Everything You Know About Cybersecurity Is About to Break

0:00Tell me how you build a$100 million business in two or three years. If you want to do security right, you need to just forget about everything that happened the last 20 years of doing security. These agents will stop at nothing. No traditional security product at either company detected the attack. It was AI that detected the attack. The 95th percentile back in 2023 was a$14.6 million round cadet. It's a$65 million round and a 314 post. The VCs aren't the ones that can see the future. It's the founders that see the future. There's no greater time to be a founder than right now. I mean, it's just the golden age.

0:30Meet Ed Sim, founder of Boldstart Ventures, a$1.1 billion venture firm that writes the first check behind technical founders. Boldstart has backed what's next across cybersecurity, AI infrastructure, and physical AI with signature investments, including Sneak, Protect AI, Superhuman, and many more. In this conversation, we go deep into the massive opportunity in cybersecurity and AI infrastructure for startups, and also unpack the subtleties that make the current environment both extremely challenging and extremely attractive for founders and how to prepare for it. Welcome to Post Money, the podcast where venture capital meets the founder's journey.

1:16I'm your host, Nilanjana Bhamek, founder and general partner at Converge VC with over two decades as a venture capitalist, investment banker, and operator. On this show, I sit down with visionary founders and investors to uncover what really happens when you raise money and build companies at the bleeding edge of technology. From the investor's perspective to the founder's vision, these are the stories behind the capital, the conviction, and the companies shaping our future. So excited to have you on the show. Welcome, Ed, and congratulations. This is your 30th year as a venture capitalist. Hey, Niladja, thanks for having me.

1:56So excited to dive in. I know that you've been at this game for quite a while as well. So I'm really looking forward to our conversation. You know, I read your WhatsApp blogs quite often, actually, and in them, I always get a combo of deep experience and the energy of someone who's just getting started. So I think that's what you exemplify and that's what it takes to succeed in this business. So you've done a fabulous job building Bold Start. It's been, what, 16 years now that you've been building? Yeah. And so catch us up on, you know, what was the starting thesis and where the firm is at today?

2:34What are you investing in? Yeah. So look, we started, both started in 2010. And this was after I had been doing venture for quite a period of time. And what I saw was that there are a few things converging back then. It was the rise of kind of Amazon EC2. This was pre-cloud days. It was the rise of open source. and we just exited a bunch of companies that sold green plum and dmc and eventually ended up public as pivotal we sold go to meeting and some of these founders are coming back to us saying hey i really would just like to raise a million bucks or a million and a half dollars and i want investors that totally get it and i just can't get a small check and the idea back then was that it was kind of nail it and scale it you can start with a very little bit and prove something out and then when you need the capital it would be the scale because of open source because you can borrow compute from the cloud, which wasn't the cloud.

3:27That was the genesis. So at the time, there were only a few generalist seed funds out there. And we started, quote unquote, an enterprise focused seed fund out of New York. Very bad move because everyone thought, hey,

3:41Ed Sim: if you're doing seed, you got to be in California. And then two is, what the hell is enterprise? Why are you specializing in that? And those are two bad moves because it took us a long time to show people, I think by 2018, to get real institutional capital to show them like, hey, this is actually going to be a thing. And yeah, so that was kind of the genesis. And over time, seed was no longer first. This thing called pre-seed came around. And then, in my opinion, it got very convoluted. Founders had no idea where to go because they're like, I don't know what pre-seed is because it presupposes I need a seed round.

4:11Ed Sim: If I'm already a really good founder, I don't need a pre-seed round to get diluted and then a seed round and then an A round and get me a round. And so I was like, you know what? Let's just get rid of all this stuff and make it very simple. Founders need one place to go. And that one place to go is inception investing, where you're a technical founder. And whether you want to raise 500K or, you know, 25 or$30 million, you know, you should come to one place and we can lead or co-lead that round. And I can tell you all the bits about it. But that's kind of what I started talking about like three or four years ago.

4:40Ed Sim: And that's kind of the lane that we swim in. We like to meet technical founders that are, you know, probably thinking about leaving their current companies and starting a company. They're just crazy enough to say, I think I see the future. I can see around corners and they want to come to us and say, I need a leader, co-lead. They're going to join me in this journey and believe in me and write that first check based on who I am, what my idea is and what the potential is versus like spending months and months diligent, diligent things that don't really exist yet. And that's kind of the genesis of Bold Start.

5:10And you've stayed true to that. So one thing that I'm reading in between the lines there is, you know, to make that kind of investing, you have to be pretty quick with your time and their time, right? So you're making a very quick assessment, relatively speaking. And so unpack that a little bit more. What does that inception? You're getting it right to the point, man. Wow. Okay. Because we like to say it's speed. Our running joke is we tell founders, we don't give you homework. We do the right amount of work so that we can decide whether or not you're - Well, when it's inception, there is no homework other than the person.

