20VC：为什么光靠基金回报制造者已经不够 | 为什么 Sequoia 在最糟糕的时点采取了最佳策略 | 如今做好 Series A 和 B 需要什么 | Benchmark 领投 Manus 轮：美国基金是否应投资中国 AI

Fabrice Grinda

1X is not good enough for me anymore at this point in life. It's not worth it. I want 3X the fund.

Rory O’Driscoll

The problem with Series B is, if you get it wrong, you end up effectively paying Series B prices for Series A risk. And it's exactly the same now at Series A: you're paying Series A prices for Series C risk. But the whole point of having to be good at this job is being able to figure out which is which.

Guys, I am so excited for this. First, thank you so much for joining me today. There are 3 of you, so I'm just going to dive straight in. We have more LPs listening than I realized, because I get pinged by all of them.

I wanted to start with: what the fuck are we seeing in venture right now? What is your 1 takeaway from the last week? Fabrice, you're our new joiner. When you think about what you've seen in venture in the last week that has struck you, what is it?

1. Venture Enters The AI Bubble

Fabrice Grinda

I'd say what struck me in venture is that we're still in the middle of an AI bubble, if you look at the amount of money that's going into AI versus every other category in terms of valuations raised and the number of companies getting funded. I thought we were at the peak a year ago, and then venture did $100 billion in investments in the category. So it doubled from Q1 to Q4.

The other categories aren't seeing much love, and I think venture as an asset class has been somewhat in the doldrums because LPs have felt overexposed. They haven't had distributions in 2022, 2023, 2024, and frankly, 2025. We were hoping the markets were going to reopen, but so far the IPO window hasn't reopened and the spigot of M&A hasn't reopened.

I'd say it's become an unloved asset class. I think it's the best time to invest. But in a way, investing in venture right now is contrarian, and I think you should be investing in funds that are not going after the all-AI-all-the-time category.

Rory O’Driscoll

Yeah, I agree on the timeframe. The big comment, stepping back, is that nobody cares about our problems. We're VCs, right? One thing I've learned is that when you start bitching and moaning, you need to realize, step back, and talk to anyone else on the planet. They're like, “Let me get this straight. You have this wonderfully interesting job. You have reasonably good current comp. You get paid a percentage of the upside. You have visibility over 3 years in your comp. Nobody cares about our problems.”

And then when you start explaining that our number 1 problem is that valuations are too high, they go, “Let me get this straight. Your number 1 problem is that you have to pay me, the entrepreneur, more money for my company? I like your problems. I'm glad you have problems. I want you to have more problems.”

So let's start with that. That said, I do think it's a weirdly tough time right now because it's simultaneously a hard time to put out new money because things are expensive, and a hard time to get money back because exits aren't there. If you think of the 2-by-2 quadrant, normally 1 of those things is hard and 1 is easy, and it's good.

In 2021, it was easy to make money, easy to get money back, and hard to put money out. It was expensive. Right now, we have the low-low quadrant. It's very expensive to do these new, great deals and, at the same time, exits are tight. So it's a tough time.

But again, after I say that, I go back to first principles: nobody cares about my problem.

I just had a conversation with one of the largest endowment funds in the world, and they said, “The lack of liquidity—the big question that we have around that is: is it a temporary adjustment in the venture ecosystem and public markets, or is it a permanent structural shift in company building, maturation stages, and liquidity cycles?

“If it is the latter, we cannot be in this asset class any longer to the extent that we have been. If it's the former, we will be patient.”

Rory O’Driscoll

The sentence has the answer. The only time it will change is when they get out of the asset classes. That's what we said 2 weeks ago: while they keep putting money into the asset class, it will remain a stay-private-for-longer asset class. And once the capital in the asset class reduces, then we'll go back to going public earlier.

Economics is about seeking equilibria, and things push things back into balance. So the very statement sets up the solution.

2. The AI Gold Rush Accelerates

Jason Lemkin

I would just say 2 things. Harry, I think you asked what's changed in the last week or something. What was the question? Some version of that? I'll tell you my microlearning.

I think even in VC and LP land, we live in this new-news cycle. I had a company close a unicorn round on—I think April 7th. There was actually a downturn in venture because of the Nasdaq, right? We had this Trump drama. I think the stock market fell 15%, didn't it? The Nasdaq or something.

Everyone was freaking out in growth-stage VCs for at least 3 days. And already, this company that did a unicorn round has an offer to do a top-up round at an even higher valuation. We forget that it used to be, when there was a downturn or something, it would shock the system for at least a few months, right? Now the shock doesn't even last until the next 20VC comes out, right?

I hear all this stuff, but all I see is a gold rush. All I see is a gold rush in the San Francisco Bay Area. OpenAI says they're going to grow 1,000% by 2029. Everyone's just looking to make an insane amount of money, and they don't care about the nickels and dimes and the casualties, and they will deploy every dollar they can.

I think we can talk about liquidity and all this, but I honestly think that today is a much bigger gold rush than 2021 or other times we can think of. I've never seen such a gold rush, because this gold rush goes all the way down to high school kids. This gold rush is up and down the stack of tech, and it's both fun and a little cynical and a little crazy.

In the 2020 gold rush, you had to be a crusty old B2B guy who had been doing this for years or have something interesting in consumer. Now every 17-year-old kid is dropping out. Harry, you know what this is like.

But they’re dropping out, and they’re in the gold rush, and it’s glorious and weird and risky and full of baloney, but I’ve never seen anything like this gold rush. Never. Never, right?

Again, we talked about this before. I honestly think almost every GP and market will deploy every dollar during a gold rush. This is the way booms work, right? You will deploy every nickel possible in the boom, and then we pull back. For those 3 days, there was a pullback in growth, April 7th through 10th, and then back to the gold rush.

Fabrice, are you not concerned that, given that gold rush realization and the speed of revenue trajectories being what they are today, I have new AI companies like Manus and Lovable in my portfolio who are scaling at the most insane rates? Then I’m in ERP systems for concrete businesses that are going from $1 million to $4 million, and I’m going, “Lovable does that in about a day.” Do you not worry that you’re going to miss out on the gold rush?

Fabrice Grinda

90% of our investments remain in AI. It’s just that I try to be disciplined, and I try to invest in companies where I feel they have differentiated data sets, valid business models, and where I’m seeing less competition.

What worries me with the gold rush is that there are too many companies doing the exact same thing, going after the exact same category, and it feels unclear who’s going to win. Historically, I’ve waited until there was an emergent winner before investing, so I was happier to pay up. Frankly, you don’t necessarily even pay up because the prices adjust with traction over time. The traction valuation metrics start aligning as you get to a Series B, C, D, et cetera.

What worries me about the Lovable-type examples is that they can go from $0 to $18 million in ARR in 3 months, but I think they can go back in the other direction. I don’t know if people remember that AI company where, for a month, everyone basically just changed their profile photo using AI. Their ARR, or their MRR, went from $250,000 a month to $30 million, and then they had 99% churn a month later and were back to $500,000. They raised at that moment in time, and these things are riskier than people think they are.

Do I think that GitHub—or maybe GitHub is not the best example—can launch coding tools to compete with the likes of Lovable and Cursor, et cetera? Yeah, absolutely. I don’t mind missing the bubbly elements of the bubble.

I worry that people are underestimating the orthogonal risk of disruption from either OpenAI extending their stack. I built my own AI for fun. I built Fabrice AI, and I started using LangChain and Pinecone, et cetera. Then OpenAI released GPT-4o and became so much better, so I ripped out my entire backend stack and moved to OpenAI. I was using a different stack for text-to-voice, and I used Whisper.

I feel people are underestimating the risk of zeros, even though something can go from $0 to $100 million ARR very quickly. Usually, they’re priced accordingly. What do you need to underwrite so you can still get your 30% IRR? I think the risk of zeros is too high.

3. Speed Changes Fund Returns

I’m going to say this is for the whole group, but Josh Kopelman tweeted this week something that really struck me, and I respect the shit out of Josh. He said, “Speed matters in venture. How fast you win matters as much as how much you win. Being able to return in 10 years versus 17 years is a huge difference to the performance of your investors. 3X in 10 versus 3X in 17, obviously enormous, and people just talk about 3X funds.”

I wanted to hear your thoughts and lessons on the importance of speed of returns and how big that is.

Rory O’Driscoll

Yeah, I saw the tweet, and obviously it’s true because we’re being evaluated versus other uses of the money, and the common denominator of capital allocation is going to be some kind of rate of return per unit of time. The tweet is definitely true, so, yay.

I think the interesting thing is that it’s actually the hardest of the 3 things to control. Let me tell you how I mentally think about it. There are really 3 things that determine your return within a fund level.

There’s picking. Out of 20 deals, 4 of them have to be great deals. That’s 100% in your control. If you can’t get that right, you should lose your job.

Then the second thing is valuation on the way in and valuation on the way out. That’s controllable a little, but not as much. See the conversation on valuations going in now, and then on valuations on the way out.

We have an excellent 2014 fund that exited in 2021, and we always tell LPs, “One entire turn of that fund we don’t deserve.” It was multiple expansion. The other turns we do deserve, right, from our picking. Sometimes you get lucky on the upside on return, and sometimes you get unlucky on the downside. Valuation coming in, valuation coming out: a little bit of control.

