20VC：OpenAI 的 60亿美元 Jony Ive 交易｜YC 既是 Chanel 也是 Walmart——并已正式胜出｜Builder.ai 崩盘、Hinge IPO：谁赢谁输｜Seed 好卖，Series A 残酷，以及 Late-Stage Venture 的肮脏真相

The wars that you choose to engage in dictate what it takes to win. We have to let go of this vision that venture has to involve fund returners. The only way the math works is if you stuff money into the very best company, and you don't end up with a balanced portfolio. You end up literally with 1 company having 20% or 30% of your fund in it, and that company turns out to be the big winner. I think your absolute assumption has to be, “YC has won.” It's the greatest—one of the greatest—equity businesses ever.

We're going to dive right into 2 Insight deals. A winner and a loser happens—

Yeah.

—in this game. Insight had both Builder.ai, which raised $500 million, I think across rounds, and then shut down over false projections. I read that they were projecting $200 million, and they actually got $45 million in revenue. Let's start there. Losing $500 million is a lot. It's over a $100 million hole for Insight. I'm sure it's more. How did we analyze this one? This was a big hole.

1. Builder.ai Exposes Venture Risk

At first, I read it and I was shocked. It almost seemed like fraud, losing $500 million on an AI website builder, right? But then when I read it, Viola Credit—or whoever held the debt—shut them down because they missed their projections, is what I read happened. That's interesting. Rory, when is that okay? Have you ever had a portfolio company, or done an investment, where the founders were more aggressive on their projections than the numbers they actually hit? Have you seen that in your storied career as a VC?

I'm shocked to discover that not every plan happens and not every company has 100% attainment of plan. But let's start with Insight's perspective: $100 million. I think the last fund is $12 billion, so it's approximately 1% of the fund. No one likes losing $100 million. Money is money, but in the context of the game they're playing, does it matter? Is anyone getting fired for this, honestly?

Stepping back, does it matter? Of course it matters. Should you get fired is the first question. If you do deals, you make investments. If your unit of account is $100 million—which is an impossibly large amount to any of us on this call, still less any of your listeners—but it's just the nature of the checks you write if you have a plus-or-minus $10 billion fund. You write $100 million checks. Some of those $100 million checks don't work out. You lose the money. If you lose more than you win, you get fired.

The good news is, I think Jeff Horing himself, the founder, was involved. He's made scads of money. I think he was a big and early investor in Wiz.

I don't think Jeff is going to lose his job for dropping $100 million when he's returned a couple of billion dollars somewhere else. They'll all be just fine. So I'm not spending any time worrying about whether Insight will be just fine despite this loss.

Yeah.

So—

On the flip side, though, they had Hinge Health, which returned 5X the money, or $400 million. Great win. It is in a $6.2 billion fund. My take on that was, is that just the nature of the game they're playing? They're never going to have fund returners by nature of it, or is that actually a pretty immaterial exit, as awful as it sounds?

I think it's the former. We have to let go of this vision that venture has to involve fund returners. I think as the fund size gets larger, the kinds of bets that investors make are—you have more bets, and they're more later-stage. I think the probability of any 1 deal, quote-unquote returning the fund, goes way down. It's not impossible at any level, but at seed it's almost a necessity, and I defer to you guys on that. You have a better feel for it.

For example, our model is that we do 20 deals: 30% are bad, 50% are solid and return 1 to 5X, and 20%—4 companies—return more than 5X, with an average of 10X, which by definition means about 0.5X of the fund. We've had fund returners, but our mental model is that we have to get 4 of them right, each of them good enough to return half the fund. Therefore, you get 2X from your big winners, another 1.5X from your base hits, and there you are. That's a model at our stage.

Now, you go to a $6 billion or $8 billion fund—10X our size—and you're probably looking at some kind of model like that. If you're doing roughly equal-sized bets, it just gets harder and harder to assume that a single deal is going to transform the whole fund. So you're left with this dynamic of having to, as you say, make $500 million, have a happy day, and then say to yourself, “We're 10% of the way there.”

Rory, when you look at the historical data, where have your predictions of that value dispersion been off? Have you had more losses and more high-upside returns? Where has it been off in the historical data?

It's been pretty accurate overall, but what you'll internalize—and this is where all venture is the same—is that, if you think about the degrees of freedom there, you actually have 6 numbers. You have the percentage in each bucket, and you have the value of each bucket. The biggest single variable that can influence things is the return in the best deals, right? In other words, that's the thing that can, quote, save you to the upside.

If you have 30% in losses and you get 0.5X back on those deals, it doesn't matter much if it's 0.2 or 0.8. It just doesn't move the needle enough. By definition, the number of deals in the middle bucket are middling, and so by definition, a middling number can never change anything. So really, the only thing that counts is that you have to have 20% of them in the amazing outcome, and then the tail on that amazing outcome is what dictates the overall fund.

I remember a venture guy I knew who's been in this business a long time. He basically said, “You have to have a model something like I just articulated.” And then every 3 funds, something utterly amazing happens that you can't really forecast. Instead of that winner being a 10X or a 15X, you get that 1 20X, 40X, or 50X, and that fund's just amazing.

And then, of course, it's a fatal error to assume you're going to do that every time, because it's just not likely. But that's mentally how you think of the model.

Tying it back to the beginning of the conversation, Insight Partners owned 43% of monday.com when it IPO'd.

Yes.

So today, I hope they get more, right? That's harder today—to collect 43% of the next monday.com, right?

What you're wrestling with here for these large funds, you're absolutely right. You can run the math, and then you do what Josh Kopelman did and ask, “How likely is it that there are enough exits at that size and stage to allow you to achieve your objective?”

When you run that math, you realize, “I need 6 $10 billion exits in a year, let's just say, and on average, there's only 4 a year.” The math doesn't work, especially if multiple people are doing it. So what you end up with is that the very tippy-top of the tail only works when, instead of having 20 equal-sized bets, the only way the math works is if you stuff money into the very best company and you don't end up with a balanced portfolio.

You end up literally with 1 company having 20–30% of your fund in it, and that company turns out to be the big winner, right? Because there just aren't enough big winners. That's why, as I said, you look at Founders Fund and stuff all the money you can into Anduril, because it's the only way to deploy that capital.

Brian Singerman always told me the enemy of great venture—

Yeah.

…returns is capital concentration limits in an LPA.

Yes. It's especially true at scale. It's not Jason's enemy, it's not your enemy, because the truth is, at your stage, you probably don't know enough to have the certainty to put 30% of the fund in 1 deal. Because let's be honest, at seed, you know jack.

And this is why the game has changed. When companies stay private for longer, the correct way to play the late-stage game is very different from the correct way to play the earlier seed, and even A and B, game that we play. And that's why, going right back to it, these mental rules of thumb that we have—“Oh, 1 deal has to return the fund”—are wrong and irrelevant for the game those guys are playing.

What they have to do is return that $8 billion in paltry $500 million chunks and make damn sure that there's 1 deal where they have 20% of the fund invested in it, and that gets $2 or $3 billion back. That's the mission. It's a different business.

2. Speed To $100 Million

One number that everyone is so focused on, especially founders—they are just so fixated on it, and I think a little bit because of the Twittersphere—is the speed to $100 million. We have Mercor, Lovable, and Bolt, all in the race to $100 million ARR. It's this AI wave that is so focused on it. As investors today, how much weight do you put on the speed, the time it takes to get to $100 million ARR or X-number ARR?

You'd be an idiot not to weigh it at some level, because we're all traction junkies. Once you move beyond seed, traction is the best proxy for overall commercial success. But you'd also be an idiot to weigh it at 100%. It's an interesting qualifying proxy, but it's by no means dispositive. That's the kind of boring nuance answer that I specialize in.

My worry today, when I think about that, is that it'd be nice to say, “Listen, those are all great examples, guys,” but a company that grows at a great rate but a saner rate, that is less dilutive and has a bigger moat, is a better bet or is just as good a bet. We'd like to think that, right? And maybe it's true, but what I worry about are 2 things in today's world: access to talent—

Yeah.

…and access to capital. And there is just…

Yeah.

Listen, talent has always been a moth to a flame with the hottest startups, right? But AI has just amped it up. It has just amped it up, and every smart engineer, every smart kid wants to work at the hottest AI company. They don't even want to work at Rippling or Deel. They want to work at OpenAI or Windsurf.