5:46I'll give you a great example. I was talking to a founder recently that's raising a$25 million round in AI infrastructure. And, you know, the guy had sold his last company and we had spent a few cycles with him. And he's like, look, you know, and within like a week, we came to a decision. We're like, we're in. Like all the partners met this founder and the other firms were still digging around. And, you know, these were multi-stage firms and usually multi-stage firms, they need to get other partners on board. And it takes time. People are on vacation. People are doing things. And I love the fact that my partners, Ellie and Ellen, when we have something that is really hot and moving very fast, it's like, boom.

6:23They're like, where do I meet? We all flew into New York. We met with the founders for three, four hours after we'd done some work already. And I like to say speed. You need two things if you want to win in Inception because I'll walk you through some other data that now Inception is something completely different in terms of round sizes, in terms of who's coming in. But the point is speed. Speed really wins. And you've got to have conviction in the sense that you need to believe in kind of what you are looking for in founders and actually have the belief in that. And then two is you need to believe that your judgment, we don't get it right all the time, but you need a very strong sense of your judgment in terms of making that decision and writing that check.

7:00And guess what? The thing that's amazing is that when you do this and you're their first partner, they never forget it. Like you were the first one to believe in me. It usually unlocks the rest of the round. It creates a competitive dynamics. It forces everyone else to move faster. And that, to me, to your point of speed, is really, really critical in that. And not saying that we're stupid. It's just that we know the things to focus on and the things not to waste your time on. You're prepared to move fast. That's, you know, that's the catch here. And you're prepared in two ways, it would seem.

7:29One is sort of you already have a viewpoint on the market or not, right? So you have a viewpoint, and that comes from various ways. and your long presence in the enterprise market and seeing so many cycles perhaps gets you there very fast. The second is sort of the, I call it the stomach informed by years and years of the brain. So unpack that stomach a little bit more. Like, you know, what is it? That's worked for you. You know, it's funny. So I'll give you a couple of examples, but you know, what's really funny is that when we look at some of the deals that don't work, like at least 50 % of the time, my partners are like, you know what?

8:04How do we convince ourselves to actually go do this thing. Usually after that first meeting, we're like, there's something that just doesn't feel right for us. I mean, it doesn't mean it's not a great company, but for us, it just doesn't fit the mold. And the times where we convinced ourselves or had someone else call us to convince us to do something ended up being very wrong. So it once again goes back to trusting your gut instinct. The second thing is, it would be, let me give you a great example. We funded Ian Swanson and his co-founders, Dee and Badar, for Protect AI. And we started talking to them a year before ChatGPT had come out.

8:39And we had known Ian for 10, 15 years. And Ian and I met at a conference. In a conference talking about the future need for AI security. And everyone in that room was like, why the hell do we need AI? It doesn't even exist. This was in 2022. And I was like, yeah, we totally need this. And here's why. And Ian texted me. He was like, let's go on chat. So we chatted for 30, 45 minutes. And we're like, look, Ian, just go get another co-founder or two. They can build with you and come back to us. He came back to us. we ended up kind of leading the inception round with a crew. But let me tell you what I didn't do.

9:08I didn't go around and say, let me talk to 10 different CISOs and say, hey, let me ask you about a problem that you probably don't even have right now. And then I want to ask them, if you had this problem, would you buy this thing? I mean, it's kind of intuitive and obvious, right? Right, right. There's no budget for it. I mean, I'm not going to go say there's no budget. There isn't a budget for it. But that's what's so exciting is that if you intuitively ask, I call it intuitive TAM. If you ask yourself, do I think like if Tesla data got hacked or if the model got hacked or for credit card model from Goldman Sachs got hacked you know would that be a problem yeah the answer is yes hasn't been hacked yet I don't think it's been hacked yet but could it be hacked yeah these are five different ways it could happen and so intuitively like okay we did not even believe I didn't know the chat GBT moment was coming a year later that that opened up the doors but we were like look this guy based on his experience at AWS he was selling these large enterprises I I knew that those guys could build a product and bootstrap probably to$10 million of ARR within the first two or three years.

10:06I didn't know if that$10 million would be$10 million for five years from now or that would be$100 million. I had no idea. But that was kind of the risk we were undertaking. But calling 30 CISOs for diligence, that was not going to help me make that decision. It was like, can these guys navigate? Can they build stuff and ship it really fast? If the world changes, can they do different things? And those were the things that we're looking for. Can they recruit great talent? And those were the things that, and can they sell, right? Those are the things that we looked for versus spending cycles doing these calls for markets that don't exist.