The hardest thing to control is IRR because of the timing of those exits. Everyone now is suffering IRR degradation. They’re going to get the same return in 2026 that they thought they were going to get in 2024. It’s just very much outside your control.

It’s a very unsatisfactory answer, but the nature of our business is that we are judged on IRR. It’s very hard to impact it directly. You’ve got to get the other parts right.

To some extent, if you do your picking really well and you have reasonable control of valuations, then delayed exits simply mean more compounding and perhaps good IRRs. See, for example, the Founders Fund example from last time. They compounded for a decade and a half at 20%. Sadly, a lot of the time, pushed-out exits just mean lower IRRs. It sucks, but to some extent, it’s the hardest element to control.

Jason Lemkin

I was thinking about this a lot, Harry. It’s interesting that Josh wrote that. I was literally thinking about it this week, and I wanted to hear Rory’s thoughts. I thought it was very interesting that Fabrice led with IRR to describe his seed fund.

I was literally looking at my 2017 fund, and it is at 4.31X with a 32.56% IRR, okay? I was doing modeling of different scenarios, with the winners doing a certain amount. Honestly, unless I was delusional, the IRR would never go up.

It’s at 32.56%. A 5X fund, a 6X fund, even an 8X fund—let’s say this fund does 8X—would require a lot of things to go well. I’m not a chest beater. It would require a lot of things to go well, okay? But the IRR still wouldn’t beat 32%.

I asked my LPs if they cared, and here’s the thing: they didn’t care, okay? But I think maybe that was too simplistic an answer. I think the answer is that, for early-stage managers, we’re just looking at multiple, right?

I’ve never understood this since I started investing because I’m looking at this and I’m like, “I can’t do better.” I cannot do better than 30%. Even 32.56% is great, but that’s all the way to get to 4.3X versus the Nasdaq. It’s still a tough one, the IRR thing, right?

4. Secondaries Unlock Faster Returns

Fabrice Grinda

The cheat code I use for IRR—why Salt can get 30% IRR today—is the virtue of being diversified. I own 1% to 3% of any of the companies, and what I’ve been doing is getting secondaries on the way up.

I’ve been doing the anti-VC strategy of selling my winners. When I feel a company gets overvalued and I can no longer underwrite a 10X, I’m underwriting a 3X in the future, and a round is happening. Sequoia, Andreessen, and Greylock are all competing to get in, and the founder doesn’t want dilution of 15% a pop, so it’ll take 30% primary and 15% secondary. I’ve been doing secondaries.

Now that you have things like SharesPost and Forge and secondary marketplaces, the vast majority of my exits in the last 3 years have come from secondaries.

Jason Lemkin

Yeah, that’s how you get it. You have to get the secondaries to get that IRR doing what you’re doing, right? Nothing wrong with it, right?

There’s another tweet, Harry, and it was talking about tax efficiency at the GP versus the LP level, right? For me, I’m not selfish. QSBS is good, don’t get me wrong, but I have some personal liquidity. I’d rather go long, right? It’s so much more valuable to me, selfishly. I don’t care about IRR as a GP, right? If I can get another X as a GP, tax-deferred or tax-free, whoa. I can’t beat that, right?

Rory O’Driscoll

Everyone wants that. Coming back to it, to be very clear, I don’t think the algorithm is to maximize IRR. I think, subject to some caveats, the real algorithm is to maximize multiple subject to a constraint on IRR.

No one wants you to maximize your multiple by holding at 7%. But if you’re at 32% and the cost of funds, or the target for your sector risk-adjusted, is 20%, and you can hold another year and get 20% for that extra year, which brings down your overall IRR, you should do it.

Jason Lemkin

Well, explain that to me. My IQ isn’t that low, but I don’t get it.

Why should I do it if that only matches their cost of capital versus Nasdaq? Why should I do that?

Rory O’Driscoll

Okay, 21 then. 20, 21. You're right, 21, right?

Jason Lemkin

Yeah. Yeah. No, I'm not challenging you. I'm just trying—

I'm honestly just trying to learn, right?

Rory O’Driscoll

Anything at or above the target return, right? Because it’s extraordinarily hard to find other places to compound money at 20% plus. So I don’t think the algorithm is to pick the fund with the most IRR, the highest IRR, because you often see that in a single year, 100%, but it’s not sustainable.

I think a smart LP—and they are, on aggregate, smart—is looking for 700–800 basis points above small-cap, pretty consistently. So 20% plus, you're probably crossing the cost of capital, and therefore, to your point, short-changing a multiple for the sake of optimizing IRR is a mistake.

For you and for the investors, right? Because the truth is this: what you end up doing—and I want to come back to Fabrice's thing in a second about when you should take secondary—is selling the company that I know really well, where I've been on the board, I have a pretty good sense of what it's doing, and I'm 6 years in, to reinvest at a slightly higher revenue multiple in a company I know nothing about, right? It's such a risk escalation.

If you have good shares, hold onto them. Now, Fabrice is right; there are a couple of reasons that you don't. The first is the institutional imperative. If you or your LPs need capital back, then give them capital back to show you have a pulse.

The other thing is, if people offer you—and this is where it is tricky, and I'd love to get Fabrice or Harry, your thoughts on it—when people offer you, quote, “a crazy price,” where you kind of go, “I believe in this company, I love it, but I'm getting, I don't know, 2 years' forward credit for revenue,” at what point do you say, “Hey, even though I'm a believer, it's a smart thing to take some money off the table”? I think sometimes it can be.

Fabrice Grinda

By the way, these are the only companies I can sell, right? No one's buying your dogs, right? The only companies you can sell are the ones that everyone knows are winners and thinks are amazing. Look, I sold so many companies at 100X ARR in 2021. I love the founder, I love the company, I love the traction, I love everything.

But the price is too insane. What I need to believe to underwrite this valuation is every star in the multiverse aligning, and I just don't see that happening. And that's even just to justify the valuation, let alone get a 3X or a 10X from here.

And, by the way, I disagree with you when I see that a company is only at 2X or 3X from where it is, because on average we've been shooting for the 10Xs earlier. I'm very happy to recycle.

Jason Lemkin

My problem with selling in the secondary is that, because you have so many names, there's not pressure on every number to return the fund, okay? I feel like, for me, I want every winner to return 3 times the fund, not 1X. Because if I just return the fund—let's say that's my first distribution—I don't make any money.

1X is not good enough for me anymore at this point in life. Honestly, it's not. Because, listen, maybe it makes someone on Twitter happy, but that doesn't even put me into real carry mode, does it? 1X the fund, right? It's not worth it. I want 3X the fund. It's just not worth it.

But Jason, what I love about you is, A, I'm always learning, but you said last week, “If I know I can 5X a check, I will do it.”

Jason Lemkin

Yeah.

And then you're also saying this week, “1X the fund isn't enough; I need to 3X it.” Those, to me, seem paradoxical.

Jason Lemkin

I think they're paradoxical, and they're both true. Imagine you're in carry mode, right? Then you'll be like, “God, if I'd only put 5 million in Lovable, and it went from $1 billion to $5 billion, that's another $20 million, that's another 5 million bucks in my pocket.”

I mean, that doesn't go far in London, but I could take some good holidays. I can't buy a good flat, but I can live well in Monaco or wherever for 5 million. See, that's the weird thing I'm still trying to learn: you want your winners to be huge, but the ones that aren't winners, the ones in that next bracket below, just making money on them is terrific, right?

And the other weird thing I'm still learning is that, in that 2017 fund we talked about, I got an extra deal in. I got an extra deal in 4 years later that's almost a fund returner. An extra deal. So think about this extra deal in that outlying year of the fund. I just think the extra money is where you can make money too, right? It's not always in having the one huge winner.

The extra money, literally, you can buy at least a decent flat in London for 5 million, right? Especially if you're like Fabrice and you pay no taxes, then your 5 million goes further.

Rory O’Driscoll

When Marie Antoinette took this attitude, she ended up with her head chopped off. So, this is why no one cries for us, Jason. Duly noted that 5 million bucks won't carry you far enough. I'm still willing to get out of bed for a million bucks. I just wanna be clear on that. With or without the table.

Jason Lemkin

Well, I'm having some fun.

Rory O’Driscoll

It's hard work, Harry.

Jason Lemkin

You said that venture isn't about little numbers, is it? It's not about little numbers, right?

5. Holding Winners Compounds Returns

Rory O’Driscoll

No, you're exactly right. And that is the part that I agreed with you more at the margin than Fabrice: it's so hard to get a winner that, if you do think it can run, the compounding in the out years is the only way in which the venture thing really, really works commensurate with other equity businesses.

We're not playing with big dollars like the PE guys. Most of our stuff doesn't work, unlike the PE guys. The only thing that's good about this business is that occasionally you can make really great companies, and ending up holding on to lots of them is the way forward. That's one of those almost-always-true rules.

There were times, however, to Fabrice's point, where it's not. 2021 was one of those periods where that's not a good rule because everything is so overvalued that they're never gonna return to those valuations. Most of the time, your compounding on your winners covers up for your mistakes, so bias long.

It's why Sequoia did the Evergreen Fund thing, which was a brilliant idea. Unfortunately, in the only year in the last 20 where it was the wrong timing, right? But fundamentally, across 20 or 30 years, it's the correct insight, and they've had it from way back when, when they did Cisco, and you distributed that $200 million before it went public. Anyone who's still holding their stock feels pretty good.

