20VC：今天做种子轮需要10亿美元：风险投资不为人知的真相 | 未来十年VC谁赢谁输 | AI时代，为什么利润率和价格没那么重要——Sameer Dholakia（Bessemer）

Intelligence is going to be a bigger market than oil. I’ve been in software for 31 years now, and the handful of other moments when I have felt magic, a trillion-dollar supercycle followed every single time. It doesn’t happen very often. I absolutely believe capital is a moat. If you’re not a giant global platform or an incredible specialist who can add differentiated value, I think it’s a hard world to make money.

$5.75 billion. That is how much Bessemer Venture Partners just raised across their two new funds. A $4 billion fund and a $1.75 billion seed fund. My word, they really took it literally when I said you need a billion dollars to do seed. Well, I saw this and I thought we’ve gotta have Samir on the show. Samir is one of the leaders at Bessemer today. He led investments in Lagora and Fireworks, two 20VC companies as well, which is awesome to work with them on, and he’s one of the best growth investors in the market today. It’s a spicy show. It’s kind of what we’d call in the UK a bread and butter show in the way that this is right down my fairway. So unlike the AI deep tech infra stuff where I’m learning as I go, this is really in my zone of comfort. And we had a great discussion, and I love doing shows like this. Like, these shows are why I still do 20VC with such passion. Samir was patient. He put up with me, and hell did we have a discussion. But before we dive into the show today, Blitzy, the autonomous software development platform that’s built for enterprise scale, is proud to sponsor today’s show. Enterprise code bases have grown beyond human comprehension. Tens of millions of lines, decades of tech debt, complexity existing tools cannot fathom. Blitzy starts by understanding. Thousands of agents reverse engineer your code base into a proprietary knowledge graph that maps architecture dependencies and business logic. That graph surfaces every CVE and zero-day hiding in your code before anyone hands you a ticket. Grounded in that context, Blitzy then executes entire epics autonomously. Modernize off legacy systems, ship new features 5X faster, or batch remediate CVEs at scale surfaced by the platform. Blitzy not only generates code, it validates and end-to-end tests every single line. And that’s why Fortune 500 enterprises are shipping enterprise-grade code five times faster. Go to blitzy.com/20vc to learn more. While Blitzy helps you build faster, Flex helps you run your finances smarter. Business owners run their whole financial life on Flex. One platform from business revenue to their personal spend. Float every purchase for 60 days. Tap capital that grows with your revenue and pay vendors in 170 countries across 32 currencies. Plus, the whole back office, bills, expenses, accounting all in one place, so you spend less time reconciling and more time growing. That’s why thousands of owners use Flex. Named one of Fast Company’s Most Innovative Companies of 2026. Visit flex.one. That’s F-L-E-X dot O-N-E, and use the code 20VC. While Flex runs your finances, Vanta keeps your compliance covered. When you’re building a company, you learn that trust is what closes deals. You may have the best product, but no buyers will sign these days without proof of your security. Here’s what happens if you’re not prepared. A prospect asks for proof of compliance. The deal stalls when you scramble. Your engineer gets pulled off the roadmap to audit prep. Every enterprise conversation turns into this horrible fire drill. That’s where Vanta comes in. Vanta is the leading agentic trust platform that not only gets you compliant fast with frameworks like SOC 2, ISO 27001, HIPAA, and GDPR, but keeps you compliant by continuously monitoring your controls so your deals keep moving and your engineers really keep building. Now access the Vanta Agent everywhere you work, even if your team lives in Claude or Cursor. And that’s why Vanta’s trusted by more than 16,000 companies like Ramp, Harvey, and Writer. So prove you’re ready for business and get $1,000 off Vanta when you go to vanta.com/20vc. That’s vanta.com/20vc.

You have now arrived at your destination.

Sameer, last time we did a show, I was about 12 years old. You put up with some very poor questions then, so I’m hoping this will be an improvement. Thanks for joining me, dude.

Dude, I’m so fired up to see you. As I always say, long-time listener, now second-time caller. I’m fired up to be here, man. Thanks for having me.

We were just chatting beforehand and said, “Ah, USV and, like, $900 million,” and you guys just raised what seems like the size of the US Treasury. I think it was a $1.75 billion seed fund. We’ll get into that, dude.

Seed and early. I’ll talk about that later. Go ahead.

Seed and early, and I think, like, five more in growth. We’ll get into that, but you said, “Oh, well, for those that have seen it before, this is that time.”

When you’re in one of these, I refer to it as magic. I’ve been in software for 31 years now, and the handful of other moments when I have felt magic, a trillion-dollar supercycle followed every single time. It doesn’t happen very often. Talk about lucky: I graduated college in 1995, back in the Dark Ages, at the birth of the internet, right in the heart of Silicon Valley, and it was magic.

You could see how people could create content and share it with everybody on the planet. You’d think, “This is magical,” and, “Oh my gosh, we could actually do commerce on this thing, and I don’t have to go somewhere to get something.” That was magical, right?

Later in my career, VMware created this layer of software that disintermediated the hardware from the operating system. It said, “I can spin up virtual machines. I can put 10 of them on that one box. And, by the way, you want to see magic? I can move that running workload from machine number two over there to machine number eight and shut the other one down whenever I want, and it won’t interrupt the workload.” Dude, that is magic. I looked at that and thought, “Holy cow.”

I didn’t appreciate that the supercycle it would create was called cloud. There’s no hyperscaler, no public cloud, without the ability to do virtualization and move VMs around, spin them up and spin them down—the elasticity of the cloud. Every time I felt magic, I thought, “Holy cow, this is cool. I’ve got to go jump into that.”

For this one, I remember I was on my sabbatical, in between Twilio. I thought I might be retiring, got bored, and was doing some advisory work. This young guy, Dave Rogenmoser over at Jasper, said, “Hey, can you spend some time with me on what we’re building here? It’s growing really fast.” I said, “Yeah, I’d love to learn more about it.”

He showed me a demo of Jasper, an AI marketing app sitting on top of GPT-3, in the winter of 2021. I thought, “Magic. Again, this is magic.” Sure enough, here we are five years later, and it’s been magical. You can see these transformational technologies, and this one is bigger than the last two combined, in my humble opinion.

Why do you say that?

The analogy I often try to draw is that intelligence is going to be a bigger market than oil, because oil powers all the machines, and intelligence powers all the humans. Not some of the humans—all the humans. Consumers, prosumers, SMB, mid-market, enterprise, government. I don’t care who you are or what you’re doing; you need more intelligence. When you think about a market of that scale and scope, you understand this is magic, and it’s going to be really big.

If we go through the stack, then: You did Anthropic. Let’s start at the heartbeat of intelligence that you mentioned. Can you talk to me about the rationale, the underwriting, what you saw, and what you didn’t see when you look back now?

Yeah. The most important bit is that I’m really proud we did it in the Series D.

What price was that, dude?

I think it was $15 billion pre, $18 billion post. It was early. It was the round that Menlo led, and Lightspeed and Bessemer joined. By the way, we were talking about my legendary partner, Byron Deeter. Thank God for him, because he and Matt Murphy go back a long way, and Matt called up his buddy Byron and said, “Hey, we’re looking to build a syndicate here for this startup named Anthropic. You should check it out.” Byron said, “Hey, Sameer, you want to come look at this thing with me?” I said, “Yeah, I’m all in on, as I said, all the intelligence in the AI market.”

