadog, HubSpot, Palantir, and Shopify: rank 1 through 4 on what you’d buy.

I hate giving advice when I don’t feel informed.

I can give you a guess if you want while Rory thinks about it.

That’s what I love. Jason—

I’m going to take Shopify because it is gaining market share at scale.

Palantir is growing at scale and maybe it is gaining a certain type of market. It is dominating government contracts, don’t get me wrong. Some people could criticize me for being simplistic, but e-commerce is the biggest part of our economy, right? Even bigger than enterprise. Rory’s got the numbers.

The fact that Shopify continues to gain share is breathtaking, right? What if Shopify has 80% market share and the economy is 80% e-commerce? Help me do the math, right? Versus 13% of all commerce already going through Shopify, and it has essentially 90% of the platform market share.

So this platform lock-in, even in today’s AI world, is kind of exciting, isn’t it? You get to have lock-in. It’s kind of exciting to have lock-in.

You know, I buy your decision. I don’t buy your logic. Let me tell you what I mean by that. If they’re all being bought at the same revenue multiple, I think the Palantir opportunity in terms of gross-margin dollars is way bigger. I was comparing 108x ARR to 17.9x.

Exactly right. That’s what I was going to say.

That wasn’t implicit in my analysis.

No, that’s where it gets— that’s exactly right. You have to say to yourself, and that’s why I apologize, because you have to say to yourself, “Yeah, I think you have a relative…” A VC would say—and this is perhaps the difference between the question and the VC—if you were to rank the size of the opportunity from here, I think it would probably be Palantir, Shopify, and then the other two, which are really top-tier SaaS companies.

You’d have Palantir 1 even with the enterprise?

Yeah. It’s the size of the opportunity, not the valuation. Please don’t misquote me. Yes, commerce is huge, but gross-margin dollars—the ability to sell $5 million projects to enterprises to, quote-unquote, implement AI—strikes me as a wonderful place to be for the next 5 or 10 years.

While Shopify is huge, it is a more mature market, right? Palantir is at the white heat of, “This is new, this is exciting, and we have it; no one else does,” which is where Shopify, Datadog, and HubSpot were 5 or 10 years ago.

But what I can’t do in my head—and I’m not going to do it, because I’m not—is, as Jason says, how do you adjust that for the fact that the market is well aware of this issue when it’s trading at 120x versus 18x? What are the other two trading at? I know Datadog is 15x revenue, and I haven’t checked HubSpot in a while.

Well, here’s the brutal one: Monday, which we talked about. Off-the-charts good, right after the correction, at 8.4x ARR.

Dude, I bought the shit out of it. I’m going to be honest.

I know, but think about how many companies we have in our portfolios that are better than Monday, or that are not as good as Monday, that may be priced around higher than 8.4x. Monday is the one. It’s kind of soul-crushing. It’s like, “You’re sure your portfolio is so good if it’s trading at 8.4x, 4x?” That’s a tough compression to go down there, isn’t it?

When the growth dies, all these businesses become very uncompelling in terms of valuation. You’re exactly right, Jason. That’s why you have to be growth bigots all the way on these deals, because once they flatten out, it’s 4 to 5 times if you’re lucky. Even when they’re still at almost 30% growth, we’re too optimistic about our portfolio companies compared to Monday.

For 20 years, up until 2019, the median SaaS multiple was 6.3-ish, and the median growth rate was almost 30%. Now the median multiple is closer to 6. Again, the growth rate is 20%, but the profitability is higher. For the longest time, you only got 6x for 30% growth.

Well done, Rory. I think you did an amazing job articulating the 4. You have Palantir 1, Shopify 2, Datadog 3, and HubSpot 4.

That’s not what I said.

I think it is.

I said in terms of the other stats. I did not say that. If you say that, I did not say that. I said in terms of unpenetrated—

We’re going to title this “Rory’s Public Leaderboard,” with your face, to see the new guest on your next show.

