Keith Rabois: AI company durability is unproven — talent density and philosophical alignment are the real filters

Speaker 2: the Renoir Museum in the latest French art heist. We'll get to that story after we chat with Keith Raboi, managing director at Coastal Ventures, returning to the show for the fourth, fifth, sixth time. It's always great to chat. Keith, how are you doing? Great. It's great to be back. Have you been following this this French art heist? Is this is this bearish for France?

Speaker 7: I I have not been following it. Okay. I don't know how I I don't know how I missed that. I was just in France last week and somehow Oh, you were in France last week. Are you guilty? Let's figure out. Maybe you know, you never know. Maybe I was stealing purloining some, you know, artworks. Okay. Okay. Well,

Speaker 2: you're you're back in America. If the markets are too frothy, I'd rather swipe some some fine art. Yeah. Are they are are are they too frothy? It feels like there's a lot of value. A lot of the a lot of the companies, they're growing revenue. The products work. They're magical sometimes. Is it too frothy right now?

Speaker 7: Well, Well, I think the revenue traction's real, as you pointed out. And then there's a variety of different categories of companies. Some are producing meaningful margin on that revenue, and others are not. And so I I don't think you can paint with one broad brush and say, hey. All these companies, you know, are worth x or y or not worth x or y. Mhmm. You have to kind of understand the fundamental drivers, but many of these companies have significant traction. I'm not worried about that. And, you know, I don't think multiples are the right way to evaluate private companies typically anyway. But if you work backwards from 2030 and say, which of these companies are gonna be sustained? Which ones have durable advantages? As I like to talk about it or communicate it through accumulating advantages, that's less clear. Like, the traction is off the charts by any historical norms. The ability to go to a 100,000,000 in revenue from, you know, two years or so or hundreds of millions in four years, basically unprecedented. But you still do have to ask questions about what's the sustainability of that revenue, not just because customers may or may not change their mind or there may be new competitors, but like different the way the ecosystem evolves around AI is not at all easy to forecast. So we struggle with this and debate this in our partner meetings constantly. The founders are impressive, the teams are impressive, the revenue and the adoption is impressive, and even the usage often is impressive. But if you still say, Is this going be an iconic company that's around for twenty, thirty years? In many cases, the question is not so the answer is not so obvious. Now, if you play your cards correctly, you may be able to get acquired.

Speaker 1: Yeah. And that's that could be fine. No. And it feels like it feels like the the the powers that be in AI, I'm thinking of one guy, I won't name him, is almost not willing to let any AI companies fail at the at this moment in time. At the moment. He is playing an interesting so that company means money. Yeah. They're growing really fast, and they have lots of profits.

Speaker 7: If you're in those shoes, you know, and your revenue depends upon people buying what you sell that are mostly AI companies, the strategy is incredibly coherent. Now, they are starting to move from being a platform company to a product company. You can see it with the recent acquisition. Yeah. As they start shipping products, maybe they ship products in other areas, they will compete with more people than they ever have before. Right now, any I AI growth is great for NVIDIA. Mhmm. But as they start having their own products, that may change how people perceive them and how they perceive others. And I think to continue and sustain their growth for whatever sort of reasons, they believe they need to have their own products. Yeah. And there's

Speaker 1: one thing that's interesting is there's so many companies that are marked up into the into the tens of billions of dollars that that are actually insignificant to the overall AI trade, but their valuations are such that if one were to fail, you would have all over the Wall Street Journal, Financial Times, all the newspapers, $30,000,000,000 company Yeah. You know. It would be like a WeWork with Theranos level failure, which like it would be a bit for a very different reason but like it would be that big. Yeah. And I think that I think that the market would be able to process it generally. Yeah. But like you said, there's just when when the cash is flowing, there's an incentive to make sure that, hey. Let's buy this company an extra two years, right, even when there's no other investors lining up to back the company. Mhmm.

Speaker 7: Yeah. I mean, it's it's in self interest. The longer the AI wave continues, the better it is for NVIDIA, without an doubt. Mhmm. And the bigger the amplitude, almost surely, is better for NVIDIA unless, you know, told there's a better version of NVIDIA or, you know, somebody obviates their competitive advantages, which could happen. It always happens in technology at some point. You know, we're seeing this play out now that what's the future of Google? Who knows? But that was, you know, even five years ago, even formulating that question kind of felt silly. Mhmm. But now it feels very realistic that, you know, a chat or similar interface may displace all of searches. The advertising revenue will lag, and so the business itself will show signs of life for a while, but people I know, and not all of my friends and colleagues in different fields are tech forward, have substituted huge fractions of what used to be searches with things they do with ChatGeepti.