Fabrice Grinda

I think it's different based on portfolio construction, right? You guys have concentrated funds. I don't, right? So, in my case, 2% of my deals follow the power law and return the fund 1X. So even a 100X doesn't return my fund, right?

Because I have 500 deals per fund, 2% of the deals return me 1X, and these are, like, 46X on average. 8% of the deals return another 1X, and these, I think, were 8X on average.

Then I have the remaining 90% actually make money, and 40%, a bit less than half of those deals, return another 1X. So for me, it's 1X, 1X, 1X, and that's how I end up with 3X and the 30% IRR, because I DPI early because of the secondaries on the way up.

And that's been true for the last 28 years, right? I haven't seen it change despite the last 3 bad years. So it's just a different way of playing the game. But so far it works, and it keeps working.

Jason Lemkin

I think the right question to ask if I'm an LP, actually, is: Are the IRRs actually reliable? I'm an LP in several funds, and I look at the underlying assets and I go, “Pfft, I know that company. I know that company. I know that. These are not where they're priced.” And actually, do not hang your hat on the IRRs of the managers, because they're not even reliable.

Bill Gurley

You're asking the very basic question: Are the numbers on the page correct? It's hard to know. I say that not being glib. We're in this weird period where we appear to be all, quote-unquote, “creating value.” We're getting unicorns, we're getting markups. Everything's working. These companies appear to be doing really well.

But, as a very astute friend of mine in the late-stage business said, because there are relatively few IPOs and M&As, there's no feedback loop. We're grading each other's exams, and we're all saying we're getting As, but the teacher hasn't graded the test yet. The teacher appears to be on strike right now, prohibited from doing anything, or caught up in antitrust.

Until the teacher grades the test, we're all just saying, “Yay, I'm great,” because Jason said I'm great and Fabrice said I'm great. At some point, all these companies are gonna have to have that horrible moment.

I've lived through it in some companies. That's where you file the S-1 and have that moment where you've got it on file, cold, and no one's seen it. Tomorrow morning, you're going to unveil the numbers, and it's like you're going to take your clothes off and everyone's going to see what you really have.

That's when we'll find out what things have worked. Until then, it's an ungraded test where everyone's getting As.

Jason Lemkin

Rory, do you agree with the Gurley statement that you should go public earlier? It's great to be public for discipline. Everyone should go public as soon as possible—that's the theory.

6. Public Markets Need Earlier Exits

Rory O'Driscoll

Bill's in an odd quadrant: pro going public early, which I agree with, but anti-IPOs. I think companies stay private because they can, and as long as they can, they will. At some point, I think we'd all be better off if they could go public earlier, right?

That will happen, as I've said. People don't do what they should; people do what they must. When they can't get cheap private capital, they'll all go public. And, yeah, being public has some good strengths. It does force discipline earlier.

It's not perfect, because you also have to deal with activists. There are some good reasons why people stay private. A lot of being public is a bit of a pain in the ass. But I do think IPOs will come back, and come back earlier, once capital gets withdrawn from the private ecosystem, which I think it will.

Fabrice Grinda

You know, when I was a kid, because I was crazy, my dream was to be a public tech CEO. Then I built 3 large venture-backed companies, 2 of which ended up being public, and I was like, "The last thing I would ever want is to be public ever again."

In fact, that took all the fun out of being a founder for me. All of a sudden, having to create the annual budget, the quarterly budget, and the update to the quarterly budget—literally, if I could never go public as a tech founder, I would do that. It is so painful and so expensive in terms of time.

All your information's out there. It's bureaucratic. It slows everything down within the company. I would rather not go public.

Rory O'Driscoll

There are significant negatives around being public that I think we have to fix as a country if we're going to have a more successful, dynamic ecosystem. But I also think the cost of that capital is so much lower in most times than what you can get privately—not for the last couple of years, but across 2 or 3 cycles, I think.

It just logically makes sense that liquid money should have a lower cost of capital than illiquid money. When I say it like that, it's pretty obvious, right? It is absurd to think that one person is getting capital that they can trade every day, and another company is getting capital where they're locked up for 5 years.

How in God's green earth, if the companies are the same, does the first company not have a lower cost of capital? But right now it doesn't. It's a point-in-time absurdity, and it will change.

Jason Lemkin

Everyone I know, directly or indirectly, who had a subscale IPO—who tried to IPO in the $100 million–$200 million range, with 20% or 30% growth today, or wherever they are—they're all miserable. They're miserable. They have to be profitable. They're not enjoying it.

Great, you have a $600 million or $700 million market cap. It's sort of illiquid as a founder anyway, right? You could sell a little bit, right? But again, it doesn't go that far in London, right? You could sell. You're on a 10b5-1 plan, selling a couple million bucks of stock a year, right?

It's not fun to be a $600 million or $800 million market cap with no analyst coverage, no liquidity, and a 3× multiple, right? With no one to buy you, and miserable employees who now know exactly what their equity's worth to the nanocent, right?

Sure, if you can go public at a—what was HubSpot? I mean, Rory, you guys invested in HubSpot. That was a rocket ship. It IPO'd at $100 million, but I think it was growing 60% at $100 million or something like that, right?

Rory O'Driscoll

Right. I mean, look—

Jason Lemkin

Those deals no one wants to do today. 60% growth at $100 million, right?

Rory O'Driscoll

There's a shift between those 2 examples, to be clear. A company at $100 million or $150 million with 50%, 60%, or 70% growth should be able to go public, is my belief. In other words, $500 million's too high.

Jason Lemkin

Yeah.

Rory O'Driscoll

You're exactly right, however. $200 million growing at 30%, yeah, the math's starting not to work depending on the multiples, right? So—

Jason Lemkin

And your life's miserable. As a CEO—

Rory O'Driscoll

So—

Jason Lemkin

Your life is miserable, I think, right?

Rory O'Driscoll

No. The bar should come down from today, where it's effectively 0. In the year of the great dot-com explosion, it was literally 350 IPOs, with median trailing revenue of $18 million, and that was clearly too early.

Coming back—actually, an interesting comment—coming back to IRRs, that was the exit environment where a bunch of 1996 funds posted 6× net, 100%-plus IRR funds.

Yeah.

Rory O'Driscoll

It was the best of times, as Mr. Dickens would say.

People are starting to realize that. We saw the CEO of Discord, and we saw the CEO of Ironclad both leave this week. Is the realization—the penny dropping—that, shit, being a public company CEO sucks? I don't want to do it.

Jason Lemkin

90% of B2B companies that IPO have a founder CEO at the top.

Rory O'Driscoll

Yeah.

Jason Lemkin

90%. Will that go down with 20-year life cycles, now that it takes so much longer to get there? So that's my number. What I've learned is you really do have to reinvent yourself as a CEO every 5 years and sign up for another tour of duty.

Rory O'Driscoll

Yeah.

Jason Lemkin

And so, if you squint, a lot of these turnovers happen in these 4- to 5-year cycles, right? Thinking of the Discord point, is Jason Citron coming to Jesus and saying, "Listen, I don't want to be CEO post-IPO"?

If you're a founder and you like big teams and you like scaling, but you don't want to do the IPO, leaving 12 months before the IPO is the right time. Right? You like scaling, you like people.

Rory O'Driscoll

Yeah.

Jason Lemkin

You just don't want to deal with Wall Street. That's the perfect time to leave, like Discord, right? And hand it to the professional guy, right?

When I look at the other ones, my concern is that they go into terminal decay, but that's a different issue. I just think they go into terminal decay.

Rory O'Driscoll

They definitely do. The hardest thing to do is a CEO change when the founder CEO doesn't want to do it, but for whatever reason you think they have, you should make that change.

And for me, that's an extraordinarily high bar, because a founder CEO with some managerial limitations usually performs a lot better than a reasonably good manager with no founding DNA. I'm not one of these grandiose people who says, "Oh, we'll never change a founder." I think statements that use the word "never" just don't tend to work as well.

When you take longer to get to the same place, you're just going to have more change. In 1999, as I said, it was a sprint. You founded the company and got public in 3 years. Most people can do a sprint.

Then it was 10 years. A fair number of people can do the 10,000. Now we've converted this into a marathon. Oh, and by the way, at the end you have to run a few more sprints just for fun.

It's not surprising that when it's a 10- or 15-year journey, lots more people tap out. Life takes care of a lot of this. People get older. People have other things going on. So, the longer this private holding-company period lasts, the more of this kind of dynamic you're going to have to wrestle with.

I find it easier to pick founders with that realization than with the shorter-term horizon. Very specifically, because you have to be a fucking psychopath. I am deranged. I run 2 marathons a week. I'm an addict in every way.

I have more energy for 20VC than I've ever had before, 11 years in. I've done 3. There is nothing normal about me, and I look for nothing normal in the founders.

Jason Lemkin

You can't make money on normal people, unfortunately.

No. But you can if it's 5-year windows or 7-year windows where you can scrape by. With the 15-year window, you have to be so deranged and obsessed by a problem. It's your unwavering life's work. I find it almost easier to pick.

Jason Lemkin

You know, it's funny. Harry, when you and I were in London this last summer at SaaStr Europa, there was a CEO speaking, and he told me that he's in a WhatsApp group. They're at scale; they're at 9 figures in revenue. He said they're a 20-percenter club. They're all telling each other—

Rory O'Driscoll

Yeah.