Byron and I grew up during those cloud days; our careers probably centered around them. Of course, we all pattern-match. We use these mental frameworks. For us—or I’ll say for myself; I won’t speak for Byron—when I looked at it, I thought, “Look, I think this market is going to be a cloud redux.” In the cloud era, you had a handful of hyperscalers: AWS, Azure, and GCP. The reason is that it’s massively capital-intensive to build a cloud infrastructure.

It turns out that in the market we were looking at, it was going to be the same thing. It was massively compute-intensive and therefore capital-intensive, so it was going to consolidate to a handful of players. But just like in the cloud era, we believed then, and I still believe now, there would be hundreds of apps built on top of them. Just like you had hundreds of SaaS apps, you would have hundreds of AI apps purpose-built for particular things.

If you have that mental model, then you say, “Okay, great. If it’s really capital-intensive, who has the balance sheet to actually invest in the scaling laws people were talking about?” It’s a pretty short list of players who could afford it.

You go through them and think, “Okay, Microsoft, yep, they partnered with OpenAI. Google, yep, they’re going to build their own, called Gemini.” Then you get to AWS. At the time, AWS had an internal model—I think it was called Nova—that was not particularly distinctive, I’ll be generous. And they had just invested a very sizable amount of money in Anthropic. By the way, at the time, it was very non-obvious: Anthropic was one of half a dozen or more labs trying to build intelligence.

Was it controversial at Bessemer?

Oh, Harry, you have no idea. According to Byron, it was the lowest-voted deal in the firm’s history during the time he’d been there. But the fact that we were both there and had conviction—and one of the things I love about Bessemer is that empowerment is one of its values—they were like, “Dude, that’s where alpha comes from. If you guys really believe, go do it.”

What did you not see that you wish you had seen?

I wish I’d been able to convince us all to write a bigger check that round. I feel like we saw it, we knew it, but there were so many risks at that stage of the company. We knew the AWS relationship was pivotal: it would provide the compute for training, which was critically important, and distribution at that early stage. I just wish we’d written a bigger check. That’s all. I wish we knew then what we know now.

When we look at concerns today, one is that the cost of intelligence shifts to open source because frontier models are simply too expensive. We see that in a lot of data produced by Vercel and Guillermo, who publishes it. We’re both big investors in Fireworks, so we both believe in open models, too. To what extent do open models not undermine the business model for the majority of frontier providers?

Look, I’m a genius of “and” on this one, versus the tyranny of “or.” This is “and,” not “or.” I come back to my comments about the scale of the market for intelligence. We just can’t get our heads around it. When we invested, my bull case was that we could get to some massive valuation, and I’ll just say that I missed the mark on how big a business could be created at the frontier by factors.

When you’re in a market that big, predicting with any precision the mix between closed tokens and open tokens is irrelevant. It’s going to be big. You’re in, right? Do you understand what I’m saying? When it’s that large of a market, the ratio is not a concern. I really do believe there are all those different use cases, from consumers up through enterprise and government, and they’re all going to want intelligence of different types and flavors.

That’s why we’re investors in both Anthropic and Fireworks. We’ll have exposure as a fund and as a firm to both closed tokens and open tokens, but I think both are going to be huge.

Do you think U.S. enterprises are willing to use open-source Chinese models?

I think that once that Chinese model is hosted by Fireworks, it’s no longer a Chinese model.

Do you think they see it like that? You and I both know that to be true, but when you speak to CIOs, they go, “I don’t want it. If the provenance is China, I don’t want it.” And you’re like, “It’s downloaded code on your local drives.”

Yeah. I’ve talked to some CIOs who share that opinion, for sure. There’s no question that’s out there. By the way, I do think we should have a great open-weight model that’s U.S.-based, and I think that would be awesome. I think there’s a big hole in that market.

Is that a great business?

It can be. For sure, it can be.

But if you look at Thinking Machines, which is priced at $40 billion today, is there a $500 billion U.S. open-model company to be built? Or is open just a much harder business model to monetize, and therefore it won’t be that large?

Oh, no, I believe it absolutely can build a big business. Whether it’s at that particular scale, I don’t know, but you can build a very large business and get venture returns if you’re the leading provider of an open-source solution. You just look at history across all these other categories, whether you’re going from Red Hat to, fill in the blank, in the database market.

You can find ways to monetize it, for sure. Most enterprises don’t want to be responsible for the open source. They’re going to want a provider with the right talent behind it. Those are all the reasons open-source businesses have been great businesses for decades.

If you were to present a bear case for frontier models, what would it be?

Do I think there’s a bear case for the frontier labs? I really don’t, because I believe that, at the frontier, intelligence can solve problems that were heretofore unsolvable, or solve them at a pace that was heretofore unimaginable. I just think there’s no shortage. Humans have so many problems to solve, Harry. We have a lot of problems in this world. We have a lot of them, and we’re nowhere near the 99th percentile of problems solved, where we’re just trying to get to the last one. We have a lot of problems to go solve, and that’s what these things will help us solve.

The majority of people on the pod come on and say we’ll see 99% of token traffic flow through open models, but we’ll see 99% of dollars flow through closed models. Do you think that’s apt, or do you think that’s kind of a bullshit throwaway comment?

I think that ratio is wrong. I really do. I don’t think that’s quite the mix, but to my previous comment, I don’t think we have to get the mix precise. I think everybody just has to understand that you have to look at the size of the TAM.

Fifteen billion seems now like ridiculously cheap. How could you? That’s amazing. Fifteen billion at the time was fricking wild. I got brought OpenAI at $30 billion, and to my stupidity I said, “Oh my God, if this is $100 billion, I get, like, a 3x, not including dilution. Oh my God.” The small-mindedness I had was just unbelievable.

Harry, get out of my mind, man. I went through the exact same thing.

So how did you get over it?

It was because of what happened in the intervening time. You remember that OpenAI opportunity that I also passed on? I had the exact same thought pattern you did, and that was right after the ChatGPT moment. The conversation was in December, and by the time we looked at that first round of Anthropic, OpenAI had gone from about $40 million to—I’d have to go back and look at the numbers. You could fact-check me, but it was well north of a billion. A billion and a half, maybe approaching $2 billion.

Clearly, we got it wrong. That is one of the most humbling things about this job: you will get them wrong, and sometimes you’ll get them wrong on really important decisions. And you have the opportunity to fix it. It’s not a once-in-a-lifetime opportunity. You can go back and fix it. That was when Anthropic came along.

Our imaginations were capped. To your point, even if it got to $100 billion, we couldn’t see beyond that. Now, obviously, you could see the growth potential of how big this market could be because of what OpenAI had done in calendar year 2023.

In a world of $2 trillion assets, do you care about price anymore? Is LP focus on price misplaced?

1. Paying for AI Growth

Entry price always matters. Every venture capitalist has taught me that. By the way, you know me: I’m four and a half years into this second career of mine, so I’m still learning. I appreciate that it’s a craft like every other, and I don’t pretend to have all the answers by any stretch. But I know how important entry price is.

What matters most is whether you believe it can grow at scale and compound for a very long time and still make you a venture return. I’m focused on looking out the windshield, as they always say, as opposed to the rearview mirror. Some of the investments I’m most proud of were at current ARR multiples that would make most venture capitalists throw up. They were difficult conversations in our firm’s IC—in the investment committee.

When we invested in Abridge, they were at $5 million of ARR, and we were paying a $700 million or $800 million valuation, whatever it was. When we invested in Anthropic in that $15 billion round, I think they were at $100 million—like 150 times current ARR. Lagora had $20 million of ARR when we invested and led that $1.8 billion round. That’s insane.