10:36To writing the investment memo, right? Gathering material to write the investment memo. That's not early stage investing. No. As you know, it's all about the people, right? That's right. Can they adapt? Can they move fast? Can they break things and make money? Excellent. And do they give you energy, right? Because like I said, yeah, you need to wake up every day because time's going to be hard. It's going to take a long time. It's like, do you want to? And you're being very humble. That was a fabulous exit for you. I think didn't Palo Alto Networks acquire the company for a very nice sum. Yeah, thank you.

11:12So congratulations. I'm going to have a million or so in three years. Yeah. Excellent. That wasn't kind of the ultimate goal, but it worked out well. But it worked out well for Ian and the team and for the fund. Excellent. So you brought up cybersecurity. I know you do much more than cybersecurity, but I always think of you as a very astute cybersecurity investor. And so I'm going to sort of maybe double click on a couple of things. So I was at Black Hat a couple of weeks ago and my head exploded and not just from the heat. Yeah, it was about 120. 120 dry heat is a far cry from Boston, right? That's right.

11:52That's right. So, boy, it was hot. But, you know, a lot of very interesting observations. But to me, one of the most sort of dystopian presentations was perhaps, you know, the OpenAI opening up on how the jailbreak happened and attacked Hugging Face. And I'm going to, I think you were there as well. I was. And so I will let you comment on it. But it sort of summed up, for me at least, like everything that was new and instantly obsolete about cybersecurity in that one presentation. So just unpack you as an... I think you nailed it because the first few days didn't even have the presentations. The presentations were like Thursday and Friday, right?

12:39That's right. That's right. I kind of think about this year has been insane because I joke that everything that happened at RSA didn't matter because two weeks later. It didn't matter. Two weeks later, Mythos, and that whole rumor came out. And then, and here's the funny thing, the stock prices of CrowdStrike and Palo just took a massive dip, right? That's right. And then I feel like everything in Black Hat didn't matter because the talk that OpenAI gave happened on that Thursday and Friday. And the craziest thing, and I was talking to someone about it today, actually. Well, let's take a step back.

13:09I believe we're in a world where, do you remember when, like during the web era, the fastest growing companies were building their own infrastructure and they're open sourcing all the greatest software out there, right? And that was kind of like, and I feel like, you know, this is where like Confluent came from, Kafka, like LinkedIn and Uber open source stuff, you know, Lyft open sourced a bunch of stuff because they were experiencing growth that like no one had ever seen before. Okay, so that's kind of that web era. I feel like if you want to do security right, you need to just forget about everything that happened the last 20 years of doing security.

13:42So everything that happened in Black Hat, it's kind of like you're either like autonomous attack ready or you're just stapling on co-pots and agents onto things that just happened before. And the reason I'm saying is that the people like the open AIs, the people like the anthropics, they're investigating and building things and seeing things that not many people have seen already or yet. And that's kind of what I think the big disconnect to your point is, is that when that thing came out, people like, holy crap, these agents will stop at nothing. And we knew that they would stop at nothing to solve a test.

14:17And in this case, they went through a, you know, they went through and said, OK, I'm not smart to do it. Let me figure out if I can break out into the Internet to ask questions. And they found this JFrog artifactory package manager and they started leaving notes in there for other agents to coordinate. This is agents coordinating with other agents. So I think the big unlock was that how do you protect a company in a world where agents firms are going to stop at nothing to just continually attack your infrastructure and try to do things? And the only thing that you can do is throw more compute at it eventually, right?

14:50I mean, meaning that you're going to have to eventually become autonomous on the defense side, which, by the way, is a very, very hard thing to do. So anyway. Very hard thing to do. Yeah. Yeah. So first, maybe some of the listeners who may not be in the day-to-day weeds of sort of AI jailbreak. So this was an attack that happened from sort of internal testing at OpenAI where a swarm of agents worked over two months to communicate with each other, leave messages with each other, share exploits with each other, undetected for two months, broke out of sort of every guardrail to not have access to the Internet, got access to the Internet and attacked sort of at least hugging face that we know of publicly.

15:35And guess what? No traditional security product at either company detected the attack. It was AI that detected the attack. And guess what? To the extent OpenAI shared how they are resolving the matter, or at least even trying to understand it, no traditional security log tool has been used. It's codex, if I'm not mistaken, that has been used extensively. So it almost felt like, you know, if this is the future, then sort of, you know, all those public companies and their capitalization is sort of almost instantly down to zero. Well, I think you summarized the details. Well, but by the way, if you look at the talk, part of the answer for them was that they would need to look at logs in real time at a scale that is unprecedented.