Jason Lemkin

—how 20% growth is great today. I get it as a human, but you can't invest in these people. They're in the 20-percenter club. They've all convinced themselves, as a form of group therapy, that 20% growth at 9 figures in revenue is as good as it gets, right?

You need people who won't join, who log out of that WhatsApp group.

7. Technical Obsolescence Comes Faster

I think there's a really interesting question, which is: are we going to see the acceleration of terminal decay? What I mean by that is, if you look at companies like Pinecone, they went up and they went down very, very quickly with changes in technology cycles.

Squarespace, Wix, any of the website builders. I'm not picking on them or being horrible, but your Lovable and your Bolt are absolutely killing them on the consumer.

Are we going to see the acceleration in that terminal decay rate?

Rory O'Driscoll

Short answer: of course. That’s almost a given. Look, first, all technology companies have obsolescence written into them from day 1. It’s sad, but true. Anything before Microsoft and Apple at this stage isn’t here, and Coca-Cola has been cranking for 100 years, right? So obsolescence is inevitable in general.

On top of that, periods of acute technical disruption are going to increase the amount of technical obsolescence that companies face. Overlay that with, as we’ve been discussing, longer holding periods. I think I said this last week, but my number 1 fear is that the technology life cycle to obsolescence is now shorter than the holding period of privately held software companies, which means every company, at least 1 time before it gets to go public, will have an existential reinvent-itself, second-product crisis.

Jason Lemkin

It’s a huge issue.

Rory O'Driscoll

Yeah.

Jason Lemkin

All of them today—they’re all too old. All of them today. All these ones that scale are too old, aren’t they?

Rory O'Driscoll

Are you talking about VCs?

Jason Lemkin

No. Well, I thought you were talking about the portfolio companies. If it takes you 15 years to IPO, you’re so far architected before the AI age. No matter what agent you add on top of it, you’re having an existential crisis, right?

Rory O'Driscoll

The quick, snide, witty response would be SpaceX, but I would be intellectually honest enough to admit that would be a snide answer, right?

Jason Lemkin

Yeah.

Rory O'Driscoll

For really hard technical problems, you’ve probably got to get a 20- or 30-year run. But you’re exactly right, Jason. For core first-generation, what we refer to as plain-vanilla SaaS, it’s highly likely that 10 or 15 years in, if you haven’t exited, you have some degree of technical obsolescence.

Now, it’s not fatal, but there’s often a reinvention act required, which actually circles back to the founder comment. The ability to take that company and drive it to a new architecture, like Zuckerberg did with mobile, or add a second product is, in my view, what will separate the companies that peter off at $100 million at 20%, going to 15%, going to 10%, versus the companies that have a year or 2 bumping at 20%, get another product out the door, reaccelerate to 40%, and make it.

I have some of both in my portfolio, honestly. I have 1 in particular—I’m thinking of Lattice—and I’m just so impressed with the CEO. We hit a 20% year a year and a half ago. We did some acquisitions, and we really cranked in engineering. We didn’t cut engineering in the downturn; we doubled it. Now it’s back up to 40% plus.

But it’s going to be existentially tough, and it gets back to you guys. That’s going to take winner CEOs, wherever they come from: founder or hired.

Fabrice Grinda

By the way, I think that’s true in B2B SaaS. I don’t think that’s necessarily true in other categories, right? If I look at what my bread and butter has been—marketplaces with B2B and consumer-facing—the AI disruption is actually benefiting startup incumbents because they have the data moats.

It doesn’t benefit eBay because they’re slow and big, but it benefits an incumbent startup that already has liquidity, is the most efficient to operate, is at scale, and now is adding an AI layer. They have all the data to make better decisions, improve the funnels, et cetera. I don’t think in these cases they need to be at risk of disruption. They just improve.

Rory O'Driscoll

I totally agree, Fabrice. That’s proving once again that a diversified portfolio has that advantage. No, you’re exactly right.

Is eBay not advantaged because of its touch points to end consumers, the sheer distribution, and the remaining brand that it has? I would just argue that it has terrible internal policies, slow decision-making, and poor teams compared to startups.

Fabrice Grinda

No, I don’t think that’s the issue. I think the issue is actually simpler: they’re horizontal and multi-category. Even though they have all the information and all the categories, their tech stack isn’t built such that they’re best-in-class in every single vertical.

The collectibles and Pokémon marketplace is going to do better than eBay, and that’s true of every single category you can think of. We’re in a handbag marketplace called Rebag, and they have this AI where you take a photo and, boom, it tells you the model, whether it’s fake or not, the quality, and the price. Everything’s done.

Even though eBay has the data, its tech stack isn’t that flexible, so it can do best-in-class transactions in every vertical. It may just be a vertical-horizontal play where the verticals do better, but eBay will buy a lot of these.

The new team is amazing, for what it’s worth. They’re going back to the basics. They’re going back to collectibles and used, and not trying to be an Amazon clone.

Rory O'Driscoll

That’s a super-interesting and nuanced point, by the way, Fabrice. It’s equivalent to what you’re hearing in enterprise software: with AI, the deep verticals will do better than a broadly horizontal platform, because what you don’t need is just a horizontal transaction-management platform. You need to fully solve the problem, and it’s incredibly hard to solve the problem in insurance, solve the problem in retail, and solve the problem for manufacturing.

Fabrice Grinda

But here’s the interesting point: I think that’s true at the application level, but at the fundamental LLM level, the horizontal GPT may just win most categories. In the same way that Google won search reasonably writ large, except maybe Kayak for travel, I think OpenAI and ChatGPT win most of the LLM-type categories.

I used to use Midjourney, but DALL-E is so good and it’s part of my subscription. I just don’t use Midjourney anymore, and that’s true for many things. Now, it doesn’t end up being true in every category. With Cursor or Lovable, I’m not so sure. These are maybe still verticals that work.

But at the application layer, I think the hyper-vertical, where you solve the problem end to end, wins over the horizontal.

I think all of us are idiots, by the way. I spoke to a friend of mine who is doing a $250 million SPV into OpenAI at $300 billion, and you’re looking at that going, “Christ, there is a nonzero chance you’re going to 3–4x that SPV on a quarter of a billion with a 20% carry.” It’s not a bad deal. I’d do it.

Jason Lemkin

It’s not a bad deal.

Not a bad deal at all.

Jason Lemkin

Especially because it’s an SPV. Going to Rory’s prior point, you don’t have to talk about it if it doesn’t work out. There’s no downside to the perfectly constructed SPV, especially if there’s no GP commit. You just don’t talk about it.

A quarter of a billion is quite hard to hide, I think. I also want to mention incumbency reinvention.

Jason Lemkin

You put what you want in your LP report.

Your core LPs just get the main funds, and only the SPV guys get a separate distribution that I do over on Carta. I got some of the email addresses wrong.

Rory O'Driscoll

I don’t know how to break it to you, Jason. When you get a little bigger, you’ll hire a GC, and this will be the moment in the podcast where your GC has a heart attack. But it’s okay for now. Keep it clean. Keep it tight, Jason.

Jason Lemkin

Why do LPs that aren’t in the SPV need to know the data and performance of the SPV? They don’t, do they?

Rory, just get your inhaler. It’s okay. Breathe.

Jason Lemkin

The GC’s behind you. Just breathe.

I’m the most transparent person you’ll see. Forget about the big GCs. I see a lot of stuff swept under the rug, a lot of investment banks doing it. To your point, I see it, even from my little vantage point, right?

Rory O'Driscoll

Some people wear their heart on their sleeve. Jason wears his cynicism on his sleeve. That’s why I like it.

8. Incumbents Face AI Reinvention

Oh, dear. I totally agree with that. But we mentioned the reinvention of incumbents there. ServiceNow is up 24%, growing 20%—well, almost 20%—at $12 billion in ARR. What do we make of that? Bill McDermott coming out with another masterstroke? How do we evaluate this?

Jason Lemkin

First of all, at a meta level, we’re all trying to figure it out, and everyone in enterprise exaggerates how much AI has had an influence on their business, right? Mark Benioff is great, but he said they’ve had 500,000 transactions on Agentforce. We’ve had over 100,000 on SaaStr AI.

I don’t think 500,000—if you think through it, it sounds great at first blush, but it’s early, and he acknowledges that. It’s not a criticism; he acknowledges it’s early, right?

All the talk about ServiceNow and all of these agentic automations, I still think it’s early, right? On the other hand, we’re all wondering, do these guys all win the most? You can put a really, really good AI on top of almost any system of record. You can make Zendesk better.

ServiceNow is up 24%. Palantir, even though it’s kooky, is up. SAP, which was founded in the 1970s, I think, Rory, is up—14% growth at $32 billion in revenue, right? That’s the meta question: will they really benefit from AI? The data doesn’t totally support it.

What I actually think—but, Harry, you hinted at it with Squarespace and Wix—is that AI helps enterprises on balance and hurts SMB players on balance. The SMB players just don't have data that's as deep. They have a lot of data, but it's not as deep. You can disrupt them faster. The sales cycles are quicker.

I don't know this. We'll see this in 12 or 24 months, but we might see the enterprise overall get stronger with AI and the SMB leaders get weaker and weaker because they're disrupted faster.