If you’re only looking at the present and the past, you’ll say, “Sameer, do you not understand math? A hundred times? That’s what you’re paying to get in?” But if your view, as mine was in each of those instances, is, “Yeah, but on an NTM basis”—that’s what I care about—“what do I think this thing can be in the next 12 months, and do I feel like this is a reasonable multiple in the next 12 months?”

So you look at the end of next year?

Again, this is me trying to learn. You look at the end of next year, and then you look at gross profit on that.

Not even the end of next year. I look at NTM, the next 12 months. Depending on the business and where you are in the year, that could be more like 18 or 24 months, and I won’t pay that far forward because it’s really hard to see that far ahead in any business. I’ve been an operator and a CEO, and I know how hard it is to predict that far into the future when you’re scaling at these rates. But you should be able to get your arms around the next 12 months, or at least get closer.

What is exciting enough in that respect, then? When you look at NTM, what’s green versus red?

Depending on the business, it’ll vary. Something in the 10 to 15 times NTM range isn’t crazy, especially if these businesses are growing at rates that have never been seen in the history of business, and that I’ve certainly never seen in my 31-year career. I don’t think paying 10 to 15 times NTM is all that crazy, because they’re going to blow through it.

Look at Lagora. We invested at a $1.8 billion valuation, leading that round, and that seemed insane at the time. But the company went from $20 million to $200 million. I think we’re in that together, and I know you know Max.

Yeah.

They’re on fire, and I feel pretty good about that $1.8 billion.

We’re going to get to Lagora, and I completely get that. Does feeling good about that $1.8 billion change depending on customer makeup, quality of revenue, or run rate versus enterprise revenue? We see some quirky ways to calculate revenue today. Does that change things?

For sure. Quality of revenue matters deeply. I’m old-school. I’ve been in enterprise B2B software since 1995, so that’s my bias. I prefer the gross dollar retention and net dollar retention rates of highly sticky customers, for whom your product is super strategic and who are going to keep expanding.

I would pay a much lower multiple for a prosumer or consumer business. Churn can kill you.

What about margins? You look at someone like Lagora, who we both love. Their margins are good, but they’re not enterprise-level—

Look, I think—

High margins.

Margins depend on your lens. Am I worried about their gross margins in the next 12 months? I’m not. I have a five-to-seven-year hold period for a growth-stage check, so I think about what I expect the business to grow into over the next five years, and why.

You have to be really thoughtful if the gross profit picture doesn’t look pretty and you’re envisioning some spreadsheet magic where it’s suddenly going to hit 60% or 70% in five years. You’d better be able to defend that and know why you think it’s possible. But in a lot of businesses, economies of scale matter, and market share and position matter, too.

I can remember that, for most of the businesses I invested in, gross margins at the time we invested were meaningfully lower than they are at scale today in this AI world. Meaningfully lower—improvements by factors of multiple inside of 12 to 24 months.

When we look at a company like Lagora, I’m an investor, and I’ll always admit I’m a big fan. You invested at $1.8 billion. The legal market is $300 billion in the U.S. If we assume a 20% take for software, that’s $60 billion, and we assume it’s split between a couple of different players, including Harvey, you could get to maybe $5 billion to $10 billion in revenue. Is that how you thought about it? And how do you get to a 10 or 20x return when you enter at $2 billion on a company like Legora? Help me.

That math isn’t terribly far off. I’d just say we included the whole world, because you and I both know Max has global ambitions. When you add in the rest of the world, you don’t grow the TAM a little bit; you add another 60%. The U.S. market is only 40% of it.

I do believe you can go after the services dollars that go into the legal market. It’s the cliché everybody keeps talking about: Jevons paradox. I believe there are a lot of legal needs that aren’t served today because prices are too high.

The way we thought about it, very simplistically, is that there’s about $1 trillion of legal services spend in the world. At that size of a TAM, I believe they don’t have to take any meaningful share to build a massive business. And if they build a $5 billion to $10 billion revenue business, you can put any reasonable multiple on that, and you’ll make a very good return.

What’s interesting when you look at all markets today is that the competitiveness has changed. When you were building SendGrid, no offense, there weren’t that many competitors. There were several, but not the depth we have today.

When we look at Lagora and Harvey, I’m going to be unpopular: I think both will do very well. When we look at Fireworks, I think Baseten and Modal will do well, too. I think Fireworks will win, which is why my money’s there. Has the competitiveness changed in your mind? And does that shape how you think about investing and market outcomes?

2. Capital Becomes a Moat

I actually don’t believe the competitive breadth has changed. What’s changed is the capital intensity behind the competition, which leads to this competitive intensity. If you look at the war chests—the amounts of money that companies like Lagora and Harvey have raised—that creates a competitive intensity that’s absolutely different from what I dealt with in my era.

Is capital a moat today? Is that the differentiator?

I absolutely believe capital is a moat. I’ve always said that markets pick winners, and then it becomes a self-fulfilling prophecy. In our day, SendGrid and Twilio were the API building blocks developers plugged into their applications to send communications. Once developers had decided, it was game over. We had won. You were the default; you became the market winner.

That happens, and we see it in these markets, too. There are lots of legal tech companies and AI legal providers, but if you asked anybody at a big law firm who the major legal AI companies are, they’d name those two. I believe capital creates both the ability to scale at an extraordinary pace and a brand for those businesses. Talent then flocks to them, and you get this reinforced flywheel. They win more customers, and the virtuous cycle continues.

Do you believe in kingmaking? I do. To your point that capital is a moat, excessive amounts in a short amount of time give you inherent advantages. Do you agree with kingmaking, like I do, or not?

Yes, because I believe everything you said: capital is a moat, and it’s a signal to the market. Having built software companies for a very long time, I know your product has to be exceptional, and your company leadership has to be exceptional. In math-speak, I’d say it’s necessary but not sufficient. Having the biggest war chest and the most capital is a necessary condition for winning the market. I really do believe you’re in that kind of a game.

It’s because compute is so scarce, and therefore so expensive, that if you don’t have it, you just can’t play. So I agree with the first part of your statement, but it’s not sufficient. You then have to deliver an amazing product that your customers rave about, love, and tell their friends about if you want to be king. Otherwise, you have no chance.

I agree with you completely there. I also agree with what you said earlier: the 10 to 12x and the growth rates now are very different from what they were. When you look at growth rates today, is triple, triple, double, double—the model SaaStr, Jason, and I always talked about—wrong? Is that not enough anymore?

It’s not enough to attract capital from venture capitalists who have opportunities to put their money into these hyper-growth companies that are growing much, much, much faster. All of us as venture capital funds have scarce capital. As big as these funds are—we just raised $5.75 billion—it’s still scarce.

That said, I do still believe in finding companies doing T2D3, because the entry prices for those companies in this environment can be dramatically different, at a very different multiple, and you can still make a great venture return on them.

Do you buy that? Do you remember the kind of Excel deals—the Excel growth deals—like 1Password and Qualtrics, where they found these amazing assets in less obvious locations that were maybe more capital-efficient? They then juice them up with some growth capital, and it turns into an even better business.

Do you believe those will exist in the next generation, or do you actually have to have that capital to get to scale?

I think that over time, we’ll see a bunch of application companies that can look like what you just described. I just don’t think that’s where the venture dollars are going to flow in the near term. When you have opportunities to invest in things that are 10X-ing at scale, and your opportunity is something that’s doubling, of course you’re going to go to the 10X-ing. That’s just an intelligence test.

We call it “cake or death.” Have you ever seen Eddie Izzard? He goes, “Cake or death?” And you go, “Cake, please.”

Cake.

Like, wow. What a surprise. I wasn’t expecting that.

