Visualizing the AI boom (or bubble), with Carta’s Peter Walker

0:00If you've spent any time building AI products or leading technical teams, you know this. Transformation doesn't fail because of ideas. It fails because teams can't move together. Enter Atlassian's Teamwork Collection. It has planning in Jira, documentation in Confluence, video updates in Loom, and now AI agents in Rovo, which connects the dots across your work so nothing gets lost. It's one AI-powered teamwork platform designed for how modern teams actually build. Learn more at Atlassian.com slash TeamChanger. That's A-T-L-A-S-S-I-A-N dot com slash teamchanger Starting a business comes with its share of ups and downs, which is why staying true to your vision is essential.

0:53A non-negotiable for Romeo and Milka Bregali, Capital One business customers and co-owners of Ra's plant-based restaurant in New York. Romeo and Milka took a leap of faith when starting their own restaurant, gutting an empty space and building it from the ground up. every pipe, every wall, every detail. But building from scratch came with a heavy financial burden, which is when they turned to their Capital One business card. With the flexibility of the card's no-pre-set spending limit, they were able to spend more and earn more rewards while bringing their vision to life. Today, Raz's success is proof that with passion and the right support, it's possible to make your dreams a reality.

1:30Learn more at CapitalOne.com slash business cards.

1:39I came in to build a research team. So I would be building charts and graphs and data analytics and putting them out in big, long-form white papers and all the old-school stuff. That's Peter Walker. He's head of insights at Carta, which thousands of companies rely on, including my former startup and my current fund. Peter came to Carta with a background in media analytics and data visualization. His goal was to turn Carta's massive amounts of data into key narratives about startups and investing. But simply hosting webinars with slide decks was not working. We found it worked okay, but it didn't catch on the way that we thought it deserved to.

2:22The data is so good, it should be catching fire. And what really changed is when you started putting it out from a personal social account or you had somebody behind it. My name is Peter. I run the Insights team here at Carta. and I'm super excited to walk through our latest data on what's going on inside of VC Funds this morning with you. I think it's just a new kind of wave of marketing where it's person-driven. You know, I can respond to comments in the way that a human being would. I can get into arguments. I can have fun debates. I can have an opinion about this data instead of just stating it as plain facts, which can get pretty boring.

2:57That's been a completely surprising part of the work, but it's been amazing. Peter is filling a huge gap with his work. Startups and privately held companies are a major part of the economy. And it's where so much AI investment and growth is happening. But the way they operate is a lot less transparent than publicly traded firms. Carta has all that data. And Peter digs into it to find the trend lines that matter. He's helping investors, founders, and the public see what's coming next. The results are so interesting, so let's get into it. I'm Rana El-Khalyubi, and this is Pioneers of AI, a podcast taking you behind the scenes of the AI revolution.

3:53Peter, welcome to Pioneers of AI. I'm so excited to have you on the show. Well, thank you so much for having me, Rana. Happy to be here. I want to set some context for this conversation. You and I follow each other on LinkedIn. I'm a huge fan of your work and especially that you bring data-driven insights to this overhyped world of AI and VC and startups. So thank you. Well, thank you for the follow. I appreciate it. It's been pretty cool getting to build all these graphics for startup founders and investors over the past few years. Not a lot of data in this space, so to bring even a little data is helpful.

4:27Absolutely. So Carta is a software platform that, you know, founders and investors use to track their options and financing rounds and investments. And my company, Affectiva, before we got acquired, we were on Carta. And my fund, Blue Tulip Ventures, is on Carta. So I'm a big fan. But I was telling my daughter, who is 22 and she's an anthropologist, today that I am interviewing you. And she was like, what the heck is Carta? So for people who are not like spending their everyday waking minute thinking about startups, what is Carta? Yes, for the anthropologists in the audience, what is Carta? Great question.

5:08At the simplest level, Carta does two things. We help founders manage their equity, meaning we help them grant ownership in their businesses to advisors, employees, investors, whoever they want to give a little bit of ownership in that company to. And on the other side of the business, we help fund managers like yourself manage their VC funds. So we have founders on the one side and funders on the other. We're interested in helping founders and funders make better deals in venture and then hopefully broadening ownership to more and more people over time. Yeah, amazing. And your title at Carta is Head of Insights.

5:43So tell us more about your role and the scope of your work. Yeah, it's a little bit of a strange cul-de-sac of a role that I've found myself in. And I don't know if it exists at many tech companies, but I think it's awesome. Did you define that, by the way, when you came into Carta or did it exist already? My old CMO and I sort of co-created the role. It wasn't like it was an open job rec out there. I got in touch with her and pitched myself, hey, I think you have a data set that's worth speaking about in this way. I think I'm the person to do it. And she was very kind to take a bet on me. That was very helpful.