Rory O'Driscoll

Yeah. No. Funny, I went and looked at the actual numbers on ServiceNow, and it clearly beat expectations, but plus or minus, it's been a little over 20% growth for 5 or 6 years. It's just a really well-run, market-dominant company. So I don't have an insight on the minor question of the jump based on Q1 versus actuals. I don't know what they were thinking or what they were worried about, and why it popped 24% on what was pretty much the prior logical estimate for revenue growth, right?

But zooming out to the Jason point, I think the bigger question is—there are 2 big questions. One is: What happens to these 2 or 3 ultra-large SaaS companies? Does AI help or hurt? My mental model has always been that, at any point in time, there are actually 3 different players. There's the pre-AI behemoth, which is ServiceNow. There's the AI teenager, which is typically a company that, pre-LLMs, had been building something in AI from about 2018 on. And then there is the post-LLM YC next-generation company.

And in all these spaces, that's roughly true. You can name the ServiceNow, you can name the mid-tier players, or you can name a bunch of new ones.

What's interesting about ServiceNow is they made a shrewd move. They bought one of the teenagers. They bought Moveworks just now, and they said, “Hey, we had an AI story before we bought a small company.” It was clearly getting some traction, but clearly not enough, and they decided, “Let's take $3 billion—what's that, 1% of your market cap?—and buy Moveworks and get really relevant in AI.” So I thought it was kind of a shrewd move. I think the market's saying—I don't think there's anything more profound in it than saying this is a company that looks like it's going to be a winner in the post-AI world that still has scale, that still has profit. Probably worth something.

Jason Lemkin

The one thing, Rory, I'd be curious what you thought about is what I think about in the middle, right—the SMB-to-enterprise box. You were an early investor in Box, right? And Aaron's all over AI, and I've been in the document-management space. There are fewer spaces that you could disrupt more with good AI than documents, because all our goals since the early days of documents were: How do we take these unstructured documents? Sure, we can OCR them, we can extract a little, but you couldn't do much with these documents, right? And this is why Aaron's in love with AI.

The question is, will it be enough for Box? I don't have an opinion; I'm just watching as a student, right?

Rory O'Driscoll

Sure.

Jason Lemkin

But I think it's in the middle. We don't know. Will this re-accelerate Box because Box is a more valuable app or not? But there's no question we have a CEO that's all over it, right? An S-tier CEO that's all over how to make a 2005 company with AI better, right? I'm watching as a case study.

Rory O'Driscoll

Got it. And, yeah, my comment would be, first, not only do I like Aaron personally a whole ton, but more importantly, he is one of the CEOs I most respect. And going back to the comment earlier on gritty and hard and sticking with it, that's a team that's been there now probably for 20-plus years doing the same job, never blinking when it got tough, never blinking when the activists showed up. So there's no one I want to win more.

I do think—and again, back to the thing—if Moveworks is a teenager, Box is clearly a young adult, right? And if what it takes to add AI at that stage is great focus from the CEO and the team and just driving change, I think Aaron will deliver it, right?

Jason Lemkin

Yeah. For sure.

Rory O'Driscoll

So I'm a big fan. I'm still a stockholder. Glad to be a stockholder. And another day, I could riff long and hard about the negative parts of being public. I've changed my thinking on how best to ensure founder CEOs can be successful in the public markets. He's done an amazing job.

I'm sorry, I'm naive, guys. You're right, Aaron's all over AI, and you're right, Jason, the document-management space is the most perfect space for it. Why is that not reflected in the excitement around their market cap? Their market cap has always been pretty depressing as a multiple. Sorry, Rory, and you can just recuse yourself.

Jason Lemkin

Rory would know more. I just think it's a question of what AI will do—it may not lead to breakout growth for some of these players, right? Has it for ServiceNow? Bill says it has for ServiceNow. I'm a little skeptical. I think it's just early, right? If it was as simple as having incredible AI on top of all the data you need, Box should win. Box should re-accelerate to 20% or 30%. I'm talking about a big public company like a VC. It should re-accelerate. It has all the AI, and it has the data.

Box has trillions of documents, right, that you can't even find. Now I can talk to my documents? It should—if we're shooting from the hip as investors—it should re-accelerate, but if it doesn't, then I'm trying to learn how AI will change it, right? Because Box is going to be, in 12 months, an order-of-magnitude-better application than it was 24 months ago. In 12 months, it will be an order of magnitude better—not a little bit better; it will be an order of magnitude better—and will that be reflected in growth? That's the meta question, right?

Well, the question is: Does that actually come out in value extraction? Sorry to interrupt you, Rory.

Jason Lemkin

That's the question.

Can they have a better product but actually just have margin degradation because they're paying more without charging more for it?

Rory O'Driscoll

I think in the end, if you deliver value, you'll get value. And you know what? Contrary to F. Scott Fitzgerald, I do believe there are second acts in American life. I genuinely do believe that AI could provide a strong boost to the Box value proposition, which needed it.

You give me grief on the stock price, but it's a compelling achievement to have done what they've done. The reason the stock price is always hard is that we're competing against Microsoft and Google, who give the damn thing away for free. The fact that they've built this profitable, wildly cash-flow-positive business competing against the 2 largest companies on the planet, who give the damn thing away for free—just hats off to them, smarty-pants.

But now the question is, can they re-accelerate? And I think the interesting thing is there are examples in the past, like, for example, Adobe way back in the late '90s. One of my partners used to work there, and she said they were stuck at $1 billion in revenue for 3 or 4 years, and then they got the unlock and they re-accelerated. I can't tell you what's going to happen; I'm not on the board.

But just as Bill McDermott is looking to grab onto AI, I think Aaron is wisely grabbing onto AI, and I'd prefer to be playing that hand than the guy doing $1 billion, $1.5 billion in revenue saying, “AI won't impact my business,” because he'll be gone in X months.

Rory, you said if you create value, you'll be able to extract value. I'm worried because you see people like Windsurf and Cursor, who are creating the most insane amount of value now, doing 1 billion lines of code a day, charging at Windsurf's price—I don't know what the prices are, but they were like $30 before. I think now they're $15 or $20. Are we moving into a world where there is this dislocation between value creation and value extraction?

Rory O'Driscoll

My guess is it's probably not. First of all, given the background noise, I don't think we need to worry about Windsurf's ability to, quote-unquote, extract value. They appear to have found a higher source of value to extract on the capital-market side. So I think they'll be doing fine if they get $3 billion. But on pricing, I thought it was a clever move.

I think you're going to see this. Yes, they'll lower the price, but they have tiered pricing. It's the conjoined question you were asking about, value pricing, which is very hard to do for a product like this. Having tiers of pricing where you get lots of people using the base product and then escalating steps of value as you deliver more value is probably the simplified version of, quote-unquote, value pricing.

In much the same way as OpenAI has $0, $20, $200, and $2,000, they're going to keep giving stuff away in the lower tiers, get you hooked, and upsell. I don't think it's a charitable act or an act of madness. I think it's probably a pretty shrewd pricing strategy.

Jason Lemkin

Well, listen, I think what they're doing is vaguely similar to what HubSpot has done, which is the low end. They're going more and more low-end. HubSpot 2 years ago didn't even have this Essentials edition, which is about 45% of the new customers—

Wow.

Jason Lemkin

And more enterprise. So if you look at Windsurf, to get going, it's simpler and cheaper than it was, right? But the average enterprise customer pays like $60 or $80 per seat, and they have 200, 300 people in their enterprise go-to-market motion, right? So they've got a barbell where everyone can get on this platform.

They're having incredible marketing from it, right? In 3 months, everyone says, “This is better than Cursor.” We can all have our own opinions. There are incredible marketing benefits, even for HubSpot, from this long tail, and then they've got 200, 300, 400 sellers selling 6- and 7-figure deals at 60 to 100 bucks a month per seat. That's a quietly better model than it looks, right? It probably can scale and absorb some significant costs from OpenAI and Anthropic at that level. It probably can.

Rory O'Driscoll

I worry that these things are going to be more competitive than people think, right? Why isn't GitHub competing with Cursor?

Jason Lemkin

They are.

Rory O'Driscoll

Yeah, but—

Jason Lemkin

It's Copilot versus Cursor, and Windsurf is number friggin' 1 on Microsoft's gotta-kill list. Number 1. It's all over. It is existential for GitHub to destroy them.

Rory O'Driscoll

Exactly.

Rory O'Driscoll

So again, this is the reason I've been avoiding these AI deals. It's kind of like 2021, where in every category you were going after, there were 8 great teams, well-funded. The very fact that there were that many actually killed the economics of the category until eventually a winner emerged.

The problem is, if you invest at a very high valuation in a company in a category where there were 8 people going after it and overspending on customer acquisition or offering too-low prices, it ultimately didn't lead, I think, to great outcomes for investors. I worry the same thing is going to happen here.

Jason Lemkin

And so while I do think eventually a winner or 2 will emerge, and they'll have the proper pricing power, and I do think in the long run you're able to extract value when you create value, there may be a lot of investor value destruction on the way up because people are competing on customer acquisition and price.

If this ends up being winner-takes-most, you're willing to do whatever it takes to win. As a result, you're willing to give up a lot of value on the way there, and it's going to take a lot of capital. That's what worries me about these models writ large, and that's why I've been staying on the sidelines, especially considering the valuations they've been raising.

I'd rather pay up when one of them seems to be the dominant winner, and I suspect that then price and traction will be more aligned.

I think I've said this before, but one of my friends led a round in one of the model companies at $4 billion, and now it's at $60 billion.