And I'm just going to say something in their defense. They're right. Wouldn't you? The best advice you can give someone starting out in their career is to join an amazing company that's going to be at the forefront of things for the next 5 or 10 years, so you can be there at the start, build the connections, and build the knowledge.

It's not like the kids these days are bad. They're entirely rational in planning their careers. Join the wave, because that wave's going to last them the next 30 years. In much the same way, if you think back to starting out in SaaS in 2000—or whatever, 2004 or 2005—great call. It gives you a 20-year horizon.

Especially in engineering, the smartest people have always wanted to work on the most interesting problems—

Oh, yeah.

…right? And the most interesting problems… So the problem with the plodding pace, which I would like, is that it's hard enough to compete with Deel and Rippling for talent. Because, listen, all the best sales talent I know wants to work at Rippling.

No, yeah.

Half of my old team works there. So you're already competing with Rippling, but poor Rippling's competing with Windsurf, Cursor, Granola, and Sierra.

Agreed.

And then access to capital. I don't think 80% of the B2B VCs, like we've talked about, want to touch something that doesn't have a chance at hypergrowth. So if you're not in that category, the math looks great, but you better have your own little ecosystem where you can thrive, your own little world—

Agreed.

…where you can recruit. And you don't need as much capital, and you better be copacetic about it and have a strategy there, right?

Agreed. Thank you, Jason, for bringing it right back to where we started, which was this $100 million: is it meaningful? And I think the answer we're both saying is yes, it's meaningful. It's just a proxy for and a magnet for success. It's not perfect, because there will be churn. $100 million with low churn—that's very bloody meaningful.

Yeah, but maybe it's better to be in London. It still shocks me, the employee churn rate at OpenAI. I can't believe how many people leave. You're leaving so much money on the table to go to another AI startup.

I don't know who did it, but that was quite interesting information on the Anthropic retention rate being significantly higher than the OpenAI retention rate. That was just an interesting piece of data, and it maybe speaks to why those investors are taking that dilution.

Going back to the comment on dilution, whenever I'm on a board, on a compensation committee, and we start talking about, “Oh my God, the dilution's too high or too low,” what I always say is, “I want to see 2 other pieces of data.” I want to see attrition. Are we losing people? And I want to see the close rate on offers. Are we failing to hire people?

Because in the end, forget morality—it's a marketplace out there. If you're losing people a lot, especially for economic reasons, or if you're not able to attract the talent, then maybe dilution isn't high enough. Conversely, if you're not losing people, maybe we can manage dilution a little better.

In the Anthropic–OpenAI wars, I think they're doing what it takes to keep people. Clearly, as you say, astonishingly, people are willing to leave OpenAI despite all that.

67% retention rate for employees after 2 years—67%.

Yeah, versus 80% for Anthropic.

Yeah, 80%. That 67% is—

Big difference.

Yeah, it's brutal, man.

You're leaving so much cash on the table.

Hey.

Yeah.

Not everyone is as greedy as you are, Harry. What can I tell you?

Sod off.

I don't know. Greed's complicated. They may think that they're getting more. Some of the problem—

Fair enough.

…is if you've been there 2 years at OpenAI and you made $8 million in tender offers, you might think another $8 million is easy. The mindset's complicated. And I think—

3. Hinge Weakens IPO Protections

I do want to go back to the Hinge IPO, because there was some stuff I was tracking a week ago that we talked about, and I actually got a lot more clarity on it in the last week. I think it's super interesting.

And it goes a little bit to, yeah, Jason, some of the stuff you circulated about the state of the unicorns and where they are, right? We've had 2 IPOs in the last week: Hinge Health and MNTN. One of them is a digital health company, Hinge Health. The other is a digital ad-for-cable-TV company, right?

Solid businesses, $200–300 million in revenue, decent growth rate, wonderful outcomes for the VCs involved, right? And you take Chime as well, as being on file. It's a proxy for… There are a couple of takeaways here.

One is there is an IPO market here right now, right? And you don't have to be $5 billion. Much of the stuff that people said—the comments about how you have to be at $500 million, which is, I think, a Cooley comment, and the comment that the window was going to shut a month ago—both of them are wrong. Deals are getting done right now, in the last week, that are wonderful outcomes for all concerned.

That's the first point: $200–300 million. It is no longer $100 million. It's $200–300 million.

It's growth. It's profitable—

Good point.

Or near profitable. But that's what it takes to get something done. These are facts on the ground.

But 48% growth, right? Still pretty high.

Yeah. Hinge Health is growing nicely. MNTN's growing as well, but not quite as aggressively. But yes, solid growth. You're exactly right.

Yeah.

And, holding that thought, Jason, it's actually terrifying how few of the unicorns are close to that level, and we'll come back to that point. But the other point—I was just down in the venture weeds, but it really matters—was trying to figure out how all those super-high late-stage rounds get processed through the lens of the IPO. In other words, does the last round have a block and therefore can stop a down round? Do they not have a block, or was the last round low enough that it doesn't matter? We now have one of each in those 3 names.

MNTN, I don't think had a prior round that had a block, and I don't think they had a prior round that made the IPO a down round. Perfectly normal, boring IPO. Hinge Health is really interesting. They raised money at $6 billion in 2021. That money clearly had a block—not that they could stop an IPO, but that they couldn't make that preferred convert to common as part of the IPO. Which, in my mindset—and it turns out to be wrong—meant that they could, quote, “block an IPO.”

So what happened? If you look, there were a couple of investors, and if you read the detail of the S-1, it's very dense. But what you figure out is Coatue, who clearly paid $6 billion for a company that's now gone public at between $2 billion and $3 billion, did some kind of negotiation. They sold some shares back to the company, and they bought some common. So they basically agreed to convert in return for some kind of make-good. In other words, they were able to get around the block and come to some kind of economic deal.

Some of the other preferred investors in the last round didn't come to such an agreement. I thought they could, quote, “block the IPO,” but it turns out they didn't. The preferred simply stays in place. That preferred doesn't convert to common until they hit $77 a share, but the IPO got done. What this is is the public market saying, “Hey, you guys want to go public. You got this preferred on the balance sheet. Normally we'd say clean up the whole balance sheet, but it's like we're not going to say that here. We're going to get this deal done.”

The deal priced in the mid-$30s and trades in the early $40s. You've still got this stranded $200 million block of preferred that doesn't convert to common until $77 a share. But that's their problem; we don't give a shit.

Do they lose money, then?

Yes. They haven't, quote, “lost money,” because they still have their theoretical value. They don't have to convert from preferred to common until it's $77 a share. Well, they don't even have to convert ever, but they don't have to convert until it's $77 a share. So they didn't, quote, “lose money,” but they're sitting there in a non-interest-bearing instrument that's way out of the money.

The mark-to-market on that is now clear for all to see. Congratulations, you have a preferred stock that's a 1X, and you don't make any return until the stock gets to $77, and it's now trading at $40. What it does is, to some extent, weaken the ability of those later rounds of preferred to block an IPO. This is really significant. It weakens the ability of those later rounds of preferred to block an IPO, right?

The market is working. The public markets are saying, “We can deal with a bit of noise. We can price this. It's just stuck up there in preferred.” It's a little isolated pile of capital that's clearly underwater because the common price isn't worth it. So it's a 1X instrument, and therefore, on a discounted mark-to-market basis, it's worth less than a 1X.

If you're them, you'd rather it be bought for $2 billion by someone else.

Yeah, absolutely. If you get an M&A, you get your money back. But this is what I love about it. Maybe they couldn't sell for $2 billion, or maybe the other investors correctly wanted to go on and build a damn big company.

The big news here is that this removes the ability of that late-stage, high-priced round to get in the way of everything. You can't make them convert. They still have their, quote, “1X,” good for them. But you, as the founding CEO, and you, as the early investors, can go into the public markets and get on with your lives. Get your 10X if you were the first one, get your 4X if it's Insight, whatever it takes, and get on with your lives.

That late-stage money is stuck at a 1X, 0% IRR for the next 3 years. Knock yourself out. It's a really big deal.

It's a theme of a lot of these implicit protections we thought we had in venture. To me, the big learning was that acquirers wouldn't buy you unless 98% of folks agreed, okay? Some of these implicit protections are breaking down. One of the worst investments I made, they were only able to get 80.1% of the shareholders to agree to it, to an exit. The acquirer didn't care at all. Whatever the statutory minimum was, okay? The acquirer didn't care.