Marc Benioff is coming, and he wants to hang out with you, Rory.

Well, then you have to be nicer to me.

I’m going to be nice to you. Do you know what’s amazing? After the last show, Cliff from Canva—done. Aravind from Perplexity—done. Jeff from Twilio—done. Marc Benioff—done. All wanting to come on the show. All fans. Little do they know that when you come on, you just get asked stinker questions where you have one of your most successful CEOs in the mix and you’re just a jerk.

Harry, what’s your net worth? What’s your net worth?

Guys, we mentioned that. Where is the upside and where’s the opportunity? Carta’s State of VC Q2 2025 came out. There are a couple of things I want to dig into, and then we’ll wrap. “Highest valuations ever for seed” was an interesting one, and it goes back to what we said about comparing to Monday. How do we feel about risk-adjusted returns and where dollars are best? How do you feel when you see the highest valuations ever for seed? Do you see that reflected in your daily work?

The Carta data, and everyone else’s data, also says there have been fewer seed rounds done than 12 months ago and 24 months ago. Rory will be better at slicing the data. Everyone should be cognizant—especially founders—that there are fewer deals being done, and it’s worse than that because the deals are very specific.

Forget about the breathtaking growth of AI-native leaders. It makes sense if you’re concentrating into winners that this would happen, right? We’re also just concentrating into winners across the board. It’s not just 10 deals for 40% of the dollars; it’s everywhere we’re seeing concentration.

Agreed. To me, that was the much more interesting point: the concentration. Sorry, Harry. The concentration in late-stage rounds.

It’s called revenge, Rory. Don’t worry, it’s fine.

The leaders and the concentration in late-stage rounds is just amazing, right? We run an internal process where we look at every deal done and the total dollars. We do it every quarter.

For the last couple of quarters, we’ve literally had to back out 1 or 2 deals because they make the statistics so weird. In Q1, the deals we forecast in total—deals in our sweet spot, enterprise, B2B, all told—I think totaled $12 billion in total dollars raised. OpenAI raised $40 billion in the same period. It’s literally twice our entire addressable market for us and 100 other A and B firms, done in 1 deal. That’s a huge level of concentration.

The same thing happened in Q2. Meta is interestingly reported as an investment in Scale AI because they put the money in, so in theory it’s a venture investment, which is absurd. But you have big rounds for Anthropic and xAI where literally your entire sector is smaller than 1 round at the super-late stage. I’ve seen concentration, but never to that extent. It’s just—

The concentration is unprecedented. It just is what it is. People should be aware.

Is that a momentary element of time, where we are in the cycle, or is that going to be a continuing feature of a new age of venture and technology?

It’s probably not going to go back completely. Look, it’s not going to go back completely to where it was. Do I think there’s going to be a $40 billion round every second Monday? No.

There are a couple of industries that now appear to be venture-accessible but are fairly capital-intensive. Obviously, LLM model creation, and obviously a lot of defense. There does appear to be a higher propensity to do more capital-intensive industries. That’s 1 thing.

The second thing is that the more you stay private for longer, the more this becomes a phenomenon. Holding them longer means they become bigger companies. Bigger companies, just to run their balance sheet—you can have $20 million on your balance sheet if you’re running a $100 million revenue company, and if you’re cash-flow positive, you’re fine. If you’re running a $10 billion company, even if you’re profitable, you probably need a couple of billion bucks on the balance sheet just to manage fluctuations.

As these companies stay private for longer, there’s going to continue to be this steady stream of fairly humongous later-stage financings.

And as I say, add to that the nature of the businesses. The AI model companies and the defense companies are capital hounds in a way that SaaS or consumer internet wasn't. So I think it's maybe not as pronounced as it is now, but it ain't going back to everything being A's, B's, and C's and nothing being more than a $100 million raise. It ain't going back.