6:16I owe her a lot. Yeah, that's amazing. So insights at Carta. We have 55 ,000 startups that use the platform, and we've got a ton of different data from those startups on what they're doing, what they're fundraising for, how the founders are granting equity, the dilution, all these cool stats. And it's my job to aggregate that information, anonymize it, and then give it back to the ecosystem. So you can think of me as sort of a public researcher for startups and venture capital, because as we mentioned at the top, there is not a lot of data in these spaces. And so when you lack data, you make worse decisions.

6:50So my goal is to help people make better decisions. But one of the things that's really unique about your approach is that you are, you're almost like a storyteller and you're taking a very content kind of TikTok style approach to surfacing the stories behind this data. So give us a sense of how much data Carta is actually sitting on. So it's 55 ,000 startups, as I mentioned. We think that's at least half, probably more than half of every venture backed company in the US. At last count, we have about 85 % of all unicorns. U.S. unicorns use Carta. So if you're looking for venture data, there's really no place better.

7:26On the fund side, we help fund managers with their fund administration for about 3 ,000 venture funds in the U.S., which is definitely the majority of venture funds. Private equity, which is a space that we're expanding into very rapidly, I'd say it's in the close to 100 funds. There's a ton of different data sets. Maybe the biggest, broadest one, though, is our equity management tool helps more than a million private company employees manage their equity. So when it comes to questions of how much are you going to get in equity? What is a vesting schedule? How do you think about exercise? Those questions, I think, were basically world-class in terms of the data set on that.

8:01So we're going to dig into some examples of this data, but I want to start broad. So as you know, I invest in early-stage AI companies. So I want to know how much money is actually going into AI. And to answer this question, you're going to have to define what you mean by an AI company in the first place. Oh, I was going to put that back on you, though. I want you to define an AI company for me because I feel like I have 80 different definitions that I pick one depending on the morning. That is part of the challenge. Well, tell us what it is today. Okay. And I might push back on you, just saying.

8:35Please. All right. So AI companies, the way that we would define AI in our taxonomy currently is that you are using AI as a core part of your product offering to end users. This can mean two different things. And I think this is a big debate. Are you an AI company if you are simply using AI within your day-to-day business, right? Your marketers use AI to write copy. Your engineers use Cloud Code to write code. Are you an AI company? Most people in that case would say no. We agree. I would maybe debate that, but yeah, I think that's a fairly reasonable one. If AI is in your product, if AI is part of the value that you are offering to your end consumers, whether or not you build the foundational model yourself or you're doing chat GPT calls or whatever, then I would say that you are an AI company and I would include you under that umbrella.

9:27So that's what we've got today. I think that's a fair definition. I mean, there's a whole class of AI native companies that maybe are not building an AI product, but they are like, AI is integrated in everything they do. And that is interesting, but that is quite different also from just, you know, companies that are just using, you know, ChatGPT to like do a little bit of market research. Yeah, 100%. Plus there's like the layers down the stack, right? There's the energy companies that are thinking about powering AI data centers or chips or semiconductor technology and all like that infrastructure layer of the stack.

10:00Those are definitely AI companies because they wouldn't exist without this different technological revolution, even if they don't, you know, they don't have AI powered chatbots in their product. Yeah, absolutely. Okay, so yes, how much money is going into the AI space? All the money, every dollar. All of the money, okay. Every single dollar, it feels like. It's tough for us to say exactly. So this year, for instance, basically of all the capital that's been raised by companies on Carta, I'd say 35 to 40 % of it has gone to an AI company. So that's many, many, many tens of billions of dollars. this number gets way out of whack when you include the top, say, three or four companies, right?

10:43If you include OpenAI and Anthropic into this discussion, we can debate whether or not they're even venture-backed, really, startups anymore. They're just gigantic companies. Then the number gets blown out, right? It might be over 50%. If it's early stage, we're talking about, yeah, you can think of it as every dollar given to an early-stage startup, 40 % of those dollars go to AI these days. Wow. You know, there was a time when a company would be an internet company or a mobile company or a software company, and then it just became irrelevant, right? Yeah. Do you think the same is going to happen in AI?

11:15I really hope so. I'm getting bored with putting out the same chart every time that says this is AI revolution. And you're right. We don't talk about internet or JavaScript companies. We talk about software companies. And then we talk about what kind of software you're building. I would be really shocked if in a couple years, 90 % plus of software companies on card aren't using AI to build their software in some way, shape, or form. Okay, the billion-dollar question and the elephant in the room, are we in an AI bubble? Oh, requires a little more definition than that. So I think bubble can mean a lot of different things.