Speaker 1: Yeah. Yeah. No. The last three I don't think that's for most of us. Yeah. The last you know, obviously, I'm I'm, you know, quite quite biased here, at least, like but I went from last year not making any real, like, commercial decisions with with AI to suddenly everything I buy, all that research is happening within ChatGPT. And I feel like I'm buying better better products and I'm having a better experience. Right? I don't need to I have, like, very specific, you know, I have very specific things. Like, my frustration with Amazon was always, I just wanna buy products that where the company has been around for fifty years. Right? I don't wanna buy the the drop ship thing with a with a name that doesn't make any sense that that's gonna fail within the first year or whatever. I wanna buy the thing that I'm willing to pay 30% more to buy the thing where the company has been around for fifty years. And so now I can Now

Speaker 7: you can afford it. It's great.

Speaker 2: Yeah. But I'm talking yeah. The example he uses is like paper towels or the paper towel holder. That's very funny.

Speaker 1: Jordan, what are you talking Anyways, I wanted to get right into how you are like, what is your in like, what are you how much are you pricing in a, like, a data center freeze or or or real friction to bringing in? And, like, what is your mental model for what happens then? Because on one hand, existing compute that's in a powered shell becomes potentially a lot more valuable and starts to trade hands as certain companies are able to monetize it, you know, much better than others. But then there's potentially, you know, millions of of GPUs that aren't plugged in that the world is figuring out how do we shuffle these around the world to get them to places that that they can be plugged in and Mhmm. And the the capital flows that happen there and the impacts to the to the labor market. Like, there's so many downstream implications of that. And when you have both sides of the political aisle that are realizing, hey. I can get points for pushing back here, then it's hard to imagine that there isn't, like, very real friction ahead.

Speaker 7: So I think let me simplify sort of the AI exposure first and then come back to that. The most devastating thing to the frothy hype and the traction that we're seeing with AI and the funding of it from VCs would be, if Anthropic and or OpenAI have a blip in their public exits, their IPOs. Do they have to show metrics over the next month to first few quarters that cause people to value either company less than their current belief that they will trade in the trillions? So they could have a miss on revenue traction, a miss on margin, something something like That would be catastrophic, and it would immediately reset virtually all valuations in the AI world. Mhmm. And probably deprive half the companies that are raising money from access to capital. Well, yeah. And then I've been asking I've been asking that happen. I'm not saying that's gonna happen or is likely. And I think if they miss, it would be more on the margin side than on the revenue side Mhmm. Right now. But if they do, everybody is predicating the idea their their valuations and new investments in dollars on the idea that you can build a trillion dollar company within the life cycle of a venture fund. There's two examples currently, and there's several companies that people believe have the potential to get there. If that changes and you cannot sort of predict that more trillion dollars or companies are can be created, then the valuations that people are investing in and the dollars that are being invested do not make that much sense. Let me backtrack one second. Like, basically, in the history of venture, a very successful company was 10 to $50,000,000,000, maybe 100,000,000,000 at the max. That's like a home run. That's like we all, like, dreamed about sort of thing. Yep. It is possible now that at least one order of magnitude, possibly two orders of magnitude have been added to the upside potential of a startup, which does adjust the prudence and irrationality of a VC investing $50,000,000 at $500,000,000 or whatever the case is compared to, you know, when I invested in Airbnb, it was at 1,700,000 pre or post. 1.7. Yeah. So, like, that role may or may not be coming back, but certainly by adding a zero or two zeros to the upside potential,

Speaker 1: you can rationally look yourself in the mirror and, you know, say, okay. Sure. I'll take a shot on goal here. But how do some of these investors and and I'm seeing these rounds, let's call them, like, series b and c stage rounds where a company has a much richer rev revenue multiple than, let's say, the labs, leading labs, and they're growing, like, often, like, not not even necessarily as well, and you know they're not as talent dense. And these all these rounds are getting funded. And I and and that is that's like the that's like the place of of frothiness that I don't understand. I don't understand no sense. I mean But I don't understand how you look at the business and say like, oh yeah, this has a path to a trillion when it's like, okay. Every single person working at this company tried to get a job at like four other companies, couldn't, and then is now here. And that doesn't mean they're not great. And they could very well build a billion dollar company. Yeah. But I just I don't know how you you underwrite these things to that to that trillion dollar mark. That's crazy. No. You're no. It's a very astute observation nor do I, which is hopefully on your list of companies that are getting funded that have those three criteria.