Everyone that wants to go back or give that dated advice—we had Brian Halligan on right before, and he was so good. If folks that got this far haven't watched him, that was S-tier. Go watch that one. His point was that, at best, half of what I learned at HubSpot matters today. It's true in venture, right? We're not going back. AI is so much bigger than cloud and so different. Unless the LPs cut off the valve, we're not going back to the old venture.

Yeah.

Which has been a good fact for the large firms, just to put it out there. It's the synchronicity of people having a lot of money, putting it out in lots of deals in 2021, and that not working. You could have said—and I might even have guessed—that the next stage of the movie was real retrenchment at those firms, because you'd seen that doing 100 deals at $10 billion pre didn't work and maybe you can't deploy that much capital.

Along comes this crop of extraordinary companies, like OpenAI and Anthropic, where you can deploy large amounts of capital. I think it really has provided justification for the opportunity for those larger funds to say, "Hey, look, you can't put a billion dollars to work in 2 or 3 model companies." The rule of thumb in venture is, "Oh my God, there's no diversification. That's terrifying, and it's not our business." But I can totally see talking to an LP and saying, "This is the only way to access that risk. If you want to access that risk, you need this vehicle." And that's why they exist. It's been marvelous for them.

I've said this before. I completely changed my stance on mega-mega-platform funds being able to deliver actually great returns. Because if you can move a billion dollars into OpenAI at $30 billion, like Josh and Thrive did, I'm not sure what their initial check was. Holy shit, there are very few opportunities where you can get a 10x on a billion dollars. And actually, now with outcome scenarios being so much bigger, I'm much more bullish.

Final one, and then we'll do a quick fire: do we believe in the one-person, billion-dollar startup? It's something that's so often hailed—

As a theme, I mean, Jason articulated the numbers really well. As a theme of fewer people doing more, yes, but it's kind of in the limit thinking. It's just hard to imagine. By definition, you can't talk to any customers. What would that person do in a billion-dollar startup? What would they not do? How would you run the key functions of a business? Who'd do the accounts alone?

Stop. It's idiotic in the extreme, right? If you're running a company with a billion dollars in revenue, you're probably making some money. You've got to file the taxes. Maybe you can outsource everything, but you end up—someone's got to meet the accountants, someone's got to review the tax return. Even if your only problem is putting the money in the bank and paying the taxes, you're going to need people.

But is headcount going to be less? Absolutely. Is it going to be 1? It's a metaphor; it's not a reality. Jason's going to argue with me.

No, I—well, 2 things. One, as I've said before, I take everything Sam Altman says and think through it very carefully now, in a way that, at the beginning of this conversation, I thought was marketing hyperbole to get attention.

Look, the one-person company has a little structural risk in a billion-dollar company, a little hit-by-the-bus risk, even if it's possible. I actually think that if you allow for some outsourcing agencies and resources that come and go, we will see a bunch of 20-, 30-, 40-person billion-dollar companies. A core of engineers and a bunch of AIs will be making things, and people are going to be like, "You know what, dude? I don't want 100 sales reps."

If you're more SMB or self-serve and you don't need the forward-deployed engineers of Palantir, I think folks are going to make a trade-off with AI and leaner teams. They're going to say, "I'll do more PLG, self-serve, with a little bit of a trade-off from humans, so that 30 of us can run a billion-dollar company together."

Software—you used to write software, stick it on a CD-ROM, and you'd write once and a million people would buy it. I think AI is going to be the renaissance of that. It's going to be the renaissance of that, and we will see it. People want to work with 30 or 40 great people, and they're going to use AI to try not to hire the other 950. They're going to try to use it in any way they can.

With agencies, outsourced accounting, all that—when we say 30 or 40 people, you're going to need a lawyer, you're going to need accounting, but it doesn't need to work for you. They don't need to work for you. And that accounting firm or law firm may be AI-powered itself, right?

I kind of buy that as a better argument than the one-person company. I mean, look, in one sense, there are solo-GP billion-dollar funds today, so you can argue, there you are, but they don't just have 1 employee. Even a solo GP has some junior people, some accounting—you have stuff, right?