Got it.

They just didn't care 1%. They didn't even attempt to get the votes from the other 19.9%. They didn't care.

And I think what all of these things have in common is the capitalist universe is recognizing that there's $2.7 trillion of privately held assets that are going to have to go public, find a home, and it's going to involve a little more complexity than normal. But the great thing about capitalism is people find a way. In the case of Hinge Health, they found a way to get it public with a preferred stock. In the case of your deal, they found a way to just close the deal and accept the risk. And I think we're going to see a lot of that, because that's what it's going to take to deal with these 600 unicorns.

Yeah. Anything you can get around, people are going to get around to go public or make a dollar, right? Any rule, get around it.

And fun fact, I mentioned a third of them. I didn't know the answer a week ago on Chime: is there a block? But I got interested, as one does, and I pulled the pre-IPO articles of incorporation. What are the terms right now? And it's super interesting.

The last 2 rounds do not have a block. I think the $6 billion round says that above $6 billion, all the preferred converts to common. There's no way for that preferred to remain outstanding; it just converts to common at any price above $6 billion. So, in answer to your question, Harry, in that case, those investors will record an immediate mark-to-market. I had a 1X at $25 billion; now I have a 0.5X at $12 billion. I've taken a loss.

And it all boils down to one little term deep in the bowels of the liquidation preference auto-convert terms. That's what Hinge Health didn't have and what those guys did.

So they're going to crystallize the loss when they go public.

They're going to crystallize the loss, exactly. In the case of Hinge Health, the preferred investors are just going to sit there at a 1X in an illiquid instrument in a liquid common stock. In the case of Chime, they're going to get auto-converted, provided it clears $6 billion. And if you're still carrying that at $25 billion, you're going to record a significant loss. If you've written it down already, you're fine.

So this is not the great game that we thought it was: you get your money back, and then when it pops, you get the premium on top.

Absolutely not. That's why I said, if you zoom out a million miles, this is all about what happens to those 600 unicorns, right? The interesting fact is—and Jason circulated the SVB work—these are the best unicorns. These are the unicorns that can go public, and what you're seeing is that some of the late-stage money has protection, keeps its 1X, and has a miserable IRR.

Some of the late-stage money cuts a deal and says, “I'll roll the dice on converting to common as part of the IPO.” And then some of the late-stage money has no damn choice and just gets converted to common and takes a loss. There's a lot going on here.

Oh, God. No, no, I'm not loving that seed is for suckers anymore. This late-stage shit is hard.

Everything's hard.

Oh, God, that's not nice. If you're in Chime, you're going to crystallize those losses. You're going to lose 50%.

Well, maybe not. Maybe it trades at $12 or $15, but I don't know where it trades, to be clear, and I think it could trade much closer to that. All I can say for sure is that they don't have a block. There is a mandatory conversion where the later-stage rounds don't have a block, and if that conversion is exercised, they will be converted to common at whatever the prevailing price is.

My only takeaway from this is I want to be Jim Andelman with MNTN.

Now you're just circling back and forth. What were we— As I said, we hated seed a week ago, and now we're like, “Oh, my God—

No.

—I want to be seed.”

No, bullshit. I just want to be at seed in 2013 or 2010.

Turns out vintage is the single most important and underrated part of venture capital. Just being there for the good years.

Listen, you mentioned the brilliant report that Jason shared. I thought one really interesting element was that accelerators and incubators are 24% of all VC deals. 24%: accelerators and incubators. Does that mean Y Combinator's just won this game? How did you guys read that?

4. YC Has Won This Game

For the first time recently, there is more competition at the accelerator/incubator phase.

There's more, right? But YC is 4 batches and bigger than ever. I do think they've won. They did 2 tilts. It's not the same YC as it used to be.

First of all, bringing in Garry Tan was a massive change. Obviously, an uplift, right? For sure, a level-up, but also just a massive change on all levels. And 2, it's obvious, but massively tilting into AI when they weren't ahead of the curve and being a center for it to attract the best talent.

It's very fluid, but right or wrong—and I'm not into the brands—people want to go to Harvard, Stanford, and MIT, and the kids want to go to YC, right? They're doing this massive event for hundreds of the best kids in college very soon. It seems new, but when I look back, all of my first investments were in some sort of accelerator—5 out of 5, right? So it's not brand new; it's just bigger than ever.

I think your absolute assumption has to be that YC has won. Look, it's one of the greatest equity businesses ever, and I deliberately use the word “business” as distinct from just “fund.” The thing about being a fund and investor like any of us is you're only as good—as I've said many times—as your last game. Every day you have to get up, make good new picks, and if you blink and get the picks wrong, you're done. You're out. The world doesn't need you, right? There are 600 or 700 funds like you.

The beauty of YC is they've got a business. The definition of a great business is if the owner of that business, Paul Graham, can be sitting back in England, walking around the cute little bookstores, and the machine keeps humming. That's a damn great business. Why is it so great? Because the world needs one. The Valley needs at least one big accelerator like that where, as you say, Jason, they'll take in anyone, provided they have the smarts and the nous, and they can convert those 2 people from London, that 1 person from Sweden, and those 2 people from the Midwest into highly marketable properties in the space of 3 short months, in return for a mere 7%.

The world needs that product, and it needs it on an industrial basis. Give them credit. Give Paul Graham credit. The original stated intent was to make it easier for startups. That was the mission, and they've succeeded. And because they've succeeded, they've built, as I say, a compelling business.

You can say sometimes it's better run than others. Some CEOs of that business are better than others. The current CEO seems to be doing a pretty amazing job. It's just a great business. And I did the math once trying to figure out—seed is not what we do—but how much better is the YC locked-in return than a seed fund at the same stage?

They basically have roughly a 2X advantage. If I look at the deals they do and extrapolate based on the published hit rates and success rates, if a seed fund investing at that stage with decent picking gets a 3X, they get a 6X. It's a structural economic advantage that's very compelling. It's a great business.

And it should be a great business, because, at the risk of sounding like a defender of free-market capitalism, it met a market need at scale. It met it brilliantly, and therefore they deserve the return. Go team. Wish I'd thought of it.

For second-time, third-time founders, it's still a niche product, I think, in my ecosystem, for folks that have been around.

Totally.

For every Parker Conrad that wants to do it again at a much better deal, other founders don't get it, right? But for the first-time founders—

Absolutely. I think the—

That and Project Europe are the beacons.

Absolutely. I mean, if—

I have to say, it is astonishing. I mean, this is what gives me hope for Europe, honestly. We have 8,000 applicants to Project Europe. 300 or 400 of them are pretty fucking awesome. It's amazing.

Could not agree more. It's that kind of thing that just wasn't there. Again, going back to the mission, I give Paul Graham credit. He's just one of the clearest thinkers. I've never met the man. I've read a lot of his work; he's just such a clear thinker. The objective was to make it easier for founders to start companies, and that's the kind of thing Europe needs.

If you make it easier for founders to start companies, more companies will be founded. Most of them will be mediocre. That's life. But some of them can be freaking amazing and cover a multitude of sins. That's the way it's meant to work. So, go Project Europe.

I think the thing that's so amazing about it is I always say the future of venture is won by Walmart and it's won by Chanel. Chanel, the incredible brand with a very specific customer base. Walmart—whatever product you want, they've got it. And what I think is so special about YC is it's Walmart and Chanel. They have scale, and they've retained brand.

Absolutely.

It's still an aspirational brand that has managed to do scale.

That's very good.

Let me repeat: I believe it to be one of the great enduring equity businesses, in the sense of businesses. There are lots of funds; there are very few enduring brands that occupy a clear niche. And the test you run is: if they went away, would someone else rise to fill the gap? Absolutely, because the world needs that product.

If the 600th venture firm went away, we might just stop at 599 and say, “We're good, thank you.” And that's the difference.

5. Series A Becomes The Hardest Stage

If we move slightly down the funding spectrum from seed to Series A, you know I like to start with some comment that is completely unsubstantiated with data, but I feel like Series A is the hardest place to be investing today. When you hear me say that, do you agree with me, and what's your take?

I guess my question then is, if seed is so easy, then that's really troubling, the combination, isn't it? I mean, it means we're underestimating the failure rate to Series A, right?