11:52If we're talking about purely valuations for startup companies, inarguably yes. 100%, yes. That doesn't mean, by the way, that some of these really, really highly valued startups aren't going to be worth it. Some of them will. Most of them will not. That's not very different than normal startup world where most startups don't end up being worth what they say on the sticker price because most startups fail. That's just the way that startups go. So it's actually not that different than usual, but yeah, we're in a bubble in terms of some of these valuations are absolutely wild. If you're talking about a bubble in AI generally across the economy and some massive percentage of the growth of GDP is based on data centers that are fueling AI and all this kind of stuff.

12:35I really like Ben Thompson at Stratechery's take on this, which is probably we're in a bubble, but if useful things come out of that bubble, I think it might be worth it, in particular energy. There's a question as to, can you compare this bubble to the last bubble in say the dot-com boom? In that sense, dot-com, a lot of companies were started, They flopped, but we got fiber out of it. And that fiber now runs our internet. Is there a similar thing where the energy that we're going to get from data centers will make things much better broadly for the economy in 10 to 20 years? I don't know. There's a painful decline coming at some point.

13:12I just don't, I have no idea when it's going to happen. Coming up, who might win and lose in an AI-driven future?

13:25Thank you.

13:55Be free. Go ahead. Listen to Freakonomics Radio wherever you get your podcasts.

14:05Okay, so from an investor's perspective, where do you see the biggest opportunity? And I'll share my bias and kind of my answer to this question. I'm particularly interested in vertical AI, and that's applications of AI specifically in, I don't know, antiquated businesses and industries like healthcare, legal, construction, publishing, where you can bring in an AI solution that can really kind of reimagine these industries. I'm curious what you're seeing. Are you worried at all that some of those vertical use cases will end up just getting eaten by the models over time? That I was going to go to that next, defensibility, because I worry about defensibility not on the spot, but longevity of defensibility.

14:51I actually tell founders when they pitch to us, I'm like, if you wake up worried that the next version of ChatGPT is going to render your product obsolete, then this is not a defensible product. I don't want to be an investor. So I think defensibility is an important consideration, but I want to hear, you know, what are you seeing in your data? I think, I mean, it's definitely maybe, it's perhaps the number one question that VCs are asking founders these days, which is, what is your plan to make this model proof or moat proof? Like, where is your moat coming from? It used to be that moat would be the ability to build a product, the technological skills that you had to do it, et cetera, et cetera, and the speed at which you could do it.

15:33And now it feels like some people are saying speed is the only moat. I think that's kind of silly. Like if you are just, if the only thing that you can do to build a moat is run as fast as possible. 996. Right, exactly. Sleep in the office, you know, get cots for everyone, never see your families. I think that's kind of toxic in some ways. Although, to be honest, I've done it at young startups and it's great for a little while. You're all just mission driven and you feel so enthusiastic about it. But, you know, someone with kids, it's a little bit harder, that kind of thing. It's not sustainable, yeah.

16:05In general, where do I see the opportunities with AI? I think vertical AI is one of the biggest ones that is coming up. Also, the idea of the mix between hardware and software. So in some ways, if software itself is more difficult to identify as a moat, then perhaps the moat comes from building physical things and infusing AI into them. So a lot of cool and interesting business models around robotics or Internet of Things or all sorts of stuff like that. The other thing that I think is really, really clear is that, and this is perhaps not in your investment thesis, there's just going to be a generation of companies that are not venture scale that can work now that don't need VC.

16:48And they're going to build wonderful, smaller businesses with a much leaner team, AI-enabled tech services, et cetera. And they're going to be amazing companies that are just not fundable by venture capitalists because the outcomes are not big enough. And so like a little small explosion of, you know, individual, we used to derogatorily call them lifestyle businesses, but they're more than that. I'm interested to see those pop up. Profitable, right? Like profitable, like good old profitable companies. Who wants to turn a profit? That's so lame these days in venture. But it's actually really great.

17:22Yeah, it is really great. I absolutely agree. All right. So we asked you to bring a few of your favorite charts, and I have a few too. and we're actually referencing a slide deck. The title is How Much Capital Do You Really Need? We will pull these charts up for our viewers and listeners. We'll try to explain them for our listeners who are not seeing them visually, but we will also link to the PDF in the show notes. So let's dig in. My first question to you is around AI valuations. There's this huge spread right now. There are these companies that are raising rounds in the hundreds of millions of dollars There's a crazy, mazy, like billion dollar valuations, sometimes pre-product, pre-revenue.