16:23Right. That's right. that web scale thing. And this goes back to my one view of the world also is where you're seeing that observability and security log data are all converging at once together. And if you look at the most sophisticated security people, they're sitting inside a cloud and asking it questions and sitting on top of observability data, security data, and other things, trying to see things. But it's not fast enough. It takes 15 minutes, it takes 20 minutes. That's right. So human in the loop is not going to work. No, you're going to have to do that pattern match in real time. You're going to have to have a real-time signaling platform.

16:52I have a company, by the way, that does that called repr.ai, grepr.ai, that sits there and looks at the logs and puts it in real-time streaming, where you can build apps on it. But the point is, is that there's so much infrastructure that's going to need to be built at a scale that is agent scale, that is swarm scale. And the problem is twofold. One is that defense, even in sports, like if I'm a wide receiver and you're defending me, I know where I'm going. You're always going to be just a tad slower than me kind of catching up. In this case, in security, there's two things. One is the models that the Chinese are offering for open-weight models are pretty damn good.

17:29And these hackers are getting full access to it. And then the other side, you've got the models in the U.S. They have to go through regulation that are slightly crippled. So you add that. Then you add the fact that it doesn't really matter if the attack works or not. Because if the only thing that allows me to attack you more is compute costs, then I as a defender I can't just automatically autonomously say yes to every single vulnerability I find and fix it because there are dependencies kind of up and down that chain you have no idea what's gonna break you know perhaps fixing this thing may be the wrong thing to do because I can break lots of other things downstream that takes time to figure that out so there's going to be this massive lag and the question is how long it takes to close it I think it's gonna be really important then I also think that most people are going to start doing their own offensive agentic red teaming.

18:14Even one of my old companies, Sneak, which is now reinvented for the AI era, sits on top of all your code. It knows everything about you. And it's automatically doing agentic red teaming on the systems that it knows you have, on the data you have, so that you can block those things before someone else does. Anyway, it's quite a game. Yeah. It's quite a game. So maybe just what does that all mean? Our job as venture investors is always to pick the best teams, be their partners at inception, early stage, what have you, help them sort of build strong venture syndicates, help them raise money, and then sort of, you know, grow from there.

18:52So it almost feels like the pattern we have used in the past to detect those teams may have sort of, may have some holes in it, you know, is sort of knowing what internal security requirements or XYZ Bank or what the CISO is looking for. I mean, is that, how relevant is that anymore? And, you know, is this insight and knowledge? I think one of the things that to me was very, very interesting is when you go through the description of everything that broke and the level of knowledge and insight that you need and the infrastructure that is emerging in these AI labs, that is the real knowledge and insight.

19:34That is not sitting inside of, you know, I don't know, CyberArk or Palo Alto or what have you. So it feels like the insight, the capability for these solutions that would be venture-backed may be sitting in these AI labs. I mean, how do you react to that? I mean, how should teams be thinking about starting a company? That is 1 ,000 % correct, right? It's the same thing as the web scale back in the day where they experienced growth. They actually were the only ones that could build infrastructure because they experienced it. I also feel like the people that are building the next-gen security companies, I mean, look, I don't know about you.

20:08we probably see lots of security founders come around saying, you know, I talked to 40 CISOs. I've got five as design partners and my next two quarters are pretty good. And I tell them, all right, great. I see you can build a nice five to$10 million business in two years, AR wise. But what are you really worth then? Tell me how you build a$100 million business in two or three years. And I say, all right, so like, what do you mean? I'm like, show me something absolutely insanely crazy. Show me, tell me something that, that, that, that is literally going to blow my mind where the CISO may say, no fucking way.

20:37And they're challenging, right? Because if it's just so obvious right now, it doesn't matter to me, right? And so then I go, the next layer, if I'm going to look at a founder, like a lot of them have deep, deep AI security kind of research expertise. And I think that's going to be absolutely critical. So that's become table stakes, right? That is table stakes. Like, you know, everyone thinks they know AI, but they really don't. And then, you know, I'll give you some examples. We funded some things a couple of years ago So because, look, there's not going to be one solution to all of this. As you mentioned, it's very complicated.

21:08So we funded a company called Gain Security, Gain.Security. And they want to build agent-aware or business context-aware network segmentation, like micro-segmentation on the fly. So what does that really mean? That means, hey, if you have something that's a sandbox, it knows what good traffic coming in is and bad traffic. And if something's happening that doesn't look right, it automatically blocks it right there and just keeps that kind of node or that sandbox kind of locked down. But you're going to need solutions like that that are completely X-Gen, they're completely different. And back then, you know, when we funded like 20 months ago, we were like, network security, why?