Seed is easy because it's Walt Disney versus Jerry Maguire. Walt Disney is, “Tell me the story.” There are lots of people that tell a good story and come from great companies. And then Jerry Maguire is, “Show me the money,” and there are actually very few people who are showing the money in a way that's actually true quality, sustainable, and attractive for a Series A investor.

And so I think that's why seed is good and Series A is hard.

Weirdly enough, I agree with you, despite the data. On the one hand, the conversion rate from seed to Series A, per the Carta data, has gone way down. On the other hand, I think what you're saying is correct: for the stuff that's working and in the chosen hot markets, there is mass competition.

Everyone wants to get the early traction in those AI companies that are just starting to explode. So you are correct: if you are in the chosen sweet spots, almost every venture person is looking at those deals. And yes, those are the very few deals that are wildly competitive.

Look, we competed in a couple of deals in a kind of broadly recognized emerging AI space. Brutally competitive. We didn't win. I think we got outpriced in one and out-beauty-contested on the other. So that's pretty tough.

You say to yourself, first of all, what about the other 75%? The odd thing is, what we're saying effectively is it's hard to invest in Series A companies while 75% of the companies you could be investing in are dying and struggling for capital. So there's a little part of me that says maybe I need to figure out the non-obvious Series A and make some money that way.

So, yeah, it's only tough when you're competing for the best. Because, again, it turns out to be hard to make money.

It is hard. It's funny, going to Rory's point of whatever you look at looks harder. I was thinking the other day about RevenueCat, where Harry and I both are investors. It announced it raised its last round at a $500 million valuation, which actually was low-ish. They just did a deal in 1 hour at a $500 million valuation.

I was the first investor in 2018 at $7 million pre-money. Okay? At $7 million. That was right before YC, and I thought about a deal I did in the last batch that was probably the same risk profile at $30 million.

Now, obviously, deals are done higher or lower, right, at YC. But I'm saying there are some similarities between these 2 companies, RevenueCat and this new one.

Yeah.

So, $7 million versus $30 million. How does that math work if the fund size is the same? Do I need a fund 4 times bigger? Am I taking 4 times the risk? Help me think about this, because I'm roughly thinking—now, granted, upside has gone up, right? That's the meta point, right? But they're still—I would say they're the same, but at 4 times the price versus 2018.

So do you adjust check size or adjust ownership on entry?

I feel like I have no choice. If the round is bigger, you write a bigger check. If the round is smaller because it's a YC deal and they don't need as much money, you write a smaller check. I honestly feel like I have no choice. That's not new. That's not—

When I invested in RevenueCat as the first investor, I wanted to put twice as much money in. I just wasn't allowed to, right? I could only buy 10%. But that is a question, right? Either way. But it's 4 times—where do I get the 4X? 4 times the fund or 4 times the risk?

In mild consolation, Jason, I'd probably say it's only around 2.5X worse off, in the sense that you do have to give some credence to the fact that, as I've mentioned earlier, we've just had huge GDP inflation over the last 5 or 7 years.

Yeah.

A buck 10 years ago is probably 60 or 50 cents today on a GDP basis. Not just inflation, but inflation plus growth, which is what you gotta look at. So it's probably not 4 times worse, but you are correct: it is probably at least 2 times, and maybe 2.5 times, as risky per dollar as it was. That's the first thing.

Yeah.

The second thing is you're exactly right. You do have to expand the check size, because the truth is, you have to play the game on the field. We've seen that at our stage, which I think of as typically at least a stage later than you, right? We've had roughly the same ownership targets for 15-plus years and the same typical deal size, but the check size has gone up to get roughly the same ownership. That's just the nature of the beast.

I have a slide in our deck where we point it out to the LPs and just say, “Look, this is the dynamic of the marketplace we're in now, and this is the scary part of the game.”

Do you guys honestly get 15%?

We typically get around 10% to 11% ownership, and that's been pretty consistent across 5 or 6 funds, all the way back to 2009. It's a pretty typical median. We don't navigate off ownership targets because, as I tell people, I would happily go later and take less ownership if I could get our target return.

But in the pricing environment for the last, frankly, decade, it's been almost impossible to have clarity that, on average, a later-stage round would give you your target return. Obviously, some later-stage rounds have given amazing returns, and we can come back to that discussion some other time. But on average, probably not.

It's typically been those early-revenue companies at $2 million to $3 million, growing hyper-quickly, where you're glad to get 10%, and 20% is not on the table.

For sure. Everyone gets less ownership than they claim on Twitter, right? Every seed manager is like, “We've raised a new $30 million fund. Our target ownership is 15%. We're gonna do 25 of those in the fund.” The math just doesn't make sense on planet Earth, right?

6. Dilution Compounds Over Time

For me, the bigger learning, Rory, you figured this out. You'll laugh when I say this, but this snuck up on me: Now that companies hold for so long, these investments—you hold them for so long. I did not fully understand the compounding nature of dilution.

Totally.

If you model 6% dilution per year for a portfolio company for hiring, and you hold that investment for 15 years, what is 6% compounding to over 15 years? Is it 9% ownership, isn't it?

It's huge. I mean, it's obviously not 9%, but you're exactly right because it's a descending scale. The impact of dilution over time is huge. On the other hand, you can't avoid it because you do want to hire the people—

You can't avoid it.

No, it's an interesting discussion, because honestly, the least enjoyable, least rewarded, but most necessary part of my job as a board member is that I'm often on compensation committees. You're trying to set up policies for companies not in their first 4 years of life, but in years 7, 8, and 10, where you have to grant new shares and re-up founders.

Sometimes I think that's a very legitimate thing to do because you want them incented. But at the same time, you have to manage overall dilution. Trying to keep it down, probably not to the 6% or 5% level, but to the 3% to 4% level, and manage that over time is just really important.

You're exactly right: 6% a year for 6 or 7 years, or even 10 years, is just a huge impact on everyone, including the founders and the initial equity investors. Spending a lot of time on refresh policies for mid-stage tech companies is, as I say, the combination of low-value, low-joy, high-impact work.

It's so funny you say that. I was with an investor this morning, and he said the challenge with LLM investments—and we have one of the best—is that the employee stock is 9% a year. It's not 6%; it's 9% to 10%.

Yeah.

The level of dilution on the employee stock grants is so much higher than traditional that it makes it an even harder venture category to invest in, which I thought was interesting.

I think it's gone up. I wish I had all the data, but Harry's point—when I look at it, my insight is this: The exits I had in 2021, when everyone had a lot of exits, had a much lower dilution profile than today. I haven't had a billion-dollar exit since 2021. It may be quite a while until I have one.

I made up the term “dilution profile,” which probably makes no sense, but that was a term I made up on SaaS a while ago. It was much lower than today in 2021.

I look at those exits and I'm like, “Man, that was pretty good.”

What was the dilution?

If I look at the companies today, I'm like, “I'm not gonna own that much at those exits,” right? They better be much bigger exits, because otherwise, man, I'm gonna own so much less.

When I go into a deal, I assume 40% dilution from my entry. Is that a reasonable heuristic, or do you think I'm over- or underestimating?

It's too low for seed. For Series A it might be okay—or no, it's way too low. I think you gotta assume now two-thirds dilution from seed to IPO. If you don't do pro rata to IPO, two-thirds.

I think there's 2 types of dilution, obviously. There's the following-round dilution and then purely the option dilution, and I'm not sure which you guys are talking about, right?

I'm combining them all to two-thirds.

Combining them all to two-thirds.

It used to be half. I think it used to be half if you didn't do it. Now I think it's approaching two-thirds.

Exactly what Jason said. Absolutely. There's a very pure test to this, with lots of data associated with it. You just look at Y Combinator, because they have fixed ownership every time. They have 5,000 data points.

If you felt strongly about calculating the answer here, you just look at their ownership at IPO, and you have a statistically valid sample. Go do the work, right?

And that's why they've dramatically increased their ownership.

Yeah.

They're no dummies there at Y Combinator.

I think we should all be—

First, moving to post-money instead of pre-money, supposedly for the benefit of the founders. Then raising their ownership, having anti-dilution, and investing more. They're the only ones that have caught up, I think, as to why seed and dilution matter more. Power to them.