What a tough question. All that said, do I think we can make money investing in companies that are T2D3? Absolutely. You can. So then it becomes a bit of a mixed question: how much do you want to have in one versus the other?

You mentioned the $5.75 billion. It’s quite a lot, to be honest. Let’s start with the $1.75 billion for seed and early stage.

Yeah.

Do you need a billion dollars to do seed today? We just saw USV announce its $900 million funds earlier today. Do you need a billion to do seed?

Seeing the prices and valuations we’re seeing for the early-stage deals coming in from my teammates on the early-stage side, and thinking about the number of shots on goal you need to find the winner, I don’t know how you’d have a small seed fund and be successful. I don’t.

What are you seeing in terms of pricing and ownership levels?

It depends on the business. If it’s a capital- and compute-intensive business—and a lot of these AI businesses are—then it also depends a little on where you want to play. Where in the stack do you want to play? How capital-intensive a business do you want to invest in? Maybe you don’t want to invest in something that requires hardware, like robotics, or in something that’s training a model and requires compute. There may be places where you don’t need that capital intensity.

But in this era of intelligence and AI, I think a lot of them do. If you assume that’s the case, the math just won’t math for the entrepreneur or for you if you’re not paying enough. In my day—I’ve been doing this for 31 years—a seed check was $3 million to $5 million on a $20 million to $40 million valuation. It was a dramatically different time.

And what is it today?

I don’t have the specific numbers, but I know it’s factors higher. It’s way more. When you’re factors higher than that, you really have to hope that you’re a great picker.

This gets into one of my other worldviews, if you will. I’m just a humble business guy, Harry, and I think about the venture industry like any other industry. When I was evaluating coming in and starting it as a second career, I looked at the venture capital asset class and said, “This looks nothing like what I grew up with in the mid-to-late ’90s,” when everything was hardware and software in Silicon Valley, principally at the early stage.

Now it’s global, in every sector you could imagine, and covers everything from a pre-seed check all the way up to IPO, where you’re investing hundreds of millions of dollars. Every one of those vectors I mentioned—location, stage, and sector—is dramatically different and broader. It’s just the professionalization of this asset class.

If you think of it in that context, in most industries, when they professionalize and scale, you either have to be at one end of the spectrum—a giant global platform that consolidates a whole bunch—or an extreme specialist. You might say, “I’m doing AI deals in San Francisco, at the early stage.” You have to optimize for specialty on all three dimensions. You can be one of those two things, and if you’re in between, good luck. I think that’s a rough place to be.

Bessemer is a big global platform. But not being a giant global platform or an incredible specialist who can add differentiated value and be better at picking because you’re a specialist makes it hard to make money.

So what happens to the billion-dollar funds?

I don’t know, but I’m glad I’m not at one of them. I think you have to be on either end. I really do. I’m sure there’s a path to success for everybody. Good luck. I’m just giving you my outside-in, unvarnished view as a business guy looking at an industry. I’m analyzing it clinically, and what I’ve seen in the history of business is that when markets get big, they consolidate and have specialists.

Take the automotive industry. It doesn’t matter what industry you pick. You’ll have your Big Three in the US, or Toyota and Honda in Japan. You get a handful of giant players on one end, and then you can be a Porsche or a Ferrari, a specialist on the other side of the spectrum. Being in the middle is a tough place to be, I think.

It is. I think it’s a very tough place to be. Can I ask you, when you look at the two different vehicles today? I had Venky from Menlo on the show, a mutual friend of ours.

Yeah. Love Venky.

Yeah. He got some stick for saying that early-stage and seed investments were option bets, and that you build a portfolio—

Oh, I saw that. Yeah.

It’s hard not to have seen that. I think we’re at about 5 million views. But he’s right by definition. If you have the majority of your dollars at late stage, then by the nature of doing early-stage investing, unless you said, “I’m deliberately not going to invest in anything I invest in early,” which would be utterly moronic and irrational, it has to be an option.

Having been on both sides of the operator and venture thing, I think what people took umbrage with was the notion—maybe this is what they were thinking, I don’t know—that an entrepreneur with a small company doesn’t want to just be an option. I know Venky personally. He’s a great dude, and they’re a great firm. That is not what he meant.

What he meant intellectually is 100% correct: it is an option, and if it works, then I’m going to put more capital behind it. Anybody in venture capital who listened to that or thought it through understands that to be a truism. There may have been a bit of flak on behalf of the human entrepreneurs doing the hard work of trying to build these things, who don’t want to just be an option.

But I would challenge those entrepreneurs, having been one myself, and say, “No, man. Understand that you have to earn the right to keep scaling, to earn the right to the next round, to get more capital, and to keep building toward your dream.” That is the game. If you don’t think that’s true, you’re not understanding the game.

Can you help me understand? I totally agree. When you have the $1.75 billion, does that have reserves baked in, or is the idea that you have no reserves there and then you double down from the $4 billion, which is growth?

Both. We absolutely have reserves baked into the $1.75 billion, but we don’t want to just be doing pro rata checks. In the companies that we believe are taking off, we want to back up the truck, as we’ve done. We just announced a round with Legora AI where we’re going to be investing nearly $200 million. When we see the winners that we think are in big markets and are going to build great companies, we’re like, “Yep, let’s go.”

What does that interplay look like, then? Do you worry about the signaling risk of Sameer jumping all over one company and not another?

No, I don’t. At the end of the day, I believe the truth will set you free. If the signaling risk is that we haven’t backed up the truck on my company yet, the signal is simply that the business hasn’t yet demonstrated that it’s going like this. That’s okay. I led one of those. It takes a while. It’s hard to build businesses.

I don’t worry about the optics of stuff like that because I’ve been in the seat. I think entrepreneurs and CEOs know the business they have, they know what they’re working on, and they know what they have to do to take it to the next level. If they’re not yet at the next level, I don’t think most CEO-entrepreneurs are going to want to pretend they are.

So you have $4 billion in growth going to your segment of the market. How does that look? When you look at the portfolio, how concentrated is it? What does the check size look like? What do the reserves look like? I’m fascinated by how you construct a $4 billion portfolio.

It’s incredible, because the numbers are obviously staggering. They’re large. But I don’t think the fundamentals are all that crazily different, in that we still want to have concentrated bets.

A concentrated bet means we may write $200 million to $400 million into a given opportunity, into a given company, if we have that much conviction. It’ll probably be over a couple of different rounds, over a few rounds. But you’re still looking for relative concentration, whatever it is.

What percentage of the fund would you be happy to put into a single company?

For our late-growth funds, 5% to 10% is what I think our average check size will be. Early growth is a little bit different, but for late growth, 5% to 10% feels fine. I’ll just tell you: in our last growth fund, Century II, I think we had 20% in one name. The name was Anthropic. I feel pretty good about it, so we’ll flex it up if we believe it’s a generational company.

You have an early-growth and a late-growth fund?

We do, yeah.

Fucking hell. What the fuck? Fuck off, Sameer.

Early growth is very different from late growth.

Come on.

It’s so different.

No, it’s not.

Okay. Tell me. Go ahead.

They’re both growth.

No.

Why don’t we do a mid-growth?

Companies—

Why don’t we do an early growth, a mid-growth, a not-so-late growth, a late growth, a pre-IPO growth?

You’ve got to draw a line. I’ll concede, Harry, that there’s an arbitrary cutoff somewhere. But qualitatively, there’s a big difference between investing in a company in the $20 million to $40 million revenue range, or one valued at, say, half a billion to a billion dollars, versus coming into Fireworks at $15 billion. You’re talking about a very different stage of company. What you know about the business is different, the risk profile is different, and the check size is going to be different. All those things are different.