18:05But then we're also seeing companies, and that's kind of where a lot of my investments are, that are just like good old pre-seed, seed stage valuations as you'd expect. So I would love to get your thoughts on what is happening in terms of valuations and also maybe size of the round. So maybe we can pull up slide 14 to start with. This is a constant question that we get in the VC world, which is what is the right valuation for a small company or a young company? Sometimes they're not quite as small as you'd expect. Look, there's no, I think people want a right answer, but there is no right answer to these questions.

18:49On the one end, let's talk about the biggest, shiniest valuations first. You know, you're raising$50 million at a$500 million valuation as a seed round and everyone is losing their minds. They're like, how could a company possibly be worth that? Well, it isn't really worth that. Valuations don't have that much to do with the underlying value of the business today at seed or series A. They're still very young companies. What it has to do with is who is founding these companies. These people that are getting most, the vast majority of the valuations that are at the upper end are incredibly legible founders.

19:24And legible in this case means they have credibility badges that other people don't have. So for instance, the paradigmatic example is Mira, Mira Marati, who is the CTO of OpenAI. She was CTO at OpenAI and then she founded a company called Thinking Machines and she raised$2 billion as her seed round. The reason she did that is because she was CTO at OpenAI. Like there was no explanation needed. this was someone that VCs were clearly going to fund. And you can debate whether the valuation was too high. She didn't have a product at the time. I wouldn't have done it, but I totally understand why they would because I think this is an incredibly important, viable person that might make me generationally wealthy.

20:07That makes sense to me. I don't know if the valuation's the exact same. But that is not the norm, right? Because a lot of my LPs are like, oh my God, did you see this headline, this headline? But I think that is not the norm. No, it's just the stuff that gets the media attention, right? So I try to show in all sorts of different ways the gap between the part of the market that gets talked about and then the rest of the market that actually is. So for instance, in this chart, you see valuations over time for companies. The AI valuations on a median basis for 2025 at a seed round is 15.6 million.

20:42That's pre-money. So you add in maybe three and a half or$4 million raised. you're talking about a 19 or$20 million company at seed. Now, is that expensive? Yeah, that's pretty expensive historically, but that is not a crazy off the beaten path valuation. You can make that make sense. But those companies don't get media headlines. They don't get tech crunch headlines. They don't get talked about on Twitter as much. They're just kind of floating along as they always have in the middle of the market. And there is a really big middle of the market still in venture. Yeah. Let's switch to that same chart, but for Series A companies.

21:21So that's slide 16. There we go. All right. So what pops out for you in this chart? And maybe, again, explain to us what we're looking at. Totally. So this is a chart that shows a line graph over time of valuations for software companies. And then we split it by whether you have AI or not AI as your software denomination. So again, AI companies get a premium. Those lines are always higher than the non-AI lines. At the moment in 2025, we're seeing series A, median series A for a software company is about$55 million pre-money. That's the valuation. Whereas a non-software, you're talking about 42 or so.

21:57So a pretty sizable gap there between AI and not AI. And the AI companies are also raising more money. So one of the biggest questions that I get from founders is, okay, I get that AI valuations are higher. Does that just mean they're selling less of the company? But it doesn't really. The dilution is the same most of the time. They're just raising more capital, which is odd. At the very beginning, there was this idea that, oh, with AI, you won't need as much VC because building a business is going to be easier. You don't have to pay as many engineers, all this stuff. And yet, many of the AI companies are raising more money than ever.

22:33So how do we square that circle, I think is an interesting question. Yeah, I was actually going to ask you about that because this is, I've not seen this in Carta data, but maybe you have data to support that, that AI companies are more capital efficient, like outside of the foundation model, like infra companies, infrastructure, AI infrastructure companies. And the thesis was, okay, you are getting further with less capital and in smaller teams and in less periods of time. Is that supported by any data that you've seen? The thesis was correct. They do have smaller teams. They could be more capital efficient, but what it discounts is the compute costs.

23:17Most of these companies are not profitable. And they almost never have been at early stage, right? It's not kind of an expectation, but everyone always harps on, oh, the margins on these AI companies are terrible. Well, they're spending a lot of money on the AI itself. And if it's easier to start a company, every one of those spaces gets more crowded more quickly. So you have to spend money on distribution as well. So on the whole, I would say that AI companies are not necessarily more capital efficient than their earlier counterparts. They're just spending the money in different ways. In different ways, interesting.

23:52So instead of like FTEs, it's actually a lot of it is going to the compute and kind of pushing your product out there. For sure. These companies, to be clear though, as a distinction to say the last boom that we had in 2021, they have more revenue. They have a lot more revenue. So in that case, is it more real? You moved from, do they have any revenue to, okay, these companies, a lot of them do have very significant revenues. Are those revenues gonna stick around is the real question. But if you look on it on a basis, these companies are quote unquote better than their 2021 counterparts. They're generating real dollars right now.