How big is the house that you're buying in Yellowstone? Is it big enough for me and Rory to come—

—to support you and Rory in London for 3,000 people? London has some physical costs, unfortunately. It's not a 90% margin, right? But the work does tie to the work, in a sense, right? And there's a trade-off: we could be doing more revenue, we could be doing more marketing. I don't want to do that stuff, and we don't want to do it, so that's why I think—

But as I said, our AI BDR that we deployed this week set up 3 meetings in the last 2 days for 6-figure sponsorships. That's the trade-off people are going to make. They're going to say, "I'd rather spend—I'd rather orchestrate these AIs than deal with 100 people quitting every day because they want to go to yoga or weightlifting, or they just don't want to do it. This generation isn't going to want to do it. They're not going to want to do it. I don't want to do it."

How do you judge this generation? Maybe it's because I have kids.

I love everybody, but people just don't want to. If you can choose an AI, you're going to choose an AI over an unreliable resource.

But when I show up to SaaStr in London, there's going to be 50 or 100 contractor people running your business. Just to be clear, to be factually correct—

Yeah, there'll be 100 people scanning badges, and so that's my original point: there'll be transitory resources. But the headcount—and going to your earlier point, to Harry's point—the equity will not go to those people. The equity will get more and more concentrated. That means people will make more and more money from equity, right? But it will become isolated in fewer and fewer people making more and more money.

We're already seeing this in the big AI wars, right? We're already seeing the $100 million engineer packages. We're going to see the best salespeople who are orchestrators be able to make $10 million in sales, not $1 million, which is what a top rep is making today. We're going to see the top salespeople make $10 million, but they're going to be AI-fueled, and they're not going to have 200 reps under them. They're going to have 10.

Yes. And the other—

—and they'll get $10 million.

And to be clear, the other 90 reps will have intermittent jobs scanning badges at SaaStr. If you had a $500,000 business, you won't starve because you can get jobs scanning badges at SaaStr, but you won't be living the life you led. That's your point about fear. Going back, I've been thinking about what you said earlier.

It's a real issue.

And I think that's why I really hope we don't live in a world where everyone's scared, because that's not great. But if you're getting more than the national median wage, you should ask yourself: are you contributing more than the national median amount? And the more you're getting, the bigger the gap. And you're right, there are a lot of jobs in tech.

I do believe that the gap between what you're getting now and what you'll get in your next-best use could be 70% to 80% of your current income, and that would make me scared. So I'm coming around to the view that you're right. Unfortunately, more people have to be scared. It doesn't make it a great world, but you're right. Even at the $100,000 to $200,000 level, you probably won't get a replacement slot. Jason, you see, we're even convincing the great master.

No, no, no. I'm convinced by facts, and even Jason's statistic about 2 people running his business blows me away. I'm just—

But it's not. We're at the bleeding edge. We went from the non-bleeding edge to the hairy edge of the bleeding edge. It wasn't all by design. It's just so hard to find people, right? What's happened is we've inadvertently done the Tobi thing. We're like, “Enough already,” right?

We've had people who made $500,000 in sales quit because they didn't want to work past 5, or we had one quit because they didn't want to enter his data into Salesforce. It never made as much sense. So you've got to be like, “I don't want to go through that again.” At this point in life, I'd rather hire an AI. We spend 2 hours a day orchestrating them.

Rory, here's the thing: We spend 2 hours a day getting them to work, but they don't quit.

I'm really intrigued to see what that orchestration input looks like in a couple of years' time. That's the job you need to get if you're not running your own fund. You need to be a chief orchestration officer. This job is worth $500,000 today. Everyone's hiring for this one. This is the job you need.

We're going to do a Kalshi quickfire. Kalshi is obviously this prediction marketplace that bets on cool, real-life things. I'm going to create my own first one, following our conversation, because I really want to: Palantir over or under a 5-year market cap of $2 trillion? Its current market cap is circa $450 billion.