No, but I want to go back to the foundation model comment, because I think it's something that we should internalize and get humble about. We, the capital providers, are not the most important people in the equation, and you better just internalize that's the way it is.

The important people in the equation are the people with the IQ and the STEM knowledge and the ability to generate these models, and anyone running those companies is gonna pay those people what it takes to keep them. Frankly, tough shit on the dilution for the capital provider side. That's the nature of competing in a huge talent war for small numbers of people who can do amazing things.

So yes, you can bitch about it and say, “Oh my God, that's awful,” but it's a cost of doing business. We just had the largest single instance of dilution in a foundation model in terms of the acquisition this week, which is obviously, at some level, an acqui-hire.

What you can see is that it turns out it takes 2% dilution, or $6 billion, to hire 1 really great VP of hardware engineering and a solid team under him. The capital providers are along for the ride. You better like the terms of trade and internalize them. You don't have to like them, but you have to accept them if you wanna play.

The terms of trade are being set by the leaders of these companies and their need to attract that amazing talent in a very competitive world.

It was funny. I was looking at MNTN. How do you say it? Is it MNTN that IPO'd? How is it—Mountain?

Mountain. They call it Mountain, by the way. MNTN.

I just figured you'd know that.

It's easy.

I was looking at it. Founded, I think, in 2009. Jim Andelman, who's at Bonfire, whom I've known—

Bonfire.

I immediately flipped the prospectus and said, “How much does he own after all these years?” I think it was probably in Bonfire I. It was probably a very small fund. We could look it up.

Yeah.

$20 million to $30 million, and he still owned 9-point-something percent at IPO. That's old school. This is gonna be a good deal for him, right?

Even if the market cap is only—and I'm putting this in quotes—only $2 billion, if he owns $180 million on a $20 million or $30 million fund, that's a great outcome, right? Today, at a company just like MNTN, you don't have that at IPO or less, right? The fund size might be 5 times bigger for seed.

So instead of a 5X or 6X performer, it might be a 1X, right? Sign of the times.

You actually had this with Michael Kim at Cendana, who said that Erik 12X’d DPI the fund with Honey for Mucker.

Yep. Yeah.

Returned $280 million to them. I thought that was incredible.

But would that happen with Honey today, right? Probably not, right? The thing is, when you look at these old deals, they’re great because they were all modeled on much smaller exits, right? And much less competition, so you could get the ownership. These old-school founders often don’t see the same dilution. I read through the prospectus, and I didn’t see the— But I bet the Mountain guys were very conservative, right?

Yes.

There’s no way Jim could own 10% over all those years if the founders were giving away 10% or 12% of the company a year, right?

But I also think it’s, as they say, horses for courses. They were playing a different game, and even today there are different versions of the game. I think Mountain was trying to build a profitable company in a pretty—I won’t say well-understood, but a defined—space where you’re not competing against a gazillion companies.

Even at the time, you weren’t competing against the largest market-cap companies on the planet. It’s very different from if you’re building an LLM today and you’re competing against Microsoft, Google, and OpenAI. The wars that you choose to engage in dictate what it has to take to win.

And the bigger takeaway from this, Jason, and you’re right, is that you can create meaningful economic value for yourself, for your investors, for your fund, and for yourself as a founder, in markets that are significant but by no means as huge as the AI foundation-model bet. Mountain is a great example of that, as is Hinge Health, the other IPO last week.

They’re 2 really solid companies, with $200 million-plus in revenue, growing 30% to 50%, depending on the 2 deals. Solid outcomes, multibillion-dollar outcomes. Everyone involved made money, but we’ll come back to some nuances on that. Great classic venture outcomes, just a very different game. When you’re trying to compete with someone who has publicly stated it’s going to take another approximately $50 billion to get to cash-flow breakeven, it’s just a different game.

7. OpenAI Bets On Hardware

I do just want to take this a bit in turn, because there are so many elements here. We mentioned Jony Ive. Obviously, we saw OpenAI’s $6.5 billion acquisition of his design studio/company. Is it a simple acqui-hire? And when I say simple, I don’t mean cheap. But is it a simple acqui-hire bringing Jony in to do a hardware play for OpenAI? How did you guys read it?

Well, look, first of all, it was really interesting that he’s not joining full-time. He’s still managing his design firm, which was an important point that I’m sure has been worked out, but it’s super interesting. It was clear he’s not joining full-time. He’s still managing his design firm. They’re just buying the startup that he’s a founder of.

For the $6 billion—is it $6 billion? Is that what it was? They’re not even getting him full-time. I bet they’re getting most of his time. I’m sure they’re their top client. Of all things, I thought that was a sign of the times: that you had to pay $6 billion, but you could get away with not getting the guy full-time as part of the deal.

Sam’s so smart, right? And he was clear in that video. He’s like, “I want the third device,” right? The laptop, the phone, and the third device. At first, I laughed. I’m like, of course, that’s what every tech dude in San Francisco wants.

But then they said, “Listen, ChatGPT has crossed 20 minutes per day for the average user.” So going from 20 to 200 with a device for a small percentage of your market cap, so you could have 10X the coverage, 20X the coverage of your life—that, if this is the right guy and the right team, might be the best investment they could make. To go from 20 minutes to 200 minutes, from 20 minutes to 200.

I think we’re going to live in a world where our AIs listen to us 24 hours a day, one way or the other, whether it’s on our watch, this device, or on our screen, or in the background, like Granola or Notion. It’s always going to be listening. And I think it might be a war he has to win, to always be listening.

“Might” is, of course, the word here. It might work. My perspective is that every single significant software-platform company develops, at some point in its life, hardware paranoia—the feeling that somehow the hardware guys are going to screw them—and the only way they can stop themselves is by spending a whole ton of money attempting to build a hardware platform.

They almost invariably fail. Making sure that doesn’t happen to you, and scratching that terrified itch, is just a part of doing business. If you look—if you step back—at Microsoft: “Oh, my God, Nokia’s going to… We should buy Nokia. Oh, my God, we should build a Surface.”

If you look at Facebook, it’s like, “Oh, my God, we should build these VR devices, because otherwise we’re going to lose in the metaverse.” If you look at Google, “We need to, quote, ‘own the phone,’” so they crank out Pixels at absolutely no margin.

Everyone does it, so who is Sam to break this time-honored tradition of spending a lot of money on a hardware device? My guess is that, statistically, the likely outcome, just based on the priors of the other companies in the space, is that 3 to 5 years later, it turns into a fizzle. It didn’t pan out, but it’s okay to try.

No way. It’s going to be huge. It will launch in a year. It will launch in a year. It will be massively subsidized, so it’ll be $20 for this device, okay? They will figure out the form factor. I don’t know what the right combination is, whether it’s embedded in your ear, like the bracelet the cool coffee guys wear, or it’s that lid you wear backwards. And within a year, we will be living all day long in AI, and the timing will be perfect.

This is great because we now actually have something that we can track and discuss. What you’re saying—I’m saying I don’t think it’ll produce anything meaningful, a significant revenue driver, but it’s okay to do. I’m not saying it’s dumb.

Yeah.

I’m simply saying it’s an itch that every platform vendor has to scratch. What you’re saying is you believe that we will look back 2 or 3 years later and go, “Wow, they shipped a meaningful device with meaningful hardware revenues as part of the overall OpenAI business model.”

I think he said it’s going to ship 200 million, and I don’t think that was a throwaway comment. Here’s the difference between us. One difference, though, is I spend almost 2 hours a day in AI already.

Yes.

Four months ago, I didn’t. I spend 2 hours a day in AI, between our AI tools and everything, okay? I don’t do anything without AI anymore. And so I can already see it.

Listen, I’m scared about it. I’m scared that ChatGPT will now rewrite itself so it doesn’t shut down. I don’t think that’s a joke. I don’t think it’s a joke that Anthropic’s Opus 4 is threatening researchers with blackmail over affairs. I don’t think it’s a joke.

But I’m already at 2 hours a day with AI, all day long. Literally, at SaaStr Annual this year, one of the folks who helped put us on rewrote the best summary of the day that day, of his day. He led an entire day. He hosted our Chief Customer Officer Summit, John Gleeson. And then that day, he wrote the best summary of the entire day.