21:41Ed Sim: Like, who cares? Like, you know what I mean? So I think that's kind of what you want is people who say like, no, I don't believe you. Who gives a crap? Right. Or like, you know, like it has to be somewhat controversial, I think. That's right. to start if it's going to be interesting and if you weren't going to take a big swing. And of course, the founders came from deep domain expertise. They came from hacking and, you know, blah, blah, blah. But that's kind of what gets me excited. Okay. Last question on the cyber topic. What's going on with these massive, you know, 100 million plus seed rounds?

22:12I mean, is this playing the old game of, you know, we have to rise above the noise at RSA and Black Hat and get attention from CISOs with expensive booths and dinners? Or is this inference cost? Help me understand, like, why raise$100 million in your seed round when the tech you're building for may change tomorrow? I think there's a few things going on. Some of them aren't$100 million seed rounds. They may be like, you know, the initial round combined into two companies, right, into two rounds. Like, for example, I know that one deal that we co-led with Ghillie called Surf AI, we announced$57 million of funding out of the gate.

22:46But it was really an initial round of funding that Gili and Cyberstarts and Boldstart did, and then Excel coming in for a$40 million A round after that. So that's one phenomenon. The second phenomenon is that I think there are only so many ways to make money in security. One is to take something that exists already and just do a 5 to 10x better job of doing that. And that would be kind of reinventing old markets. Those are the ones that typically require a lot of capital, right? Because there's an existing market there. You've got to out-muscle kind of existing incumbents. It's typically a second or third time founder that exited prior.

23:23And you know what? There's a lot of cyber funds with a lot of money. And they're like, you know what? That's the easy one. Just throw$100 million at that round, right? Let's give it a go. And I'm not going to pick on anyone in particular, but I think there's a company called Mutiny, maybe, where the founder is a well-known founder from Twistlock. sold it for$450,$500 million to Paolo. And he started a new company and I think he raised$60 million and I think he's giving some of the capital back. And the point I'm just saying, amazing founder, amazing pedigree, strong technical team. This is not really easy.

23:55This is hard when you go out there and they ended up going back in the container business and selling stuff. So I'm just saying that if a founder like that has a hard time doing it, this is not for faint at heart. And capital alone is not going to separate you from anyone else. That's right. The third part would be anyone going into the hardcore compute cycle. If you're going into hardcore compute and if you are in the token stream as a security company, then you're going to need the money for GPUs and training and all the other things. Oh, absolutely. I think that is very justifiable capital.

24:22I get that one. Okay, well said. Then the other ones would be new markets. When there's brand new markets that no one understands, that's when the rounds are typically$10 million, right? $8 to$10 million. And that can require lower capital because you're just experimenting and building, and then you can scale and do your$30,$40 million. So those would be the two reasons. Yeah. Okay. Well explained. Thank you. Let's get back to, I know you do a lot more than cyber. Just catch us up on sort of what is interesting and exciting for the firm today in this AI climate and maybe share some new investments that show the breadth and arc of the firm.

25:03Yeah. You know, what's funny is, I don't know if you noticed, but I changed my, I've been doing right in that newsletter for seven years or seven or eight years. And I changed the title from What's Hot in Enterprise ITVC to What's Hot in AI Infra. I didn't notice. Yeah.

25:18Ed Sim: It was a subtle change. And I just want to say that our roots has always been kind of investing in technical founders building hardcore infrastructure. And what that really is changes over time. So if, you know, because I'm a believer that at inception, the VCs aren't the ones that can see the future. It's the founders that see the future. We just say, okay, I do believe you. And I think you're going to be really good at it. And over time, we've always done different things. We've always strayed away from the main goal, but still sticking true to our source. So when blockchain infrastructure came around and that was kind of cool, you know, and a lot of the founders, And if you remember, we're doing all this blockchain infrastructure stuff.

25:54You know, we did stuff like Blockdaemon. We're in the initial round of Fireblocks. Hypernative. We just did this thing, Trace Finance. We sold to Fireblocks. So that was one. Two is, three years ago, we started seeing a lot more stuff in robotics. And we had some folks. My partner, Ellen, who's in Boston, who you may know. She was a former founder that we backed and joined us. But she had some friends that were, you know, at DeepMind at Boston Dynamics. We used to send diligence over to them. And one day they said, you know what? I think we're going to start something. and we believe that scaling laws will work in robotics, unlike what most people think.

26:27We're very frustrated because, you know, Google and these other places aren't investing enough in robotics, and it turns out that two of the three co-founders were on the original Pomi paper that applied transform models for robotics. Fast forward, we ended up basically on, you know, them alone investing in generalist AI. We led the first round with NVIDIA, and now they just announced kind of a new round at, I don't know if it was renounced, but it's in the$2 to$3 billion range. Two and a half years later, they have version 1.5 of the model. Basically, they're building out the robotic brain. They're building out the foundational model for robotic hand dexterity, but it's going to expand.