Speaker 7: None of them none of them none of them are buying investments. But, like, if if you told me, like, hey. Company is growing less less rapid clip than the Frontier Labs with less density of talent. At lower scale. At, like, at 100x lower scale. Multiple. Unless a higher multiple, my answer would be no. Yeah. You know? I I have some competitors I'll introduce you to if you want, but, like, you know, and that I'll be in France

Speaker 2: stealing some things in France. I'll

Speaker 7: go find I'll go find a startup in France. Like, there's one or two apparently that are quite good. Yeah. But the fundamental driver though you identified, which is key and a lot of people lose lose sight of, so it's worth double clicking on. The critical density of talent does matter. Yep. And so if a company is in the short term at low scale growing at, you know, x clip, that's not that critical. If they have the right team assembled against the right challenges, that is investable, you know, up to some dollars about. And whether you're, like, multiplying, you know, using Meta's multiple or OpenAI's implied multiple or NVIDIA's multiple doesn't matter in the earlier stages if the team is stellar. But your point is the team is often not stellar. And it doesn't have to mean the same people that would work at Anthropic or OpenAI. But they have to be world class at what they do, at their craft, and you have to be able to assemble a density of them. In some ways, want to find people who are mispriced and that the labs don't necessarily value you as much as they should, and there's an archetype you can scale a company on, But you still have to believe you're scaling with incredible talent, like top one basis point, 10 basis point on some dimensions to really believe that you're gonna create a trillion dollar business from scratch in in less than a decade. Do you think that

Speaker 2: differences in philosophy, whether it's P. Doom or AGI pilledness, can create sub pockets of talent density? We just had Scott Wu on the show, brilliant mathematician, cognition's on a tear. He I don't know if this is just because he's at, like, the application layer, but not a lot of doom, lot of optimism because he's interfacing with the real world. And I I love that energy. And you see see with Ramp two where incredible talent density, but not of the crazy doom scenario, weird, like, you know, that pocket is a different group. Those folks are really talented, but they have different beliefs. And so I'm wondering if there's if there's, like, the hyper talent actually scatters across different pockets of philosophical alignment.

Speaker 7: Absolutely. You know, it drives back from Peter's old expression of, you know, you're building a cult. Yeah. Well, think about it. Cults have tenants. Yeah. And so some of that is a philosophy of life. Yeah. And I think that's right. Like, at Ramp, we're optimistic about the future. Yeah. We use technology to improve society and we're optimistic about, a, our ability historically to do that and our ability in the future to do That's why people come to work at RAMP, and that's what they do every day. And all of them are thinking, the world's gonna be better than it was a decade ago, and it's gonna be better than it was twenty years ago. And we're gonna we're gonna wave our magic wand to make it better if we have to. Yep. Yep.

Speaker 2: Yeah. I'm a firm devotee of the cult of saving time and money. That is for sure. But what what what do you make of the conversation around doom? There's been this back and forth around regulatory capture and the evolution of this. And I've watched some people from the outside, friends who are not in tech but critique tech from a political angle, and, they were upset about the rhetoric leading to potential regulatory capture, but now they are worried about some of the developments in AI, and they're calling for regulation. And so they're sort of begrudgingly giving the labs their due on the calls for regulation. But how have you been processing just this year? Has it updated you at all on how we should think about the probability of the bad outcome and doom and all of that?