I get all of that, but I think what matters is your multiple on invested cash. And actually, if you look at a company like Fireworks, I think the upside is just freaking enormous if this works.

Enormous. Yeah, sure.

And I adore Lagora, but by nature, I think the upside is just less because it’s a vertical market. It’s not intelligence for the world. I think Max would agree with that. It’s law, and even if it’s tax and compliance as well, it’s still relatively verticalized. I think your multiple on invested cash is what you should be looking at, and I’d argue that would actually be the same.

I agree that many of the opportunities we’re investing in—we’re trying to invest in the very best at both stages, early growth and late growth—you can find 10x and 20x companies on both sides. We want nothing less than 5x, and that’s a really high bar in growth.

What did you do in early growth that you didn’t do in late growth, and that you regret not putting into late growth?

That’s a good question. I’m really excited about another vertical AI company called Fieldguide. It’s AI for auditors. We’ve done AI for doctors with Abridge, and AI for lawyers with EvenUp and Lagora; Fieldguide is AI for auditors. I invested early. They were under $10 million in ARR at the time, and the entry price was probably a little under $200 million. They’re just ripping. They’re growing really fast, and I’m really excited for them and proud of them. We haven’t yet been able to back up the truck.

Their inflection has been extraordinary. I think a lot of companies saw their solutions become much more compelling after that magical time in January and February, when the models took a leap forward. All of a sudden, Fieldguide’s solution became wildly more compelling, and the uptake is showing it.

I’m going to push you, but is that really a generational company? No offense. It’s fascinating, but Fireworks is. Cognition or Factory could be.

Sure.

And again, I mean it with zero disrespect to Fieldguide. It can be a phenomenal business, a $10 billion or $20 billion business. But to be in late growth, as Accel calls it—the Leaders Fund—is this really generational?

It depends on how you define generational. If generational now means you have to get to a $100 billion market cap, then I agree with you; it’s probably not that. Some of those others certainly can be. But I think it’s an extraordinary business, and that’s one I wish I’d backed up the truck on more.

And if the price is right, then, to the point we made earlier, the multiple on invested cash can actually be the same.

Exactly right.

Do you have separate teams for early and late growth?

No, it’s the same group. We’ve got four great GPs and an amazing team of VPs and associates going after this. We’ve been working on it. Harry, I’m so fired up about our growth team and the practice we’ve built over the last few years. I’d put our growth portfolio up against any growth team in the world. I really do think we’re going to have one of the best portfolios in all of growth, in all of venture capital.

I spoke to many people on your team beforehand, and they said you really focus on culture. I thought, “Oh, gosh, how boring.”

Come on, man. It’s the most important topic.

For every CEO, it’s like, “I want to talk about culture,” and then you speak to their team and it’s different. Every marketing team is like, “Talk about culture, talk about culture.” I did hear something, though, that I’d like to ask. Talia told me to ask you about “teammates and friends beat colleagues and coworkers every day, and twice on Sunday.”

And twice on Sunday. That’s right, man. I’ve used that line repeatedly for a decade now.

How does that affect how you build a venture firm today?

First, I’ll explain where the line came from, and then I’ll talk about why I think it’s so important in firm building, just as it was then. It came from when I was leading and serving at SendGrid. I’ve always believed that, when building businesses, the hardest thing to do is align masses of humans—so many people—to work in a coordinated and orchestrated way toward a singular goal. That is really effing hard to do, man, and every CEO listening is like, “Yep.” They hear that. They know that. That is the hard bit.

I’ve found it was a strategic competitive advantage if your team—if your people—actually liked each other, cared about each other, and thought of each other as teammates, not colleagues. The reason, Harry, is that teammates work harder for each other. They’re not working for themselves; they’re working for their teammates and their team. You never want to effing let down your teammates.

I get you. I obviously agree; I can’t disagree. But how do I put that into practice and turn coworkers and colleagues into teammates? What can be done?

I’m going to start with the most important basics. You have to be aligned on culture and values. How we do our work has to be super clear, and everybody has to align with it. There’s no changing the standard. You can’t have an asshole who’s a great performer but doesn’t abide by the rules. Whatever your culture and values are, you have to get everybody working in the same way. You have to protect that.

Everybody on the team has to be empowered to say, “Nope, I’m going to call that out if you’re deviating from our culture and values.”

On culture and values, I think a good value is something that could be argued for or against.

Disagree. A value isolated from the organization—sure, anybody could agree or disagree. But you have to have a set of them where you say, “This is our way.” That was the SendGrid way of doing things. We had our four Hs: happy, hungry, humble, and honest. Everything we did had to reinforce those four. That was great for SendGrid.

But you can’t argue that you shouldn’t be happy or hungry. We want unhappy, unambitious liars.

No.

What? No.

Okay.

Well, no.

I’ll push back. Here’s an example. I always said that I loved SendGrid and our culture; I thought it was amazing. If you tracked down any Gridie and asked, “Did you enjoy working at SendGrid? Was it a special team?” I know to my bones that 100 out of 100 of them would say, “Yes, it was.”

The reason is, take the humble H. It would’ve been seventh hell as a place to work for a certain group of people, right? Our humble H was “we, not I.” We’d ask people in interviews, “Harry, tell me about a professional accomplishment of which you’re most proud.” Then we’d sit back and do a pronoun count. How often does Harry say “we”? How often does Harry say “I”? I want the people who say “we.” That was our value. That was SendGrid’s value.

That doesn’t mean it’s the right value. To your point, you could argue either side of that. I don’t care. But for our company and our team, the way we work together is this way, and if you’re not that way, that’s fine. It’s a big labor market. Go find another place to work. That’s great. No harm, no foul.

But, man, it matters that the people you bring in want to work together in a certain way. That’s how you build a team, not just a group of colleagues and coworkers. And, by the way, that’s number one. Number two, which is equally important if you want teammates and not just colleagues, is to spend time together not doing work. I know this sounds so crazy.

When I was leading SendGrid, we flew the entire company down to Mexico every January for an annual kickoff. The whole company—not just a sales team or a President’s Club. We’d lock everyone in a room for a day and get aligned on the vision, the mission, why we were doing what we were doing, and the goals for the year.

The whole second day was: go find somebody from another office and go kayaking together, go mountain biking together, or sit by the pool and have a drink together. I don’t care what you do; just do it together. The reason this matters, Harry, is that all the CEOs listening will understand when I say this: shit goes wrong when you’re building businesses. It’s hard.

When you have a group of colleagues and coworkers, you know what they do first when things go wrong? They point fingers at each other. It’s blame-placing: “Sales isn’t responding to the leads fast enough. Marketing isn’t generating enough pipeline for me.” Or, worse, they just head for the doors because they’re thinking, “Well, I’ll go to the next company, because I’m just a colleague. This is just a job.”

But teammates and friends come together in hard moments when you’re building a business—and believe me, there are many. They don’t start pointing fingers. They come together and say, “Harry, dude, I spent time with you mountain biking or kayaking or sitting by the pool having a beverage. You’re a good dude. We’ve got to figure this out for our team.” That’s how you solve problems. When times are hard, teammates and friends will beat colleagues and coworkers every day and twice on Sunday.

No, I totally get that. I completely agree with that headline. I think I’m very trained and strongly believe that values should be differentiated and arguable. We believe that hard work—the hardest of work—leads to better results. I don’t buy rest days in the gym. You train a different body part. You do cardio, not weights.

Amen.

A lot of people don’t agree with that. I’m chastised for it in Europe. I’m the spawn of Satan for saying, “No, we shouldn’t.” Seriously, it’s not popular. I’ll get booed on stage at labor conferences.