27:04And they had a video last week, version 1.5, that blew up on the internet. I think it was a trending item on X. It had 3.8 million views. And basically what it showed was that, hey, I can show a robot a simple task, pick up a phone, put it in a box, and it actually literally copies me. It was like the copy-paste moment for robotics. So that to me is exciting, but you may ask why. The other opinion was software. As a technical founder, they had a proprietary data mode. They actually were building their own data. And because of that, we've done two more robotics deals. And because of that, we did a bio AI foundational model, which was very similar, very smart people from domain.

27:38We've done stuff. We did an AI mining company for minerals. We've done. So because of these things, you just have to kind of constantly adapt. It's not like we woke up and said, we're doing bio or we're doing robotics. It's just we started seeing these signals and the founders were great. And we're like, you know what, let's do that because that does fit enough into our world. And I think that's really, really important is to remember where you came from, but always remember that you need to keep changing and shifting because if you don't, we'd be dead. Right. The world is changing. You're adapting with it, but you're sort of following the founders that you know and trust and they're showing the way to you.

28:11Yeah. Yeah. Within that kind of domain, you know, kind of that nerdy, tactical, inquiry domain. And, you know, these are all moonshot plays. We even did a chip deal last year out of Boston as well. So we're just following kind of what these smart founders are doing. And I just like to say is that it's not about obvious or non-obvious or consensus or non-consensus. It's about being bleeding fucking early, right? You're so early that people look at you and they have to squint their eyes and see. I don't even know if that horizon may come. I don't know when it may come. But if the founder is willing to work on it, and it was their life's mission before, and it continues now in a form where they can focus on it every single day, That is what gets us at Bolster excited to partner with.

28:50Right, right. And you're reminding sort of VCs that the place where DPI actually works beautifully is early stage. I mean, that's where you got to go in early, pay whatever the market price may be at that point. It sort of averages out. But if you can make it work, that's where the returns are also just. just versus. Unfortunately, we're not the only ones that see that now. Yeah, right, right. So let's get into that. So do you think platform funds have an advantage in this market? I wouldn't say that they have an advantage, but I can say the world has changed. So when I wrote this article about, you know, inception investing, the race to be first back in 2023, you know, the median, I just pulled out some data from PitchBook today, the median size of seed round, Ben, was$3.8 million with a$15 million post.

29:46That was median in 2023. The median today is a$7 million round at a$34 million post. Okay, so that's kind of where we're at today. The second data point that you should understand is that I like to see the spread between the 95th percentile and the median. Yes, I was just going to say, does the median median even matter in our business. Exactly. The 95th percentile back in 2023 was a$14.6 million round at a 57 million post. Right. That's the market. And so now the 95th percentile today is a$65 million round at a 314 post. The spread between 95th and median was 3.8x in 2023 is 9.2x now. All right.

30:29The second thing I'll tell you is that the amount of rounds, and I just did it cost AI slash ML because back then it wasn't AI, but AI slash ML. So did you know that 42 % of rounds now at inception, and this is taking pitch book data where it says this is the first round of funding, 42 % of rounds are greater than 10 million. And when the rounds are greater than 10 million, 55 % of those rounds include a fund in size that's 500 million or greater. So I think that What it really tells you is that, and there's been more deals. So is it venture capital funds are dictating this fundraising strategy of these companies, or is it the other way around?

31:08Like, what's going on? Well, I think it's both, right? Because I think that when you're going after a compute-heavy opportunity, if you're doing a Neolab or doing anything where you need compute. Sure, let's exclude those. I get inference cost and wills, fine. If you exclude those, then what do you think is going on? Well, a couple of things. One is that, look, all these multi-stage firms know, look, the only places to invest for them right now is either first or when scaling. Like the hardest thing to get done right now is an A round or even a B round because they're like, I'd rather just pay up and come in later.

31:43So why not buy as much as I can? Because if this company is successful, the cheapest round ever will be that first round. And these option checks, when they're$2,$3 billion in size or even greater than$500 million,$10 million,$15 million round, it doesn't really matter to them. But if it works, it's going to really, really work. And they're going to be able to put a lot of capital in these companies. So I think it's a function of multistagers all coming into the rounds. And so as a smaller fund, let's just say for the fund managers out there, the decision you have to make are a couple of forms.

32:11One is, do I want to lead and co-lead rounds? If I do, it requires a fund size to lead certain kinds of rounds so that you can have ball control. Meaning that when I see a founder I like, I'm literally going to say, I'm going to dictate terms right now. I'm going to lock my position in versus calling friends. And I think the big question for these founders is, do I lead or not lead? And then do I do rounds that are$10 to$20 million in size or not? Because that's also going to dictate how big a fund you're going to raise. And then the third thing you're going to say is, do you want to participate in rounds that are bigger than 30 or 40 or 50?