How did he do it? He just had Granola running on his phone 24/7. Granola—I’m not running Granola now, but I might next week, so I don’t have to do anything. Granola and the new feature from Notion are pretty cool and pretty creepy. They run at the hardware level. You don’t know it’s a note-taker. You can’t see anything that’s recording every minute of the day.

But agreed—

—that’s recording every minute of the day.

Agreed with all that, Jason, but actually, the key sentence for 200 million consumers is that it ran in the phone. How many people are going to be willing to spend $200 or $300 or $400 for another device, and then make it part of their daily lives for—

It’ll be $20. It’ll be $50, and it’ll be cool. The thing is, Jony Ive will make it cool. If it’s cool, it’s the elusive next device, and I think all Sam needs is for it to be cool and to work, right? Just think about the Ray-Bans, like the connected Ray-Bans that everybody has. It’s wildly successful.

But going back to my first comment, I think it’s okay to try that. Because when you’re at the stage that OpenAI’s at, you should make those bets. And it gets right back to the comment on dilution, even, right?

Just think about it. 2 people got 2% of OpenAI in the last 3 months. One of them wrote a $6 billion check. Have any of us ever seen $6 billion? Have any of us ever had it in our account? No.

And then the other one signed a part-time working deal, sold his 55-person design studio, and got the same amount of money. If that doesn’t show where the capital providers stand in the hierarchy, in the great AI race, nothing does.

I’m sitting there as SoftBank or someone like that. I just wired you $6 billion, and you effectively took that $6 billion and gave exactly the same ownership to a 55-person startup.

I think we will need to remember that Sam will need to go out and raise more money. He says he needs to spend $50 billion. What this enables him in terms of a storytelling—

I see.

Narrative is unbelievable. Now he's got a hardware play done by the guy who did fucking Apple. The Saudis will give him more money than he needs with this new chapter, this new challenge that needs funding. It's a great story.

Totally. Absolutely.

20 minutes to 24 hours would be a great slide. That'd be my slide. We're going from 20 minutes a day to 24 hours a day. If ChatGPT could be monetized per minute, and then per human—

Mm.

That's a lot more revenue. 20 minutes to 24 hours. That might be the strongest PowerPoint argument that ChatGPT is underrated, right?

But I think without this, to get another $20 billion for the next model, that's hard.

It's a good insight.

We're starting to exhaust the capital.

It's a good insight that—

Yeah, yeah.

I probably missed. Yeah.

Totally agree with you. In terms of storytelling, this is the kind of thing you do. And that's the thing in these hyper-growth markets where almost nothing is certain: you would far prefer to take the dilution and cover the base than be wrong and get sideswiped. But right now, making that kind of bet totally makes sense.

It's also 2% of market cap. It's not a Hail Mary.

Uh—

Sorry. And Harry, I can say this, as a—

Not a Hail Mary.

Go on, Rory.

You're sitting back there in England and you must look at Jony Ive and go, “Oh my God, if every STEM and design graduate from every high-quality London college isn't figuring out how to get on a plane and go to San Francisco, I don't know what they're thinking.”

Do you know—

$6 billion.

You know what I'm thinking, Rory? I'm thinking, ha, Americans still have to buy Europeans to get some taste.

You know what? The other thing I kind of liked about the deal?

Okay, Harry, you're exactly right, Harry. That's why Europe has 1 very rich entrepreneur who does taste, Bernard Arnault, and we have the other 9 who do tech and have all the money. So knock yourself out with your taste. Taste you can buy. Cold, hard tech lasts forever.

You know what—

So yeah, you can feel good about that.

You know what I love? I love the American banking sector. After years of trying, you guys create fuck-all enterprise value. We have a Russian in London who creates a $100 billion behemoth that makes—

Well—

Chime and everyone else look like child's play.

I want to go back on that. I didn't realize we were going off on that tangent. I'll tell you very directly: that's partly because you have a mediocre incumbent banking sector there, and thus there's a huge amount of surplus value to be extracted. What's the company in Brazil? I'm sorry—Nubank, right?

The crappier the existing banks, the bigger the opportunity for fintech. Broadly speaking, in the United States, with a few exceptions, some of these companies—these existing guys—are pretty efficient. So you're right. There wasn't the same inefficiency to attack there. The opportunity for fintechs was a tougher business to get to scale.

Now, you had the countervailing fact that you have the Visa and interchange revenues, which are pretty exciting in the US, so that's been an advantage to the US over Europe, where they're more capped. But yeah, I don't think attributing success in fintech in London to the greater entrepreneurial qualities of Russians living in the UK is perhaps the most logical analysis, Harry.

I agree. We should instead look at a Swede who creates Spotify, changes the music industry, and creates a $100 billion company.

You're right. Yes. Look, I'm delighted to see you guys have some wins. Genuine comment. It's wonderful to see Europe have some wins, but unfortunately, the data just shows the vast majority of market cap in technology has been created, first of all, in the United States, second of all in China, and then Europe is a far distant third.

You'd love to change that. Broadly speaking, despite recent events, Europe is broadly, quote, “on our side.” It would be good for the United States if Europe would get its act together and have amazing technology companies. But for some reason, you just don't seem to be capable of doing it.

Ow.

Sorry. That was harsh.

Well, maybe I—

This wasn't even on the agenda, Harry. This wasn't even on the agenda, so we're riffing at this point.

No, I know. I don't disagree with you.

8. Europe Fights For AI Talent

Actually, Harry, can I ask you a question about it? A vibe check, because I'm curious. I've been coming to London for years and done a lot of Europe-to-US, but the pull of SF for AI is so powerful. We can argue over France and elsewhere, but it is powerful to founders.

Is it powerful among the entrepreneurs that you meet across Europe? I mean, EF is hybrid now, isn't it? I mean, EF is basically—

I'm going to get in trouble for this. EF sold itself out by going to San Francisco. It's a complete sellout. To your question of—

Yeah.

Actually, we have a huge number of people who say you can only build companies in Silicon Valley, and so people listen to this—

And I'm not one of them.

And I'm not one of them either. The founder communities are very aware that it is incredibly hard to retain great talent in SF. It's incredibly expensive, and you're competing against OpenAI—

Yeah.

And Anthropic. And so yes, it has the allure of Hollywood for our industry, but I think when you dig beneath the surface, the smart ones are going, “Actually, I can get better people for cheaper in London, where DeepMind is, where unbelievable AI talent is.”

Absolutely. You can't afford anybody in the Bay Area as a startup. It's—

No.

The inflation is so high, to our point. But the SF Bay Area is not what it was pre-2020. In some ways, it's smaller. But if you're a solo founder, if you don't know anybody, if you're an outsider, the sense of community in AI that you get in the Bay Area is so powerful.

In 2019, I would tell founders to come to the Bay Area because if you're in B2B, you walk down the street, you're going to see everybody, because we're all working in an office.

So I would say—

But it's more this community. I underestimated the power of sitting in Dogpatch, seeing every YC founder, seeing Sam Altman on the street, seeing everybody. The density is actually higher than 2019 for AI—for founders only, not for SDRs, not for marketing managers, not for everybody else. But for founders, it's nutso, the density. It's nutso.

It's a smaller community, for sure.

Yeah.

But it actually makes my life easier because we have 3 companies which are crushing and have created an ecosystem just in themselves: ElevenLabs, Synthesia, and Granola.

Yeah.

And all 3 of them—

Yeah.

Have created a mentality that you can build amazing AI businesses in London. So it's actually easier for me because it's a much more concentrated supply of great AI talent that's not as distributed as the Bay.

As long as you're in the hackathons, hanging around ElevenLabs and hanging with Mati, you're kind of near greatness. It's easier.

But do you feel, as a founder—and this is going to sound facetious, but it's not—do you feel in London, with Granola, ElevenLabs, and Synthesia? I'm a fan of all of them. Do you feel like you're failing every day as a founder?

Because that's the special part of being in SF: you feel like you're failing every day compared to everybody around you. If there's only 3, I might feel like I'm doing pretty well.

It took me a while to internalize what you're actually saying. It wasn't—I thought you were saying, “Are you failing in London?” What you're basically saying is the core San Francisco value prop is a feeling that, no matter how well you're doing, someone else is doing better and you just gotta compete more. Is that what you're saying, Jason?

Listen, right now I'm in—

It sounds like the story you're telling me, Jason—

I'm in SoCal. I feel like a failure the instant I get off a plane in the Bay Area. I'm not joking. Even I feel like a failure—

Wow.