Jason Lemkin

Yeah.

And their multiple is 3.1×.

Jason Lemkin

Yeah.

And it's because they're shedding 9% a year in employee stock comp, and they've raised billions and billions and billions.

Jason Lemkin

Mm-hmm.

It's not really a venture-fundable asset in that respect.

Jason Lemkin

It's not the best venture—

Oh, that's a bad sigh. That's a bad sigh, Rory. Don't hit me. Come on.

Rory O'Driscoll

No, no, I hear you. Look, it might not be the best venture return, but I just think that, on these broad trends, not showing up—just walking away from the trend entirely—is too hard. I think they're still up 3×. I think they have upside from there. In retrospect, they're probably still glad they did it. They might be nervous about the burn.

We're probably saying some version of the same thing. The best of all worlds would be a wonderful tech trend and low capital availability, which means only 2 companies get funded, they slug it out, and they both make money. That's welcome to 2010, or even 1994, because I was in the business. It was awesome.

Jason Lemkin

Yeah.

Rory O'Driscoll

I'd still prefer a really strong tech trend and 5 or 6 competitors to no strong tech trend. I do love where we are from a tech perspective right now. This is some of the most amazing technology we've seen in 20 or 30 years.

We get so used to it. You sent me your questions, Harry. I just loaded them up to ChatGPT and then, my God, I'm an expert on everything on the podcast 10 minutes later. It's just a wonderful world here, so there's a huge amount of value happening, right?

But I do agree, it's incredibly tricky, figuring out how to play with just the capital. The capital is definitely massively eroding the returns. You find yourself as an investor trying to find a way to play the trends. We're gnawing at this every day. It's so hard.

Do you go a little earlier and try to catch it just before? The bad news is, you don't know who the winner is and there are 8. Do you pay up for the winner at $500 million or $600 million when there's only $1 million? Is that enough traction? It's damn hard. I'm not saying it's the only game in town per se, but it is the biggest game in town.

Jason Lemkin

For sure.

Rory O'Driscoll

And pushing away from the table entirely is a bit much.

9. Series A Inherits Seed Risk

Rory, if I was on your team, I would be saying to you, “Play the trends? Dude, we're pre-seed, seed, and A. Our game is backing generational-defining founders. We don't play trends. We just try and find the best founders in this business.” How would you respond to me if that was what I said to you?

Rory O'Driscoll

I'd say to you, at the pre-seed and seed, that's probably true. And by the way, I liked your tweet, just to turn it right back on you. All you have is 3 things. You said—I'll give you the 3 things you have per your tweet, and I'll contrast it to where we are. Slightly different.

You said there's only 3 things: “I've got awesome freaking founders, a roughly directionally correct market, and economics that make sense.” And you're exactly right. Those are the 3 things. But you did add—the second one is there, right? You do need to have some semblance of a directionally correct market.

We're probably paying a round and a half. By the time you get to an in-revenue A, which is where we play—early revenue with product-market fit—you have to have product-market fit. That's why we use the word.

If you pay up for a company that has product-market fit and then you lose it and have to reacquire it, by definition you've overpaid. By the time we invest, I want to at least say I believe that this is the right solution that has some element of product-market fit. We're not going to tear it up and start again on something else.

That happens and, as we discussed last week, it happens more in AI than anything else. But I can't afford the luxury of just saying, “These are meat-eating founders, and they're going to figure it out. Who the hell cares what they do?” I want to at least know they've locked into something that can hunt. And then the economics, like you, have to make sense.

Yeah, but I don't think you can find number 2 with number 3 today, really, at all, because I see these deals every day in San Francisco, Rory, and they're $400K in ARR. That's not product-market fit.

Rory O'Driscoll

Okay, you caught me.

That's the big problem, dude. Okay, you nailed it. Exactly. That is absolutely the issue we're wrestling with: by the time you have our first criteria of product-market fit and a good founder, the third criteria—valuation? No. That is why venture capital is hard. No, you're exactly right. Is the solution to try and do what you do and go pre-seed? I don't know.

I think you have to. I think you also have to, given the fact that pre-seed funds are now $400 million like mine. I can do the seed and the A. Good luck trying to take my best pre-seed, because I'm going to cling onto it as hard as I can.

I think that's why you're seeing Neil Masak? lead the seed for Windsurf, because there's a realization that there's no freaking way you can get in unless you are there.

Jason Lemkin

But even there, that one—I mean, this is a good question. When both Windsurf and Cursor, Anysphere, were seed-funded, they were radically different companies. They weren't even the same company. They were nothing.

I think I have the chronology right. When Greenoaks doubled down on Windsurf, it was Codium, which, essentially, the company abandoned 3 months ago. Every engineer was repurposed to build a better version of Cursor called Windsurf.

This is like betting on an S-tier founder, which Varun is, but those bets don't work out a lot. I don't think those bets work out most of the time, even with S-tier founders, right?

Rory O’Driscoll

The classic B2B investor is at $40K MRR, to your point, and is attached to a trend. The world is changing. It's AI, it's Web3, it's mobile, and they've attached to an early trend that even you couldn't see, like in the old days, right? There is early product-market fit at $400K ARR, right? It's just we used to be able to invest in those companies in the teens.

I'm sorry. In San Francisco, is there really early product-market fit? You're a YC company. You've got all your YC mates around you. You've left a firm, or you've left Square, or you've left whatever.

And you say, “Hey, come on, sign up for a $20K contract.” You get 20 $20K contracts.

That’s not early product-market fit. It’s having good friends with some money.

Fabrice Grinda

Agreed.

Jason Lemkin

I think 10 folks that weren’t in your batch might be, though.

Fabrice Grinda

Totally. And more importantly, Harry, it turns out it’s hard to make money, right? Tough shit. Yes, we’re paid to figure out which product-market fit is bullshit and which product-market fit is not.

I remember back when the round that we now do—the A—used to be the seed. The problem with Series B is that if you get it wrong, you end up effectively paying Series B prices for Series A risk. And it’s exactly the same now at the A: you’re paying Series A prices for seed risk.

But the whole point of having to be good at this job is being able to figure out which is which. There’s an element of the job that is picking, and it turns out that’s what they’re paying the comp and the salary for, and you’ve got to get it right. And if you don’t, you’re going to lose money.

In a more forgiving capital environment, your margin for error would be much higher, and that’s the really hard thing about today. You don’t have enough room in the price to bury a lot of errors, so you’ve got to get your picking much better and your win rate much better.

So, yeah, it’s harder to make money when there are 20 VC competitors than there were 3. What are you going to do?

I love it. Fabrice, I basically get a schooling every week from Rory. Last week I was a hypocrite, so this week is a definite improvement on that.

Rory O’Driscoll

I want to apologize. I felt “hypocrite” was wrong. I should have just said you’re incorrect.

No, Rory. I have a thoroughly inflated ego, and so it doesn’t really harm me. You also mentioned something in 1994. I was born in ’96, and so I wouldn’t have actually remembered that. But thank you.

Jason Calacanis

Well, maybe if you read.

Yeah.

Jason Calacanis

Maybe if you read anything that wasn’t on social media, you might actually—

Rory O’Driscoll

Jason, that generation doesn’t read. We know that.

Jason Calacanis

True.

Oh, books are dope.

Rory O’Driscoll

Yeah.

If it’s not a tweet, I’m out.

Rory O’Driscoll

Exactly.

Jason Calacanis

It was a reel.

Guys, I want to run through a couple of fundraising elements that I just have to hear your thoughts on, and then we’ll wrap. Manus raised $75 million at 4x its last valuation, led by Benchmark. It’s a Chinese company. I love the Benchmark guys, and they’re friends, so this is no shade, but should we be funding Chinese AI companies?

10. China Raises Geopolitical Risk

Jason Lemkin

The first thing I thought when I saw it was Rory’s point from before: we’re all taking more risk. So there’s a political element here, right? I’ll briefly address it, and then other folks can talk about it.

When I saw that, Rory’s point just rang in my ear. Would Benchmark—would anyone ordinarily want, today, in today’s world, to do a China-based AI company? No. But it’s disruptive AI, at least for now, so you’re taking more risk.

You’re taking more risk that you won’t get liquidity. You’re taking more risk that the government will take away your shares. You’re taking more risk that you can’t repatriate any earnings. I don’t even know all the risks, but so many folks have walked back from China that they walked in.

I just thought it’s taking more risk to get the massive outcome. The second thing I thought when I saw it—and this wasn’t really about Benchmark—is that, in general, nobody cares. Nobody cares if you’re selling weapons. People love defense tech that’s killing people.

Now, we could argue either side of it, but I haven’t talked to a VC that thinks there’s anything wrong with that. I’m sure there are. Everyone’s like, “I’m Mr. American Dynamism now.” And maybe that’s okay, but it was just a couple of months ago we were talking about how AI might kill us all and that we needed safety and control, and I haven’t heard a peep out of that since Anthropic was formed.

No one cares, right? But there’s risk—Rory’s point rang in my ear on risk, Rory.

Fabrice Grinda

Yeah. There’s so much to unpack in what you said and then what Jason said. And silence on some parts does not indicate consent.

So let’s now go back to it and say, on the decision, I think you led with the “should.” I want to avoid the “should they, shouldn’t they” question and playing geopolitical guru, because I’m not. My comment was, as a fund investor—the way I think of it is this: on an individual deal level, there’s quite a lot of risk, but they’re probably massively getting paid for the risk, and it’s an idiosyncratic risk.