No kidding? Well, listen, I couldn’t agree with you more. I’m the son of two immigrants. Work ethic is everything—that hungry H. My line was always, “I want to work with people who are gonna run through effing walls to be successful, who will not relent.” We quote the Eminem song: “Success is my only mother-effing option. Failure is not.” That’s what I want. I want people like that.

Number two for us: feelings are expensive in investment committees. Sameer, if you and I are in an investment committee, I want to be able to say you’re an idiot for thinking we should pay that price for a company. I don’t want to have to tread around it with, “Sameer, thank you for bringing this. I appreciate the time you spent with the entrepreneur, but I think we should consider…” I’m sorry, we’re both big enough.

Yeah, sure. That can be great for your team and culture. People who opt into that are gonna be awesome together, and they’re gonna appreciate the directness. I would say intellectual honesty in the IC is actually what matters most.

But my point is that it’s great that you can debate, because I’m not right. It’s just our way. But what I don’t understand are values like Zoom’s, which I think are about happiness. No one’s advocating for unhappiness.

Yeah, I hear that. And, by the way, on the happiness one: at SendGrid, I had to write down what we meant by each of the four Hs. Of course we want employees to be happy, but that’s not what I meant by the happy H. The happy H at SendGrid meant that I wanted people who were gonna see the world and our business as half full—the glass is half full, not half empty.

There are people who are gonna be negative, who are gonna be like Eeyore in Winnie-the-Pooh, or just downers. I want people who are optimists, who are fired up to go tackle problems. That’s what I meant by happy—not that I want you to be joyful, though that’s great, too. I hope you’re joyful. I want you to be a positively disposed person who, when things go bad or are hard, isn’t gonna go negative and bring the team down with you.

If we go back to venture, we’re seeing some of the most incredible value accumulation from these businesses. We’ve mentioned Lagora. We’ve mentioned Fireworks. I think you’re in fal as well. How do you think about when’s the right time to sell? With IPOs being continuously pushed, and perhaps with many more Stripes, how do you think about liquidity events?

3. Liquidity and the IPO Window

I joke about being new to the business, four and a half years into venture, but I do know that DPI matters. I’m not that new. I know that liquidity matters. DPI matters. Our LPs aren’t in the business of handing us money in perpetuity, so we’ve got to get it back to them in multiples, and I get that.

Right now, because we’re still so early in this AI wave, it’s not a principal concern. I think these businesses are gonna keep compounding at scale, but we should look for opportunities along the way. Practically, as prudent, risk-aware business people, you should take some chips off the table to make sure you’re getting some DPI and locking in gains for your LPs and your funds. You should.

Do you find that you’re able to get liquidity, given that Bessemer is often a very premium brand in a round? This is a backhanded compliment.

Yeah, I appreciate that, and I do think we have a great brand. When we’re looking for a little bit of liquidity to take some chips off the table and bring DPI back, it tends to be at very late stages of the journey. Most people see it for what it is: they’ve got an enormous position here, they’re gonna make lots and lots of money, and the whole thing is gonna continue to compound.

To take a little bit off the table to get back to LPs, I find people are more matter-of-fact about that now. This goes back to the professionalization of the asset class. What used to feel like a personal decision—“Do you not believe in me anymore? Do you not believe in my business anymore?”—is something people now understand.

At the scale of the businesses we’re building, the size of the checks we’re writing, and the valuations we’re writing them at, I think people get it. CEOs are very sophisticated businesspeople. I’ve been on the other side; they understand.

I totally agree with you there. I think people oscillate too much on, “Oh, they’d be so upset if you sold.” I think founders are actually very rational and smart about it.

Super rational. Tell me, what percentage of the founders you work with who get to a crazy late stage don’t take any secondary themselves as the rounds of financing go through? Most of mine do, and they’re doing the right thing for their families. I tell them to. I’m like, “Shoot, you should probably sell a little bit here, bud, to make sure you’re financially secure for your family.” I think that’s great. No problem.

I think it makes better founders. I completely agree with you. You think with a much longer-term vision—

Yeah. And then they’re willing to swing for the fences, right?

Yeah.

Yeah.

Completely agree with you. Can I ask, do you think we’ll see the number of IPOs drastically reduce as Stripe shows that public markets can be harder than private markets if done well? How do you think about that?

Only a few comments here. One, I believe it’s a permanent structural shift that companies will stay private longer. The asset class of private venture capital is just that much bigger. There are that many more dollars in the market and that many more firms. Companies will stay private longer, and I think that’s a given. You’re gonna see crossover investors spending more time here.

I think LPs are allocating more to the asset class because they get that. They can’t reap the rewards of this generational boom if they’re waiting for them in the public markets. Most of that run-up will have already happened in the private markets. So I absolutely believe that is true. I also believe that we will see an unlock in the IPO markets next year, later next year, and in 2028. I think a lot of companies will go public. We’ve had this window shut for so long.

Can I ask you why that unlock will happen? What will cause it?

I’m an economist by training, and supply and demand is a thing. It’s just a thing. In my humble opinion, what’s going to happen is that a bunch of great businesses will go public and have a lot of success in the market. There will be an obvious appetite from investors in the public markets to get pure-play AI exposure. They don’t want it from just a few companies. They’re going to want it in dozens or hundreds of companies that are pure-play AI exposed. I think the demand will be so incredible that investors will be willing to pay those prices. Then supply and demand works, prices move, and I think we’ll see a lot of companies come out while the window is open.

I’ll say one last point on the IPO thing: having done it myself, I believe it makes for a better company. I deeply believe that. You run a tighter ship when you’re a public company. At SendGrid, when we were building up to our IPO, I used to say, “We need to be IPO-worthy.” And what does that mean? It’s a high bar to withstand the scrutiny of the public eye. To predictably deliver quarter in and quarter out on what you say you will do is hard, but you build a really rigorous and operationally disciplined machine. I hope a lot of companies will take advantage of that. I do.

Do you still buy that? When you look at a lot of companies—Canva released their numbers, and they were great numbers—you have similar reactions. The expectations have shifted, and these companies have those expectations anyway. I would argue that Stripe is able to do a lot of things because of its private nature that Adyen isn’t able to do, which makes Stripe structurally advantaged. I think you’re enabled to be better because you’re private, rather than by the discipline of being public.

Yeah, I hear that. There are pros and cons to anything. You’re right: you have latitude to make big product investments. You can go deep into the red for periods of time if you think you need a shift and have to take a big hit. But on the flip side, there are things you can do as a public company. I have a currency; I can go do M&A. There are pros and cons on both sides.

I get you, but Stripe doing an M&A deal for OpenRoad for $8 billion—I mean, they wanted to buy PayPal for $55 billion.

Yeah.

You have that now. You have liquidity for employees and investors. I’m just wondering: are we really seeing the winnowing of IPO attractiveness?

Maybe. I don’t know. When I talk to the companies we’re invested in—

They still want to go public.

Yeah, they still do. And by the way, in certain markets it’s a badge of credibility. Those companies’ customers care. They’re like, “Okay, these guys are legit. They’re real. They’re at scale. They’re public.” It connotes something.

Ironically, for things like Aura, I think it also totally makes sense: a consumer brand where retail wants to be part of it. Do you think retail should be a much larger part of the LP base of venture funds? If we have so much more in gains now in private markets, do you think retail should play a large part in the LP base of venture?

Harry, I think it’s hard to argue against what we would call the democratization of access to that. I do.