Every day in the Bay Area. And sitting here on the beach where I am for 1 week—

I'm not—

I'm feeling like a success story right here, right now. I'm not kidding. I haven't been full-time in the Bay Area other than this last year, and I felt like a failure every day.

The YC company that hasn't announced yet that I did—the one I compared it to—the founder's a pretty good AI guy. He's like, “I'm 26. I feel like a failure.”

“I had 1 small exit. I’m falling behind everybody,” right? He literally works 7.5 days a week. That failure feeling, I think it’s bigger than ever, and I just think it drives founders.

I yell at some of my founders, “Feel like you’re failing sometimes.” I yell at them. I’m like, “You should feel that way.”

Listen, my mother would argue that I have too high an opinion of myself, but I have a high opinion of myself. I’m a fucking machine—

Yes.

And I am not motivated by an external person.

I agree.

I’m motivated by the internal of me, and the greatest founders that I know are motivated by the—

I agree.

—internal fire. And being in London, I push myself every day to be better. Everyone around me is relatively mediocre in terms of London, generally speaking, population-wise. I don’t need them to feel failure. I will push myself, and the best founders push themselves.

I don’t think that the London versus San Francisco thing is a function of that, to be honest. I think anyone who’s built a successful company like the 3 you mentioned, Harry, the truth is, the drive that it takes to do that in a country where it’s not the norm, in my view, speaks to something even more powerful and entrepreneurial than those entrepreneurs, right? Because over here it’s almost like you go to Stanford, you drop out, you go to Y Combinator. It’s almost like it’s the preset path.

Whereas for someone who, having been an entrepreneur and frankly failed in London, in the UK, and gone bust, it’s incredibly hard in Europe to be entrepreneurial. Therefore, the people who do it have something really determined and awesome about them. So I do believe that at the human level, those guys—I’m not worried that those guys aren’t competitive enough. I don’t think that’s the issue.

I think the real issue is not that. I generally find people are roughly the same the world over. I think what is true is the systems to become successful are way more powerful in the Bay Area than in Europe. You’d like those systems in Europe to be better, but just as an objective statement of fact, your ability to bounce back from failure, your ability to get capital a second time—it’s all the other things that make it a lot easier in the UK and the US to be successful.

In fact, to make this very concrete, I always used to say to people talking about Europe, “Exceptional people rise to the top anywhere in the world.” The strength of the United States economic system is we can take mediocre people and make them damn successful. That is the secret superpower of the US free-market economy.

The truth is, the mediocre people in Europe can often drift off to government jobs, whatever, safe sinecures. In the United States, the free-market system keeps the whip on everybody’s back, and as such, a lot more people have to strive and become successful. I don’t think there’s any lack of genius in the UK or in Europe, just as there isn’t here. I don’t think there’s any lack of drive in those individual people. It’s just that the systems to transform that individual talent and drive into a successful ecosystem are so much stronger here.

We mentioned Duolingo then. We kind of go back and forth on this. Jason, you’re always predicting that AI is going to replace all of our jobs. Then we’ve had Klarna backtrack on it, and now Duolingo is backtracking on its AI stance. The question is: Are leaders getting way ahead of their skis and then walking back, and is this going to be a continuing trend?

I thought when I read that, it sounded to me like every backstage conversation I had at SaaStr Annual this year with public-company CEOs, which is that they’re trying to guide folks to a truth. Not everyone can say what Fiverr said, right? He got there really fast: “Listen, we’re going to go AI-first, and we’re not going to hire anybody we don’t need to. AI can do better than our contractors, and we’re only going to hire people when we have to.”

I think CEOs of public companies are trying to prepare their teams for it, but the backlash was too strong. So I just think he had to walk it back to 70% of the truth. I think they’re just walking back the fact that everybody knows they don’t need 30% to 40% of the team they have today. Everybody says this—not everyone, perhaps not if you’re 50 people like Granola, but everyone with 500 employees and up that I talk to off the record, including public companies, says, “I don’t need 30% to 40% of my team.”

I think we’re going to see mass layoffs in the next 24 months. I think the net headcount is going to stay flat, right? But I think he just walked it back because it’s too hard for people to hear. It’s too hard. There’s only so much honesty you can get from a CEO.

Just for some stats: In 10 years, they made 140 courses with humans. In a year, they made 140 courses. So 10 years of work took them 1 year with AI.

I think, actually, on this one, I think Jason’s 100% right. I think you see every one of the CEOs oscillating between 2 extremes on AI. One extreme is, “Oh, my God, it’s going to make us wildly efficient because I’m sucking up to Wall Street,” and I probably overpromise on that side. You saw Klarna do that and have to walk that back.

Then on the other side, when you do the, “AI’s going to change everything, I’m going to save a whole bunch of people,” your other constituents, which are your employees, lose their shit, and you have to walk that back, too. So what I think people are going to evolve to, as Jason said, is the very bland statement: “We’re going to adopt AI. It’s going to make everything better. I’m not going to threaten ‘mass’ layoffs. It’ll just make things better. Oh, and by the way, we’re hiring.”

Jason nailed it exactly. We have evolved to standard corporate speak for how you talk about AI. It’s going to make us efficient—Wall Street, wink, wink. No one’s going to get fired; you’re just going to do more interesting things. That’s the current state of the lie.

The good news, I think—separate comment—is that I think it’ll be just fine. There will be efficiencies, and there will be jobs that would have existed in the absence of this product that won’t exist now. So there will be tension, but I don’t think, Jason, it translates to mass layoffs.

We’ve had this discussion iteratively. I think it’ll take a lot more time to adopt. Some companies, especially tech companies at the very forefront of this, will see significantly reduced hiring. I saw that LinkedIn executive who posted that graduate hiring is pretty screwed up right now, in part because people aren’t sure how many graduates they’re going to need in computer science and all those things.

I definitely think there’s going to be an impact here. I don’t think it’ll be, quote, “mass layoffs.” I think it’ll be more of a steady grind of 2% or 3% less hiring per year—a tweak at the margin of the organization. You’re going to keep trying to move it forward. It’s just going to take time.

Klarna was just what everybody wants to do. They went first. They went to the extreme. But all it is is the future pulled forward a certain number of months.

Yeah.

I think it’s 12 to 14 months. You think it’s 60.

The thing we’re in consensus on is Jason’s articulation of how to manage the messaging here. The messaging from corporate America will be bland, with a slight hint of upside on the stock, but not being so direct that it alienates all your employees. That is going to be corporate speak for the next 2 years. Totally agree.

What we don’t agree on—and you’re exactly right, Jason—is whether it’s 12 to 15 months, in which case it could be a significant change in terms of employment, or whether it’s my theory: 60 months, 5 years, in which case it’ll be more gradual.

9. AI's Next Breakthrough Gets Priced

When will OpenAI achieve AGI: before 2030 or after 2030?

That’s easy. It’ll achieve AGI whenever it suits Sam Altman in his negotiations with Microsoft over the OpenAI–AGI term in that contract to declare it to be AGI, because it’s a meaningless, ill-defined term that will be used for economic leverage. Go team. That’s it.

Stepping back, because maybe listeners don’t know, there is the whole big-assed AGI discussion: When will it happen? What does it mean? No one can quite define what it is. No one can quite define what happened.

Then, oddly enough, there’s a term in the Microsoft–OpenAI relationship, which now appears to be quite contentious. It says, I think—I’m trying to remember—when AGI is achieved, I can’t remember which way the leverage moves.

I’m willing to predict that that term will be exploited by one of the 2 parties. Someone now finally has an economic reason to give a shit what AGI is and when it can be activated.

So my guess is that the determination of what it is will be driven by that contract rather than any theoretical BS, fear-of-the-world kind of stuff.

Listen, I think Elon Musk calls the ball, but he's just always optimistic about exactly when it's going to launch across all his companies. But he always calls the ball, right? I mean, he founded OpenAI too, right? The guy's pretty good. All the trillion-dollar-ish ones he founded, right? He said 2026.

So what I think—and again, I'm not an expert—is it will feel like AGI in 2026, and the really smart guys will agree we're there around 2028. This is just me. I think he's right. He's so good, right? So I'm betting 2026 it feels like it. In 2028, we agree, unfortunately, we're there, right?