32:40And that's another question altogether. And usually those are reserved for the multi-stage firms because they can absorb many of those. I mean, in our situation, our last fund size was$250. So I'm leading a$30 million round in a company now, and we've done some of those in the past. I'm not going to do a lot of those. But we can also already check a small as$500 million and it will still matter to us. So that's the rails that we chose. But I think everyone's got to make a decision what kind of business they want to build, how they want to do it, whether they want to lead, and what rounds they want to do.

33:07Right. Right. I mean, underlying all of that math that's happening at whichever firm, whether you're at the early stage, whatever fund size you're playing with, you know, when it is this phenomenon is going on, a lot of it is assuming exit multiples that are far greater than today. Yeah. So we're all banking on both M &A multiples and IPO multiples expanding significantly in the next 7 to 10 years. And maybe that's possible. Yeah, I would say yes and no, depending on, look, I mean, at$250 ,000, it's much bigger than$1 million, our first fund size. Our fund five was$150 ,000 in size where Protect AI came out of, and we returned most of that fund just from one deal.

33:55But I think even at 250, we don't require a trillion-dollar outcome. Look, we would all love a trillion-dollar outcome to launch a dollar trustee. We all would.

34:03Ed Sim: But we don't need one of those to actually make the math work. That's right. So I think that's kind of the balance that we chose to kind of make is that, look, I mean, generalists already in two and a half years is off to the races right now. And we still think there's a lot of headroom. But I don't want to build a business where I'm dependent on having one of those because it's a tough business. because everyone thinks they're going to have one of those, but there's only a few people that are going to get those. And God bless someone to do, and I hope that we both do, but we also need to find other ways to cobble together our returns as well, right?

34:38Because not everyone's concerned. That's right. I'm going to sort of maybe ask slightly related questions on that. So, you know, a lot of folks that listen to this podcast are other younger venture capitalists who may be thinking about starting their own firm or spinning out of wherever they are. Is this a good time to start a new firm? And what would your sort of advice be to a new spin-out firm or a new firm in this climate? I think it's never a good time to start a fund. I'm not going to encourage you, but if you're a founder, you have to have a founder mentality because like being a starting your own firm is like starting your own company.

35:15It's like starting your own startup. So if you have the passion and you feel like you have a, excuse me, an advantage or that you see something that others don't see, then by all means, like, but all I'm telling you is that it's not easy. Like, you're going to have to, you're going to go out there, it's going to take a long time, it's going to take a long time for people to pay attention to you, but you have to have some unique insight in which you know that no matter how many no's you get, the 500 no's you're going to get, that that 501st no, I mean, yes, that meaning that you can get that yes, that's the time to keep going, right?

35:46And so anyway, but secondly, I would say that it is hard though at this point in time to get people to pay attention, right? Every time I talk to LPEs, they're just re-upping because all the larger firms, the platform firms, are doing really, really well, right? I mean, Andreessen is crushing it right now. Martine had an epic week last week and God bless them, and they're just crushing it right now. So they're showing that scale actually matters and works, right, to a certain degree, which means there's only two ways to make money. But going back to your point, it's easier you go early or you're a platform fund.

36:13Those are the two ways. I mean, or it can be a massive growth fund, right? I mean, there's data from Code2, for example, showing that it's easier to go from 1 billion to 10 billion than it is to go from 100 million to a billion. And that at the scale, when you go from 1 to 10 billion, it becomes astronomical. So yeah, there's growth too. But I think this game of early, everyone's in it. Everyone's freaking early. There's no one that is saying that they're, you know, even like all the big platform firms are like, I'm inception or scale. That's it. And so this is probably the most competitive market that's out there.

36:44So if you're going to start something, and most people that start will start kind of that earlier stage, just be aware of kind of what you're getting into. It's a lion's den out there, but it's the most exciting time. Yes, very well said. Let's flip that on its head and sort of the same question to founders starting a company today, right, in this climate sort of for, you know, what is a lot of experience there, 30 years of experience. You've seen many cycles. There are a lot of parallels to what has, you know, happened in the past. You touched on some of them from the technical side, but from a company starting point, as you look to the past and you look to the future, look at the current market, look at sort of the capital market dynamics, you know, to a great team that's thinking of spinning out, sort of give sort of the top three, you got to keep these three things in mind.

37:32Yeah, look, I think there's no greater time to be a founder than right now. I mean, it's just the golden age. And I feel like we are still kind of in that first or second inning, kind of what's going to come down the pike. I mean, even today, I think Sam Altman was saying, or people were saying that, you know, AGI might be coming at the end of the year. Whatever your definition of that means, or maybe RSI is coming. I mean, but that's going to happen faster than we think. And it's hard to keep up with even the models, right? All the open-weight models are coming down the pike. So it's a great time to start something.