Why? Because I could argue against it. Sorry, I’m really going for it. Bessemer is a very over-funded product. There are so many people who want access to it. If you just had a random people’s fund, I don’t know if it would beat out the hospitals and pension funds. So then you have adverse selection, and retail just goes into shit funds. That would be an argument.

By the way, retail should only go into either giant global platforms with brand, leverage, and power, or super-specialized folks where you know you’re taking risk. If that doesn’t feel comfortable for the retail investor, they shouldn’t play, because it is a high-risk game. There’s a lot of argument these days, especially when the public markets have been ripping like they are, that if you’re a retail investor, you should just put it in the public market, not take the liquidity hit, and let it compound there.

But, yeah, obviously, when I look at our portfolio of companies, I think, “These things are going to continue to grow and compound and deliver gains that are not accessible in the public markets.” That’s where I would go.

Can you paint the bull case for how Fireworks, one of my favorite companies, is a $250 billion business? I think Lynn is generational.

I couldn’t agree more. Lynn is extraordinary, and so is George Hu, who just joined her team as president.

On the first call with me, I’d never met him before. Lynn gets on and gives me to him, and this is before I know whether I’ll be able to invest. I say, “George, I’ve heard so much from Mark, and…” I’m trying to be charming. He goes, “Yeah, yeah. Cool. So tell me, how are you going to increase the value of this business?”

That is my man. That is my man, George. He’s going to cut right to it.

No talk about the weather. Sure, George, let’s go straight into it. Sorry, I interrupted. You were saying about working with George.

We worked together at Twilio. He was deeply involved in the diligence when we were doing the Twilio SendGrid acquisition. We were in the same executive leadership team room with Jeff Lawson, who’s extraordinary, and a bunch of great people, in those trenches together for 18 months. I’ve been doing this for 31 years, and he is absolutely one of the very best software executives in the industry, period, full stop.

Why? What makes him so good?

First off, George is an exceptional thinker. He’s super strategic and very analytical. He’s been doing this for decades, and he knows how to build a go-to-market engine like very few people in technology today. George joining Lin and Fireworks preceded our investment by about a month or two. When George joined, I thought, “Dude, these guys are going to crush,” because you have the brilliance of Lynn, a strategic thinker and deep technologist, an unbelievable technical team from Meta, and now George building the go-to-market engine. That’s the pitch right there. You asked me for the bull case: those two together at the top of that company are going to build an enormous business.

In terms of the TAM, the market, and what they’re doing, I think their general pitch—that enterprises aren’t going to want to rent intelligence from somebody else; they’re going to want to own their own—is the simplest and really compelling tagline for any enterprise CIO buyer. Personally, I think there’ll be a healthy amount of both, but it’s hard to imagine a business saying, “Yeah, no, I don’t want to own any of my intelligence.” Every business is going to want to own some of its intelligence, for sure.

Do you buy the sovereign model argument?

I do. We’re invested in Sarvam in India, and we’re invested in—or investing in—Mistral right now.

So you think countries and geopolitical regions will need access to their own models for obvious geopolitical reasons. Is that it? And to what extent do we go down the stack and say, “We need our own energy provider; we need our own…”?

Yeah. I’ll say, in certain markets and for certain use cases, they’re different. In India, one of the things that makes it such a ripe opportunity for intelligence that’s local to the market is the Indic languages. There are so many different languages, and training the models on their own content is important. That’s multimodal: voice, text, all the things. There’s a very particular argument there. Not to mention, of course, that India has some of the best technologists at scale who can work on the problem.

In the case of Mistral, at the end of the day, if you’re trying to determine, say, your military strategy, there are certain use cases where, if it’s a government use case or a regulated industry, having it be a sovereign model is probably going to feel more comfortable to that buyer.

When you’re writing a check, how does the upside requirement change depending on the vehicle it’s in? For us, I’ll put it very simply: from the early fund, we need to see a fund returner, which means 50x. From the growth fund, we need to see 10x. It doesn’t need to be a fund returner, but it needs to be a 10x. How does that upside requirement change for you?

What you said is actually very consistent with how we think about it. We want 5x-plus funds. That’s just really hard to do at any stage. You can only get there if, at the later stage, you’re putting big dollars behind businesses, some of which may return less than that, but some of which are going to return way more.

At the early stage, by definition, your hit rate is going to be much lower because the businesses are earlier and less developed. You’re going to need some 50xs to account for and pull up all the ones that don’t work out. Obviously, it’s much more de-risked when you get to the growth stage, so the upper bound doesn’t need to be quite that high. You don’t need 50xs in growth, although some of ours may approach that. But that’s not the same bar. I think the blended return gets to the same place.

Do you think you can still do a 5x given the fund sizes being as large as they are?

100%, no hesitation, especially in this era, because of what’s happening in the world right now. When you see businesses scaling at the rates we’re seeing them scale, the value creation is incredible. It’s not hypothetical anymore. I sounded like a crazy person three years ago when I would say this stuff, but now everybody just looks at me like, “Oh, yeah, that’s true. I can think of 10 examples.”

There’s no debating that we’re seeing value creation in this era unlike anything we’ve ever seen. So absolutely, I believe we can generate those kinds of returns.

You can only invest in one of those funds: early, early growth, or late growth. Which one do you invest in?

If I can only pick one, personally, I’m focused on the growth stage, and I really do like the opportunities in late-stage growth.

But I actually think we’ll do better there. I think the outcome scenarios are bigger than they’ve been in a long time, which means the upside and the multiple on invested cash are much bigger. I think we both did Fireworks at, like, a $15 price. I think there’s a 20x in there for us.

I agree 100%. When you can get early-venture return MOICs on late-stage, de-risked companies, I’m with you. That is a great spot.

I will say that we have a world-class early-stage franchise that’s been doing pre-seed, seed, and early Series A for a very long time. If you can buy 10% or 20% ownership early for small checks, that’s fantastic, too. I would love that. We’re going to continue doing that all day long, but it’s definitely an “and.” That’s why I love our model as a giant platform, where the synergy between our early-stage practice, early-growth practice, and late-stage practice is really something.

I want to do a quick-fire round where I pepper you with questions and you give me your immediate answers.

All right. Do it.

Okay. Who’s the best picker in all of Bessemer?

I think we have a bunch: Byron Deeter, Jeremy Levine, Adam Fisher in Israel, and David Cowan out here, who’s been doing it for three decades. But if I had to pick one, I’d say Byron Deeter.

When we were going public, on the morning of the IPO, we were at the New York Stock Exchange. We brought out every SendGrid employee who had been at the company for more than four or five years, so we had a room full of SendGrid employees. The president of the New York Stock Exchange came out to welcome and greet us before the bell-ringing ceremony. He said, “I just want to welcome all you SendGrid employees who are here. You should be so proud. SendGrid management, it’s so great to have you all here. Esteemed board members, thanks for coming and being part of this special moment. Byron, great to see you again.”

He paused and said, “Byron, how many times have I seen you here in the last…” Mind you, this was a decade ago. Byron sheepishly, because he’s a very humble guy, gave a number. I can’t remember what it was at the time; he flashed up something like six. We all laughed that he’d been there for multiple IPOs recently.

I was sitting right in front of Byron, and I started counting in my head. I came up with more than six. I turned back and said, “No, no, you’ve done more than six.” He said, “Six here. NASDAQ was another big number.” He’s had so many companies go from early stage all the way through IPO. It’s hard not to say Byron Deeter.

Byron is the picker. Who’s the winner? What I mean by that is, when founders are choosing between Sequoia, Founders Fund, Benchmark, who’s the one person who closes founders better than anyone else?