Speaker 7: Well, I'm an I'm an opt I think people start off in life with, you know, orientation optimism or doom. Yeah. And, you know, maybe from the time they're, like, seven years old. I'm an optimist. I'm a tech optimist. That's what I do. And if I wasn't a tech optimist, I wouldn't be investing in tech companies. And there's a question of whether these people should be building or working there or forfeiting their equity if they really believe they're doing something that's bad or evil. In my opinion, the world is getting a little tired of some of this because, you know, we had the gloom and doom about jobs. All jobs are going away. There's no human jobs. It could still happen, but all of the evidence as of today is to the contrary. Every single piece of evidence. There's you could look through you can be with Sherlock Holmes, you know, with your magnifying glass. You can't find any evidence that AI is constantly creating job loss. If anything, it's creating jobs. Yeah. So okay. So they're o for one, let's say, as of now. Then they're like, oh my god. The world's gonna end. Okay. Well, hard to falsify that. Jobs, at least you can falsify. You were like, no, that's not happening. False. This one's a little hard to falsify, but then I think as a couple of people push back today online is, well, articulate how the world ends because then at least the pre the precepts or the predicates, we can start trying to falsify. Sure. You know, and then prove or try to disprove that, you know, there's some risk there. And then and I I do think that the reality is the world is always gonna be subject to some risk. People like to compare to the advent of the nuclear bomb. And there's a very strong argument. It's not QED proven, but the nuclear bomb in many ways probably made the world more peaceful. If you went through the arc of history and, like, how many citizens died, you know, as a fraction of the population of the globe in conflict before and after the advent of the nuclear bomb, you have a pretty strong case. It certainly wasn't negative. Yeah. And is it but was that known that it would play out that way, you know, on day one when we dropped, you know, a nuclear bomb or two in Japan? No. Even, you know, there's a scene in Oppenheimer where they don't even really know what it would do to the atmosphere, and they go to consult with Einstein to make sure it's not gonna blow up the world. Mhmm. And he's like doesn't give him a very compelling, like, definite answer? So, you know, like, the world is free of is not risk free. Yeah. And I think, you know, all of us make decisions every single day. We drive a car. We fly in a helicopter, blah blah blah blah, and take that into account. So I don't think a zero defect approach to life really works either for humans. Mhmm. But I think the burden is on the people arguing for regulation or slowing progress down of articulating, well, what goes wrong, why and how, and so that people can probe and push back and try to evaluate and quantify that risk. And I have yet to see a compelling answer to that. That's not just like, oh, I know it because I was working on this and it's really dangerous. Mhmm. How many of these how many of these ideas do you think Wait. Wait. Hold on. Let me layer on one other political point. Even if that was a 100% true, the idea that politicians are gonna make it better is also subject to historical debate. Yes. Like, if you said we have a world class problem on our hands, would you entrust it to the federal government or to a bunch of private citizens?

Speaker 2: Not obvious. Yep. Yeah.

Speaker 1: Do you think that that that basically these ideas around PDUM end up in the s one? Because I would expect them I would expect Anthropic to actually put this into As a risk factor. As a as a real risk factor. People are saying like, oh, you know, the the you know, what? They they gotta stop the employees from posting this stuff. But if but if it's coming top down Yeah. It might be just Then then Upper SEC disclosure.

Speaker 7: It could very well end up in there, and it would be a it it it I used to be a securities litigator, so I can argue a couple kind of fun points on this. You could say, hey. Like, if the real risk is the world's gonna blow up, buying anthropic shares is no more dangerous or risky than buying anything else. That's a good point. So it's not a particularly it's not a risk factor associated with anthropic. Sure. A more nuanced view might be, hey. If the political environment were to appreciate that No. No. No. We can

Speaker 1: you couldn't you make the argument that there's like that that the company could, with a runaway AI, incur such massive liability that it it is a risk to your actual investment? That's

Speaker 7: probably fair and there probably will be language like that, but it'll be so toned down that most people won't even understand it. You know, there's an old Peter Thielian lesson here too, which is if you have risk factors, you want the most possible risk factors so no one can really tell which one's really important. So you might see an anthropic, you know, prospectus that has hundreds of pages of risks so you can't, like, isolate, oh, this is the one I should actually care about. Got it. But the the most acute risk probably would be something like, hey, we scare enough politicians that Yeah. You know, they implement policy that's particularly in you know, not necessarily imprinted, but it could it would, it could, you know, constrain our growth. And that is, you know, a real risk if they can't control their employee blog posts or x posts Yeah. Sooner rather later. Is there You know, I think people are seeing through some of this. Sure. You know, there there has been significant exposure of the fact that many of these anti data center, you know, crusades are being funded by China Mhmm. Intentionally by our adversary. There is, you know, real proof that these data centers certainly do not waste energy you know, wastewater and things like that. It's kinda it's a silly it's almost like a walking IQ test. That doesn't mean that the political cost doesn't have to rebut it. Sure. But it is kind of a walking IQ test for someone if they really believe that data centers are gonna, like, make the world run out of water somehow. Yeah. Yeah. It's it's very ridiculous that the origin of that is, like, just a simple calculation error that was off by a couple orders of magnitude, and it just ran away before the whole Unfortunately, you can use chat GPT now to solve math problems. I know. But