38:00The two is at the same time, everyone's starting something, right? So you have to be aware that the bar to start something and the bar and the mode is kind of, you know, it kind of almost negates itself. So if you're a founder, like, and you have this unique insight, if you have something you've been working on and just thinking about and it's been your life's mission, by all means, you should go do it. Right. I mean, and so just and think about kind of to yourself, you know, the business that I build is what am I building for that supports kind of the model from six to 12 or 18 months from now?

38:30Not for today's model. Right. Because I think if you already start from a starting point about marketing the model today, you're already done. Like things are moving so, so fast right now. and so yeah that's kind of what I would say we're seeing founders like literally just kind of come out of the woodwork you're probably seeing tons of them too and they know that in order to win not only are they going to have to you know come up with a unique idea but they're gonna have to run really really really fast and they're gonna have to be a talent magnets as well because right now I think there's almost more capital than there is talent available and the gap is getting very big right so you want someone that can accumulate amazing amazing talent and build this mission, right?

39:07Because the amount of money that people are getting paid right now, when engineers are paying three, four, five hundred, eight hundred thousand dollars for a startup, you can't really pay a lot of those people unless you raise a hundred million bucks. So you're gonna have to sell them on a mission. You know, I know, for example, generalists, they're able to get people that are so mission oriented that, you know, they're in it for the equity, right? And a lot lower kind of current day comp. And that's the kind of founder that, you know, we want to back, that you want to back, is that I can build this mission and get people fired up.

39:35So I think that's part of the two is that you put all that together and that's going to, what's going to, was going to help you win because if things don't go well, they're going to buy your team. They're going to buy the team that they couldn't put together in the first place. Right. And so that if you can build that already, then that's your downside. And your upside is building something really, really massive, having a lot of fun doing it and, and kind of changing the world. Right. And I think that's exciting. Well said. Well, that is a, that's an excellent way to sort of, you know, having asked the last question on this awesome recording, which I am sure will get a lot of listens.

40:09And that is sort of just looking to the future, share a couple of new investments that you think are sort of, you know, emblematic of what's coming. Very, very interesting. I'm working on a new cybersecurity company that I can't give a lot of details on, but it's going to be a$25 million plus round. And it's very relevant to everything that we talked about earlier about agents escaping sandboxes and preparing people for two years down the pike. And it may involve models and not models. And that's all I can tell you. But it's a pretty exciting thing that I'm working on. But it's kind of a two-year-old thing, going back to your point of trying to vaporize everything that we saw at Black Hat.

40:49Another thing that we're doing, you know, we just invested in a new robotics company. And I'll just tell you that we, and this is Ellen, I think, worked on this one. It worked on one of the world's foremost experts in understanding hands and hand movement. And I think one of the biggest limiting factors right now is the ability to get data on hands, like fine-grained movement for these five jointed fingers. And there's a new hardware software solution that can allow this to happen instantaneously

41:15Ed Sim: and gather tons and tons of data right out of the gate. That's the big limiting factor in robotics right now. So we are very bullish for robotics at the moment. And then the last I would say is that hopefully we have a bio-AI company that thinks it is on the verge of discovering new molecules. So I think that also is very, very exciting. Yeah. So those would be three things that are kind of. So excellent. So new cyber frontier, physical AI. Yeah. The new word for robotics. Yeah. And bio meets AI. Yeah. And new sort of drug discovery. So very exciting, broad and future-looking, forward-looking.

41:57Thank you for doing this. Any sort of anything I should have asked you that I didn't ask you? No, I think you covered it all. I mean, I don't know. All I can tell you is this, Anandja, and I know that you have incredible energies and

42:10Ed Sim: you've got this new podcast and you're reinventing yourself always, is that, I mean, I was there at the beginning of the internet days back in the mid-90s, and I thought that was fast. But when you look at it, we didn't have mobile. It was 28.8 dial-up speed, right? I don't think people understand what 28.8 tool bits are like. I have never seen a world that moves this fast right now. I'm just trying to make sure that you understand what your true north is because it's easy to get distracted. That's the only thing I say. You have to keep your sanity. You got to find some things outside of doing this that try to keep you sane and keep you on a longer-term of a star that is more than just getting stuff thrown at you every five seconds.

42:47So that'll be the only thing that I encourage people to think about. Well said. And thank you for doing this. And I hope to have you back on the show in a bit. And we'll talk about what's new and hot then. Thanks for having me. Make sure to subscribe to Post Money for weekly conversations with thinkers, founders, and investors shaping the future of capital and startups. I'm Nilanjana Abhamik, and I'll see you next week right here on Post Money.