That’s a good one, too. Knowing both of those two, I’d revise my answer: David Cowan is the picker, because he finds stuff that everybody thinks is crazy, like Rocket Lab, that then becomes an enormous business, and nobody knows what he’s doing. But when it comes to closing, Byron Deeter every time, for sure, 100%. That guy is a force of nature. He’s charming, funny, and smart as all get out. His experience is ridiculous, and he knows how to have a good time. He’ll host people, whether it’s at a 49ers football game or skiing at his house in Montana. He’s an amazing, amazing closer.

What company is Bessemer not in that the firm universally rues and wishes it was in?

So many. I would say Stripe. We were the developer API venture capital firm. We led the seed, and I think the A and B, in Twilio; led the Series B in SendGrid; and had Auth0, which David Cowan led. We were the developer API company, and we weren’t in Stripe. That hurts. That’s probably the one I’ll use.

You can add one person to the venture partnership who isn’t there currently. It’s a dream add, not a realistic one. Who would that be?

If I could convince Ravi Matre to join us, I’d love to have him over here from Lightspeed. Ravi is an incredible deep thinker who goes deep on the technology and the markets. You look at some of the investments he’s led and made, not the least of which was a $1 billion check into Anthropic in the $60 billion round. He’s really good, and he’s willing to take these giant risks that seem almost borderline insane. I have deep, mad respect for both.

A billion-dollar check. That is some serious balls. You’re just like—

On Bessemer’s round.

The finance director typing those zeroes must just be like, “Wow. This is—wow.”

That’s a lot, lot of numbers.

What have you changed your mind on in the last 12 months?

In the last 12 months, I’d say it’s my deep appreciation for how big the open token economy is going to be. I just didn’t see it 12 months ago. I didn’t appreciate it. It’s a big part of the reason we’re investing as much as we are in fal and Fireworks: fal on the generative media side and Fireworks on the infrastructure side. They’re going up and down the stack. They’re amazing businesses with amazing teams. We’ve talked a lot about Fireworks. I just didn’t appreciate how big that market was going to be 12 months ago.

Final one for you. What are you most genuinely excited about moving forward?

I’ll say two things. One is I’m wildly fired up about all things robotics and embodied AI. We invested nearly $200 million into Waymo, and I just think autonomy, what that changes, and how it changes the way we work and live is awesome.

Is that humanoids, or is that not humanoids?

It can be. Certainly, it’s an area we’ve been looking at. I think we’re probably at the equivalent of 2021 in embodied robotics, relative to the generative AI era. We’ve still got a couple, few years to go, but it’s on one of these curves, man. It’s going to go. It’s happening fast, so I think we’ll get there faster than most people think.

Are you in any of the robotic data companies? We saw Maka just get done by Sequoia yesterday. You’ve got Micro AGI in Europe. Are you in them, and how do you see the data market for robotics?

We are not, as far as I know or can think of offhand. But I know we’ve spent a lot of time around robotics, so I’m sure my early-stage partners are listening and going, “Sameer, come on, man. We’ve met with 20 of these, and we’re going to make this pick.” I don’t actually know offhand. I think there’s a lot of excitement in the firm around that area.

I’m personally looking at autonomy. We did Waymo, but now you’re like, “Okay, well, what about autonomy in construction? What about autonomy in mining? What about autonomy in agriculture?” There are tons of areas.

Oh, this is a good quick-fire. You do Waymo at 115, and Uber has a market cap of 120, 125.

Yeah.

Why am I a moron?

Because you’re looking at the ride-hailing market as the only market. That’s not the only market, Harry. I have a 16-year-old son that I just bought a car for. If I could have bought that car with an option to have a Waymo driver on it and pay them a subscription every single month on that car, or any car that we bought as a family, it’d be a default purchase. Of course, I would do it.

Just like Tesla and Elon have FSD on his cars, I think Waymo has the opportunity to do that for every other OEM on the planet. Are they doing that today? It’s not their near-term focus. They’re focused on ride-hailing. They’re crushing it. They’re rolling out nationally and globally to every city, and they’re gonna build an enormous business on that side.

They’re gonna unlock use cases in ride-hailing that didn’t exist before for folks like your mom or my teenage daughter coming back from whatever. There are so many more things. That will be a very big business, but when you take that same technology and capability and apply it to personal car ownership, and then apply it to trucking and logistics, I think Waymo’s gonna be an extraordinary trillion-dollar business. Easy.

Good. One is that. What was the second one?

It kind of follows your lead, which is therapeutics. Using AI to, as they say, drug the undruggable and figure out cures is, I think, by far the greatest potential and impact we’ll see from this technology. I believe you’re just seeing the pace of things. We’re gonna automate. Wet labs are gonna become automatable, and you put an AI brain that can see and reason over massive datasets in ways that a single human simply can’t.

I think we’re gonna have a pace of discovery in biology, therapeutics, and drug development over the next decade that would have taken us half a century. I really do believe what Dario says about that in “Machines of Love and Grace.” I think we’re gonna see in the next decade what would have taken us half a century. I really believe that.

I think you’ve been so patient with me today. You’ve been— I mean, you’re like, “I came for a nice chat, dude.” And then you’re just like, bam.

I love this, Harry. You always do this. You’re so freaking entertaining, and you’re smart, and you ask great questions, so it’s fun.

But before we leave you today, Blitzy, the autonomous software development platform that’s built for enterprise scale, is proud to sponsor today’s show. Enterprise codebases have grown beyond human comprehension. Tens of millions of lines, decades of tech debt, complexity existing tools cannot fathom. Blitzy starts by understanding. Thousands of agents reverse engineer your codebase into a proprietary knowledge graph that maps architecture dependencies and business logic. That graph surfaces every CVE and zero day hiding in your code before anyone hands you a ticket. Grounded in that context, Blitzy then executes entire epics autonomously. Modernize off legacy systems. Ship new features 5X faster or batch remediate CVEs at scale surfaced by the platform. Blitzy not only generates code, it validates and end-to-end tests every single line, and that’s why Fortune 500 enterprises are shipping enterprise-grade code five times faster. Go to blitzy.com/20vc to learn more. While Blitzy helps you build faster, Flex helps you run your finances smarter. Business owners run their whole financial life on Flex. One platform from business revenue to their personal spend. Float every purchase for 60 days. Tap capital that grows with your revenue, and pay vendors in 170 countries across 32 currencies. Plus, the whole back office, bills, expenses, accounting all in one place, so you spend less time reconciling and more time growing. That’s why thousands of owners use Flex, named one of Fast Company’s Most Innovative Companies of 2026. Visit flex.one. That’s F-L-E-X dot O-N-E, and use the code 20VC. While Flex runs your finances, Vanta keeps your compliance covered. When you’re building a company, you learn that trust is what closes deals. You may have the best product, but no buyers will sign these days without proof of your security. Here’s what happens if you’re not prepared. A prospect asks for proof of compliance. The deal stalls when you scramble. Your engineer gets pulled off the roadmap to audit prep. Every enterprise conversation turns into this horrible fire drill. That’s where Vanta comes in. Vanta is the leading agentic trust platform that not only gets you compliant fast with frameworks like SOC 2, ISO 27001, HIPAA, and GDPR, but keeps you compliant by continuously monitoring your controls. So your deals keep moving, and your engineers really keep building. Now access the Vanta Agent everywhere you work, even if your team lives in Claude or Cursor. And that’s why Vanta’s trusted by more than 16,000 companies like Ramp, Harvey, and Writer. So prove you’re ready for business and get $1,000 off Vanta when you go to vanta.com/20vc. That’s vanta.com/20vc