I love the company, and they've exceeded all expectations. I just think compounding to $2 trillion from here is pretty damn hard. So I'm a no.

Gosh, I'm pretty decisive. I'm just comparing that to Salesforce. Salesforce at $40 billion is now worth $222 billion. Now I'm feeling gravity when I'm looking—literally, I'm looking at the number 1 and number 2 by market cap. Palantir is worth almost twice as much as Salesforce as we write this, and Salesforce is number 2, right, at $222 billion.

So I'm looking, and the AI is bigger than cloud. That's the reason it's going to happen, right? But I'm worried about gravity when Salesforce is only worth $222 billion as we record this. I want to take this bet, but I'm going to go the under.

Okay, we've got: When will Stripe officially announce an IPO? Before June 1, 2027 or after?

The correct response is, “Does it matter anymore?” I mean, aren't they so post-public? I don't know. It trades all the time. I'm sorry, I'm being a little bit like, whatever. Cash-flow-positive, wily. So they're just doing their thing.

But do you think they will go public in the next 2 years?

Yeah, as I said, going back to what I said: If the cost of capital gets markedly cheaper in the public markets, then maybe. But they seem to be more resistant than most, and there's nothing that's going to make them do it because they're strongly cash-flow-positive at huge scale, and there's liquidity.

So the truth is, it's an idiosyncratic bet, and the Collisons haven't shared their opinion with me. So, hell, I don't know.

Very honest. Thank you, Rory. You're not chatting to them in an Irish WhatsApp group—

—over a beer in the studio. Over a cheeky pint with Dario. That was a very important sign. I thought one of the best fintech founders in the world was moderating Dario. I just thought it was ironic.

Ten years ago, people were like, “Harry, media will never work for you, dude. Just to let you know, it's not a thing.” Okay.

True. I think Stripe does a super-interesting job in media, with the publishing and all that. More than most companies, there's an intellectual curiosity there, which I find the most attractive thing about that company. I like them to go public, which is—

No, I meant it as a compliment. I didn't mean it badly.

I know. I just want to pile on that. Yeah, it's 2 smart people talking to each other.

Okay, let's do a final one. Will xAI sue Apple? There are odds for this one, and I know you like your odds, Rory. So, yes: $100 gets you $29. No: $100 gets you $151. Elon has made it pretty clear he's very unhappy with Apple, and he's already sued OpenAI. There's a precedent there, right?

Yes. A, because he appears to have a propensity to pick fights with everyone, which is just fun to watch. B, because the only reason he would have is the dispute with ChatGPT. To some extent, you could—I saw a good tweet that basically said Elon is wrong because the whole idea that Apple is going to favor anyone—they don't need ChatGPT to be wildly successful. They need to figure out their own AI, so it's a marriage of convenience.

If Apple continues to be, quote, aligned with ChatGPT, then they probably get drawn into it, because the ChatGPT-versus-xAI fight is going to be existential for a long time. That seems to be both personal and business, because you have the whole xAI business. So that fight isn't going away.

Anyone getting sucked into that could get pulled into the mess. To the extent they are, then, yeah, it's probably not a bad bet. It's kind of like getting named in the lawsuit: You didn't do anything wrong, but you're dealing with someone that did something wrong to me, so I'm just going to pull you into this mess and depose all your executives anyway.

I think it makes sense. I mean, Epic suing them was sort of worth it, right, at the margin. And he's incensed, and he's got a lot of money in xAI. I mean, he's the richest man in the world, but he doesn't have unlimited capital. So I think it happens.

Guys, as always, I so appreciate this. I want one final question from you. You gave me a target—someone to get for the show. I think I delivered with the people I've lined up. If you were to add 1 more name, who would you say it is?

Well, look, for what it's worth, Alex Karp has always been my dream since the beginning of SaaStr. I've asked every single year since 2015. Personally, if he has the energy to do it from his farm, that would be my dream.

Guys, you are awesome. Yes.