It's a nice one. It's an over-under, so I'm under 2030. I'm with Jason on that one. Okay, will Trump cut corporate tax this year?

No. Easy bet if you read the docs. Basically, if you look at the One Big Beautiful Bill—I can't believe I said that without laughing—the corporate tax rate was changed permanently in 2017, so there's no way to touch it now. It's 21%; it's not going to be changed.

I think the current version of the bill doesn't include another change in the corporate tax rate. There are some minor second-order changes to corporate tax around international taxation—GILTI, all the stuff that makes your head hurt when you even try and understand it. There's no change to the taxation rate, and there's no need to because, unlike the personal income tax changes in 2017, the corporate tax was a permanent change.

Yeah, certainly my taxes are going up. Rory's and my taxes are going way up under the Trump bill, because we can't deduct California taxes anymore from federal taxes. So our taxes are going way up, just like they did under the first Trump. Under the first Trump, they got rid of SALT, so our taxes went up, right? Now our taxes are going again because, as partnerships, we're not going to be able to deduct our California taxes against our federal taxes anymore. So Trump just doesn't care about California, nor probably should he, right?

How much are they going up? Just so I know.

A lot.

You'll be fine, Jason. You'll be fine.

No, but it is interesting. Listen, I'm not into politics, but it is interesting that under both Trump regimes, my taxes have gone way up.

Absolutely.

Way up because of Trump. He doesn't care about California. That's the main reason, right?

Right. And if you run the demographics as wealthy inhabitants of New York and California, and you look at their voting propensity, I'm sure there was some guy in the House Ways and Means Committee who, when they realized, “Oh, this really sticks it to those rich guys on the coast,” was like, “This is the only damn tax cut we're ever going to support. Let's push it through.” So, yeah, I hear you, man. But oh well.

Yeah. But by how much will the taxes go up? Rory's better than me. 6% under Trump?

But that's personal. I don't have a sense of my personal taxes.

Yeah.

Because of the pass-through entity tax, you'll no longer get a pass-through entity tax deduction in California, so I think our taxes will go up. Plus, California's raising it another 1-point-something percent, so I think our taxes are going to go up another 7%. No, that's 7% out of the 100%, not 7% higher, which I'd be cool with. It's another 7% we're going to be paying this coming year. Hooray.

Right. First of all, no one's going to cry for us, and California's still a great place to live, so we'll figure it out.

Final Kalshi quick-fire. When will there be a half-trillionaire, so someone worth $500 billion? Will it be before 2026 or after 2026?

I'll take after, easily.

I mean, isn't it just tied to how the stock market essentially performs over the next year, right?

Yes.

It's if you think the stock market—if you're really bullish on the market—

Yeah.

Well, we need to see double-digit growth in several stocks, but overall in Nasdaq over the next couple of years, right? We need to see a return to that double-digit growth rate, right? I think this poll is actually pretty good, the 38%. That's my gut. What are the odds that we return to the great growth rates overall in Nasdaq that we saw before this instability? 38% sounds about right, right? That's what my gut says. But I'm actually making the bet it's higher. I'm all in on the market.

Absolutely.

So I'm betting your 2026 number, even though it's not really consistent with reversion to the mean for gains of public equities, right? We can't have this growth rate forever, can we?

Wow.

With one caveat. I thought this was bullshit when we started it a week ago, but now I'm thinking of making a bet on Kalshi, which just shows how easily it works. My bet is it's well after 2027, for exactly the reasons you articulated, Jason. I just think it's hard to assume that the medium-term continuation of the same level of growth is going to happen.

Look, from 2010 to this year, it's been an amazing stock market for 15 years. Now, when you look at it, it's just hard to extrapolate continued growth at the same level, right? And I think one of the things you learn in the data is it's almost impossible to predict the stock market over 1 month, 6 months, or 12 months. But over 5 or 10 years, the correlation between entry valuation and ultimate return is pretty high.

Entry valuations are high, so statistically—I mean, Vanguard published great work on this—over the next 10 years, your default assumption on the equity return from United States stocks should be much lower than it's been for the last decade. If you take that into account, then you're right: none of these guys is going to be a half-trillionaire by 2027. And then I stopped. The only caveat to that entire sentence is the only person who can sprint their way to a half-trillionaire in the next couple of years is obviously Elon.

Mm.

Because he has private stocks. The ability of the private market to mark up investments is as yet untrammeled by any form of reality. So when you own a slug of SpaceX, and you own a slug of xAI, and you own a slug of Twitter, it is entirely plausible that someone gives you such a big step-up that you have a paper net worth of north of a half-trillion dollars in the next 2 or 3 years.

Wow.

I don't think Microsoft or any of the public stocks are going to compound your way to the same level.

10. Unicorns Face A Long Reckoning

Final, final one. There are 646 US tech unicorns. How many are actually unicorns in reality?

I thought that was a really good one. That was, again, from Jason's paper—let's give him credit—from the Silicon Valley Bank work on unicorns. I think the data there says 20% to 30% max.

If you say you have to be roughly $100 million, roughly growing more than 20%, and vaguely add on near profitability, what they were saying is that percentage is, you know, mid-to-high 20s, early 30s. So, yeah, I think that's exactly correct.

What it means is 70% of those unicorns aren't worth a billion dollars or more. They're worth something, but they're sub-$100 million, subscale, subgrowth, subprofitability, so they're probably not worth a billion. And the overall return from the entire $2.7 trillion of equity will be driven by 5 or 6—not just decacorns, but hectocorns, whatever they're called. Because if they don't compound their way out, the average unicorn isn't going to get you there.

I would certainly agree. What's the exact question Harry's going to ask about unicorns?

646—it says 646. How many are actually unicorns?

You know, it's worth $1 billion in hard cash today.

Here's what I'm worried about. So, 20% was kind of what the SVB data said, right? Which sounds about right to us, right? It's an unfortunate number. It's smaller than we'd hope. It's smaller than the markdowns that GPs have taken, right?

The only thing I would say is—and I know we've talked about this, and I hope we do one of these and it changes—I'm just worried there aren't as many exits for these folks—

Totally.

—as there should be. And so I'm worried the number is half. Whatever the number is, whatever we calculate it to, I'm worried in practice it's half of that, when in 2021 it was twice that, right? Because there was so much liquidity for PE and others, right?

Yeah.

Jason, you're exactly right, which is why it's kind of this quantity: 646. And then you circle back to 2 good IPOs. Let's say we had 2 good IPOs a week for the next year. That's 100 good exits, right? If they were all going to make it, that would take 6 years to clear the total balance of unicorns.

Yeah.

It just brings it home. You're exactly right. There will be some value from this. There are companies, but the bar on exits that's now knowable and achievable is a $200–300 million-plus valuation with 30% growth. It's a relatively small number of the 646 in the herd that's going to make that.

The only fun thing I would say—

And it's going to take a long time.

In 2021, at the peak, we had an IPO a day.

Yep.

So, clearly, the markets can absorb it, right? If these companies all reaccelerated to growth, we've already done this, right? 2021 was—it sounds crazy today—an IPO a day. In 2021, it was an IPO a day. You couldn't even keep up.

Totally.

TechCrunch said in 2021, “We won't even cover companies at a billion. It has to be $2 billion and up to write a post,” right? It was so crazy.

No, that's actually an excellent point. I shouldn't have throttled it at two a week.

Two a week would be great.

If the appetite comes back—which is why it's really good that both of these companies are performing—more of them can go public. It isn't a question of whether the liquidity is there. It isn't a question of whether the ability of bankers to get transactions done is there. To your point, Jason, the real question is how many of that 646 meet the new profile of $200 million-plus, 30% growth, and profitability.

Mm.

And that's where the culling of the herd will take place.

Boys, I always love this. You have been fantastic. Thank you so much for joining me. It's slightly earlier, also. God, I love you Californians doing early mornings. This has been amazing.

Can I get back to Jason's point? We work over here, dude. Sorry, that was mean.

You're a traitor. You're a traitor. You're Irish. How could you do this? You beauty.

You mean leave England? Let me explain. I could give you 800 years of history and explain why, but let's not. We don't have time for that, Harry. I can do it just fine.

Oh, right—we miss you in the UK.

Oh.

We miss you, Rory. Yeah.

That's a longer subject. And no, you don't.

Boys, you're amazing. Thank you so much.

Take care.

All right, rock on.