So, probably from a pure portfolio management perspective, adding 1 of 20 deals that has this very weird risk, where you’re probably getting a lower price but have some exogenous political risk, could, from a pure finance perspective, be a good idea. I wouldn’t do it because there are 2 other criteria.

The second criterion is: are you taking individual deal risk, or are you taking firm risk, where the blowback from doing the deal slops over—not just from the individual 1 out of 20 deals into some impact on the firm? I can’t assess it, and to be clear, I’m not yet making a moral distinction, but I’m just saying, with so many people leaving and so much congressional pressure around that, I wouldn’t have had the courage to maybe bet the firm.

Or, at the very least, bet that I’m going to spend some portion of 2026 in front of Congress explaining this in a way that I… We saw very intelligent investors, Sequoia and people like that, get out of billions of dollars of value because they just wouldn’t want to be there. Having got out of something like that, I wouldn’t want to go back.

So I think institutionally, as a firm, I probably wouldn’t have done it. The last one is some vague moral stuff, and I’m deliberately punting on that, which is not to say… How do I put this?

Rory, your wife just said you couldn’t say anything in that regard, so—

Rory O’Driscoll

I get a list of banned topics at the start of every month, and that’s one on there. But I think there are very different risks from investing in a US-based defense contractor like Anduril versus getting involved geopolitically with China, right?

Not commenting on the specifics of the company, I probably would be happy to skip that risk. I don’t have a developed opinion on the whole OpenAI, closed AI, open AI, open-source AI debate. I don’t know enough about geopolitics in China to bloviate on it, and that product is available elsewhere from VCs, so I don’t need to fill that market gap.

But I just go, “Hmm, that would be a tough one for me.”

Look, I used to invest a lot in China and Russia back in the 2000s and early 2010s. I was an early investor in Alibaba, right, and in Ant Financial, et cetera.

The thing is, there was a possible path where both of these countries would actually be US allies, right? Deng Xiaoping really saw the 2 countries as aligned. And Deng Xiaoping was probably one of the greatest statesmen, frankly, ever, and got a billion people out of poverty.

The path he was on was pretty profoundly different from the one Xi Jinping is on. And so, with Xi Jinping having these nationalistic, conflictual great-power issues, once Jack Ma disappeared, I pulled out of China completely.

My remaining holding is Ant Financial, which should have been a massive home run, except that Xi Jinping just personally decided, “You know what? I don’t like what Jack said to the regulators. I’m shutting down the IPO.” And I did the same thing in Russia.

Russia was an amazing market for us, and then, in 2014, Putin decided to invade Crimea. When that happened, several of our unicorns were funded by Tiger, Bessemer, et cetera. All the capital pulled out.

The only people to fund them were local, well-connected oligarchs who were like, “Okay, I’m out.” So would I take that geopolitical risk today? No, absolutely not. Do I think actually backing people that are contrary to our interests is a good idea? Absolutely not, just from a moral perspective.

Now, do I think that at some point this may change and they may become more aligned, and we can get away from this great-power-war, great-game-type world that we’re back in currently? Yes, I hope that’s true.

I’m actually hoping that, in the long run, as China becomes wealthier, the masses will not want taxation without representation and it turns into a democracy. But the problem with dictatorships is, as people can see, dictators are in power for a very long time, even if they don’t do right by their people. Look at the Castros or what’s going on in Venezuela.

So for now, I would completely avoid these geographies. Frankly, I even moved away from Turkey, which remains a democracy and a US ally. But Erdoğan is, to me, going against all the principles of Atatürk and the positive revolution that happened there.

And so I stick to my personal moral principles. By the way, speaking of defense tech, I’m investing a lot, actually, in Ukrainian defense startups. My vision is that Ukraine could become the manufacturing hub for defense.

With Endural—and, look, I’m an investor in Endural—the problem is that its cost is extraordinarily high, so the way you measure this is cost per kill, and they’re not battle-tested. You want to minimize cost per kill, basically.

And so these startups in Ukraine that are using super-low…

Fabrice Grinda

Yes, I know that sounds horrible.

Jason Calacanis

I want to see a few investor decks with CPK—cost per kill. I want to see it declining over time, and I want to make sure that by Series C, your cost per kill declines.

Fabrice Grinda

Look, if you're backing defense tech, this is about ROI, right?

Jason Calacanis

No, I get it. It's a learning—

Fabrice Grinda

Like, people want to—

Rory O’Driscoll

It's a learning—

Fabrice Grinda

...make you be efficient. If you're a country with fewer resources, you need to be more efficient.

What Ukraine is doing is building an extraordinary stack that, if we are ever in a great-power war with China or Russia, we're going to need access to, because our costs are too high. Right now, we don't have the manufacturing capacity. We would lose a war with China right now, and I think the only way out of that is backing things where you can build scalable, cheap mass manufacturing. The reason we won World War II is that we out-manufactured the Axis, and that was the US. Today, we don't manufacture.

Jason Calacanis

CAC, CLTV, and CPK are the metrics I really run the fund based on.

Sometimes, a high CPK can be disguised in a low CLTV, and that's the problem with the lower-margin American Dynamism investment. Your CLTV can seem high, but your CPK is high as well. What is the magic number here for the ratio of CLTV to CPK?

Rory O’Driscoll

It's 5:1. 5:1.

Jason Calacanis

5:1? Is that how it works? Yeah, yeah.

Fabrice Grinda

Look, it sounds awful, but—

Jason Calacanis

Maybe David Sacks can help us with the ratio, like a burn ratio. We can do the ratio here.

Fabrice Grinda

Look, it sounds awful, but I think that this is existential. It's existential for Ukraine. I think it's existential for the West. I think we need to be in a position to be strong enough to defend ourselves; otherwise, we're going to be bullied—

Jason Calacanis

Right.

Fabrice Grinda

...and otherwise, we're going to lose. If we are perceived to be weak, there will be war. I think our perceived weakness is more likely to lead to a war and an invasion of Taiwan, or whatever, than if we are perceived to be strong.

It's distasteful, and I would much rather we all live in harmony, to be clear. I would rather we didn't have the leaders we have on both sides, but it's existential and essential, and we need to do this.

Jason Calacanis

Can I just add one question? I know we're over.

Rory O’Driscoll

Sure.

Jason Calacanis

But to Fabrice, since you've—I've never done any defense tech. Do all these new seed investors and folks who are excited about it have any idea what they're doing, or are they just chasing trends?

Do all these folks in Dogpatch investing in hot startups know what the hell they're talking about when they pop down to El Segundo for the day, or are they just flushing their money down the drain?

Fabrice Grinda

I think no. They're doing what Rory said people should be doing, which is investing in megatrends. Is defense tech a megatrend? Absolutely, and they're trying to latch on.

By the way, there are very good funds in this, like Shield AI in defense. They're amazing, and they know exactly what they're doing. Are most people who are latching onto this lemmings?

Look, I'm invested in Anduril, but I think most people are like, "This is the emerging winner. We need to back up the truck here and invest." How thoughtful are they, and have they really looked at the cost per kill, et cetera? I think the answer is no.

Jason Calacanis

Yeah.

Fabrice Grinda

But probably they still will do well because—

Jason Calacanis

I'm going to ask him that question at Demo Day.

Rory O’Driscoll

To be clear, they're probably just as thoughtful as the dude who piled in $250 million on the SPV in OpenAI 10 months ago.

Fabrice Grinda

Exactly.

Rory O’Driscoll

Don't conflate overthinking with the fact that momentum investors may actually be impacted negatively by overthinking. I think it's actually one of my faults: if you just want to pile on the dominant winners and the megatrends, please stop thinking, because that sentence is all you need.

Just one last comment on this. Listening to all this, I remember 15 years ago investors asking us why we didn't invest in China. We didn't then, and we have never done so. Yet, my perspective as an immigrant coming here is that the US is 25% of the world's GDP and 50% of the world's software and enterprise technology market. If I can't make money on 50% with the rule of law, adding another 10% with no rules whatsoever isn't going to help me.

I'm so glad I haven't had to do that to make a buck, and good luck to the Benchmark boys. It feels like a hard row to hoe.

Jason Calacanis

Yeah.

11. Europe Enters The AI Arms Race

Jason Lemkin

No, look, every founder—especially since I'm European, right? I'm an immigrant as well. All my French founders are like, "Where should I build startups today? Does it matter?" I'm like, "No, come to the US."

You have 300 million rich people who are early adopters and are dying to buy your product. You're playing the game of life on easy mode, or very easy. Yes, it's harder to get visas, but whatever. You can figure it out. There is no doubt.

By the way, if you're at $100 million in revenue, it's easier to go from $100 million to $200 million in the US than from $0 to $100 million anywhere else. That's true at $1 billion. It's true probably forever. Just buy the Europeans, actually.

Rory O’Driscoll

Harry, you and Fabrice have just kicked off—

Heavy pushback. Classic glib statement from an American—and an American in Turks and Caicos, I add lovingly, Fabrice.

Lower salaries in Europe, higher retention—fantastic. I'd like to have both, okay? Easy access to cash. There's a lot of money in Europe, actually, and there's—

Jason Calacanis

Huge market.