Speaker 2: but it did it it it reveals like a different problem, which is that, like, people don't do the calculations. They don't they don't even need to. They they just go off whatever. And how do you

Speaker 1: Yeah. Tell me about your read on the dynamic between effective altruism and venture capital. Because I think there's a lot of, like, VCs out there that are just sort of have, like, generally pretty, like, down the middle beliefs about the future and maybe they're generally optimistic and then you get, you know, the pitch from a budding young effective altruist. And part of it is like, you know this person is just gonna be extremely generally, at least we've seen points, like, extremely risk on. It does qualify as a pretty good venture investment. But there's I feel like now there's been enough of a pattern of of blow ups that you need to be, like, kind of wary. Right? Where there's, like, extreme upside, you know, potentially cat you know, your downside's capped as a VC, so you're kinda like, if you blow up, like, it's more your problem than my problem. Sure. But but but there's so much upside. And so I I just think that, like, that specific dynamic, there will be, like, a book written about it. Sure. You know, fast forward ten years, but I'm curious your view on it.

Speaker 7: Well, I think venture when it's practiced as a craft at its best is a matchmaking exercise between founders and a team and a vision and your sort of views about the world. So I think that some types of founders with some beliefs about the world, whether you label them, you know, effective altruism or something else, could be a good match for a VC who sits around all day and worries about that stuff. It would be a bad match for someone at Founders Farm, let's say. Mhmm. So I think that, you know, the best relationships between a founder and their investors and a consigliere or a board member are when you have alignment, kind of double entendre, between how you think about the world and what's good for society and how the founder does. I do not for example, I do watch and I'm friendly with many people who work at prediction markets, but I personally don't believe prediction markets are good for society. I believe in the Trump quote, and actually, I had actually used the line before he did, but there it's just another casino for Americans at the moment, and I don't think we need more forms of speculation in The United States for normal people. And the evidence is that, you know, 99% of the people who engage in prediction markets are gonna lose money, and I don't think that's necessarily a positive, you know, thing either. I don't really love sports betting even though I love sports. So but that's why I would probably be a bad match for someone who wants to build poly market even though I like Shane or or Kelsey even though, you know, I hang out with her too. But it's just not what I believe that the future of the world makes the world a better place. Yeah. Yeah. No. That makes And I I would just say that I think there's a lot of investors that have their beliefs about the way the world should be and then are willing to completely discard their values and beliefs to back companies because they're just like, well, the return is just gonna probably

Speaker 1: be so good. Got I got it. No. You know, percentage

Speaker 7: of investors that will discard their beliefs in order to get returns is like it feels like it's like eighty, ninety, 95%. In a bull market I agree I agree with that assessment. It's just why do you do what you do as you know, in a in your professional career. And at least at KB, and this is Sri of Founders Fund too, we wake up and say, is this going to have a positive impact in society? We have, you know, more money than we need. We have more money to deploy and invest than we need. We could be doing lots of things with our time, and we want to apply our energies to solving problems that we think would move the world forward. You know, Peter and Trey have pretty strong filters on this stuff, and Vinod and I do, and my call I think all my colleagues at KB do. But not that's not for everybody. I mean, some people, you know, work at hedge funds. I don't wanna work at a hedge fund either. Mhmm. Yeah.

Speaker 2: Well, thank you so much for coming on the show. This is always great. Great to catch up. It'd been too long. Too long. Yeah. Enjoy that. Enjoy that art. Let's make it. I I missed I missed the fur I missed the first month of you guys when I was like on every day. Yeah. Yeah. But hey. You're always you're always welcome. This came up because I was, we haven't talked to Keith in, like, way too long. It's always been. Hang. So We'll talk to you soon. Have a great rest of your week. Cheers. Goodbye. Let me tell you about Figma. Agents meet the canvas.