I completely agree, and you can sell actively into the US from Europe. Pigment has shown that and scaled to a really meaningful revenue size from Europe. You have to get on a plane more often, but it's absolutely possible. You can leverage cheaper engineering teams and hire great sales teams in the US. Very possible.

Jason Calacanis

What's the question? Is there an arbitrage here?

Point being, you don't have to. You can actually—

Rory O’Driscoll

Harry and Fabrice are arguing. We can sit this out, Jason.

Jason Lemkin

No, no, no. I always—When I started investing in European startups accidentally—Pipedrive, Algolia, and Talkdesk—it's not a great insight, but I'm always shocked by how much cheaper engineering talent is in Europe. S-tier talent.

In Paris, the best engineers are half the price, and people stay past their cliff.

Rory O’Driscoll

Yeah.

Jason Calacanis

Even at OpenAI, people leave at their cliff. Why that arbitrage doesn't work even better is the odd question. It's less understood that it exists, which is freaking true.

I'll tell you just one thing, Harry, on this. When I've asked a lot of the European founders that I still work with about this, they're all relentless. They're like, "We're—" They're just not driven. Our European folks are just not driven enough. I don't care about the cost. They're just not driven like San Francisco.

Nicolas Dessaigne from Algolia, when I saw him at YC Demo Day—everyone in every French startup has got to come now. Maybe he's talking about the YC thing as a general partner, but everyone I've worked with is like, "We don't care about cost or retention. It's just the pace. We cannot get our team in Nice or Barcelona to work at the pace of the US, and we don't care."

Do you want me to lose a load of friends?

Jason Calacanis

I know Project Europe disagrees. I'm not—I don't have the database you do.

No, I'm going to lose a load of friends in one sentence. None of those are generationally defining founders. Go to Revolut's HQ. You will not find—

Jason Calacanis

Revolut's awesome. No, no, I'm not saying that. I'm just saying—

No, no, you're right, and I agree, but the founder—

Jason Calacanis

Yeah.

...sets the cadence, and if you are a generational—

Jason Lemkin

I think the ultra-outliers will be everywhere outside of Antarctica, right? You will have the Revoluts, right? But when you have to hire hundreds of people, that's where it gets hard.

I just think then you're stuck with the average. No matter what anybody says, there aren't 500 S-tier engineers who want to work anywhere but Cursor. They just don't exist, right? They're going to work at your concrete B2B company with an AI copilot? You're lucky to have 4 S-tier engineers at that company.

If you get 4, you're lucky. If you have 1 person who would work at OpenAI join your company, it's a serious issue. You're lucky to get 1. But maybe if you're in San Francisco, you could get 2.

Rory O’Driscoll

I don't know if I agree with you, Jason, on that gross generalization about Europe, but I actually don't think you have to. The real truth is that in enterprise software, as distinct from consumer, where Fabrice is playing, it doesn't matter where the R&D is. It should be wherever you can get good R&D.

The spend is 50% in the US, 20% maximum in Europe, and 30% in the rest of the world, and the US is the early adopter. So the thing that is true for enterprise technology companies is that your R&D can be anywhere. Your go-to-market will have to be in the US. You can make that statement wonderfully without typecasting the competence—or lack thereof—of the entire 300 million people in Europe.

As a European, I don't want to do that because I do want to be able to go home sometime. So I think we have loads of investments where they have European engineering staff, et cetera, and they're awesome and great. But I think, in the end, the reason we mentally think about the American market is because, you know, Willie Sutton: that's where the money is.

Jason Lemkin

Yeah, but I just don't think you can win today if you're comfortable leaving work at 5:00.

Rory O'Driscoll

I'm going to defend the Europeans. Now it's working. You're pissing me off.

Jason Lemkin

What?

Rory O'Driscoll

I think there's a visible difference between the Europe of the people who get it and the Europe of the people who don't. Let's be clear: out of the 300 million people, not all of them, or even most of them, get it, right? But when you run into the people who do, you're like, "Yeah, you're with the program." And you're right, because it's not anywhere close to most, the generalization that you're making is correct, Jason.

But that's looking at the whole working population. There are teams—there are critical-mass areas—where people do get it and they're cranking, à la American. Look, Dublin is a great place for American tech companies, and fundamentally it's because people are with the program and cranking.

Jason Lemkin

I think pre-AI, I agreed with you. I've just changed my mind in the AI world. The value of a 10X engineer is 100X now. It's 100X.

Rory O'Driscoll

Jason, that could be true for infrastructure, but okay.

Jason Lemkin

Your companies are going to go under if they're not dominating AI in their market in B2B. They're all going to die if you're not number one in AI in B2B.

They're going to die.

Rory O'Driscoll

Harry cannot hide. Let me tell you why Harry's hiding here. Let me give you the bad news, Fabrice. What happens in this program every time is we crank up stuff, and then by the end Jason escalates, and that's okay.

But the problem is this: the social marketing team from Harry's team then picks up the most inflammatory quote that Jason made and puts it out on the internet with your face and my face. So tomorrow morning we're going to see a thing that says, "No Europeans worth a damn." It's going to be your picture and my picture, and we'll never go home again.

Jason Lemkin

No, no.

I think there are plenty of Revoluts. I just don't think going home at 4:30 to have red wine and baguettes means you're going to win today. You're just going to lose.

I agree.

Jason Lemkin

I can tell you this from all the French founders I know.

I can agree with that.

Jason Lemkin

I literally talked this week to someone who was interviewing for an SVP of engineering role, someone incredibly respected in Europe, okay? He was doing the cycle with all the portfolio companies in B2B. He's like, "Well, I just got back from a month off in Europe for my vacation, and I want to spend some time really thinking about AI." He said, "I really want to spend some time thinking about AI," right?

And so I took the chat and I put it into SaaStr AI, and I said, "Should we hire this guy?" And the AI said, "Whatever you do, pass on this guy." This was one of those candidates—you know, the ones who get passed around to all the VCs because they're hot candidates? This was a hot candidate, but with the month off, the baguettes, and the fact that he was going to dilly-dally into AI, he was going to get destroyed by the kids in SF. You're going to get destroyed.

Fabrice Grinda

But Jason, right now, if you're a 10X engineer, with AI you become a 100X engineer.

Jason Lemkin

It's true.

Fabrice Grinda

The question I have is, in the long run, could AI actually have the opposite impact, where regular engineers become—yeah, maybe they become 10X themselves? Maybe they're not 100X, but they become 70X, and that's good enough. I don't know. Maybe.

Jason Lemkin

A regular engineer today has already gone from 1X to 2X. There's no debate.

Fabrice Grinda

Correct.

Jason Lemkin

There's no debate.

But the 10X engineers become 100X engineers.

Fabrice Grinda

Did it shrink the gap?

Jason Lemkin

No, no.

Fabrice Grinda

But that's the question.

Your whole team has already been 2X'd in the last 5 months. Your whole team is at least 50% more productive, right? Even Salesforce commits 20% of its code through AI. So if you're good, you're 50%. And literally, it's 50% across Windsurf. Like, 50% productivity. No matter what anybody says, there are no 9Xs, okay? It's 50%. But everyone has these tools. As Harry said, it's 15 bucks a month, 100 bucks a month. Everyone's twice as productive, right?

Yeah. But the question is—

Jason Lemkin

So it's an arms race.

Fabrice Grinda

In the long run. Yeah. But in the long, long run, when AI becomes so much better—

Jason Lemkin

Yes.

Fabrice Grinda

Do you think that the gap will shrink between the very best engineers and the average engineer? Because if that happens, that changes the game. If it doesn't, then it doesn't. It's a question, by the way.

Jason Lemkin

I'm still learning.

But I'm going to make the bet that—

Fabrice Grinda

For all of history, 10X engineers have become better and better. They've become better and better, and that's what we're seeing in startups today. That's why they can grow so quickly. That's why Harry's described this as their small teams. It's not just that they're small teams; it's that they're much better than they used to be.

So I think we're going to see bigger teams, and we're going to see 10X engineers even better. I think half of these sales teams will be gone in 2 years. Half of these sales and customer success teams will be gone in 2 years because they're mediocre. They'll be gone. They're all going to be laid off.

We're not going to need any mediocre two-call SMB reps. We're not going to need any customer support folks that don't even show up to the QBR. They're all going to be gone with AI. And the engineers are going to be even better.

But the arms race between Cursor and Windsurf is so huge, and that arms race is going to lead to bigger and bigger engineering teams that are better and better and better. I know Windsurf's 100% in the office. I think Cursor is, too. They're going to be these aggressive, in-the-office, 6-and-a-half-day-a-week teams that don't get baguettes and red wine at 4:30.

I still love Europe, don't get me wrong, and I'm a fan of Project Europe. But I think we've got to find the Revoluts, because baguette culture, I think, is worrisome. But it's crazy. The rate of change is crazy, and I do think all these B2B companies are just going to die if they can't change fast. They're going to die.

I cannot thank you enough for doing this. This has been so much fun.

Jason Lemkin

What about Deel and Rippling?

Deel and Rippling.

Jason Lemkin

Deel and Rippling—what about Deel and Rippling this week?

Do you know what? We're going to pause before we lose more friends. You've pissed off a continent already, Jason, so you're cool. But guys, thank you so much. This has been amazing.

Fabrice Grinda

Thank you.

Jason Lemkin

Talk to you soon. Nice talking, Fabrice. Thanks for the time.

Fabrice Grinda

Likewise. Bye.