Alex Karp on Why Palantir Is WINNING | Zeta Live '26

AI's first wave and America vs. China

00:00Let's talk about Palunteer just for a minute. >> Yeah, >> you guys are obviously crushing it. >> What's really interesting to me is what are going to be the most highstakes questions that complex organizations are going to have to ask about where AI is really going to change the work that gets done. Alex. >> Well, um, the the there's like a there's the first wave of actually the first wave of AI really happened in the military, but then my version of it was partly is classified, but you see it in Ukraine and other places, >> but we're private here. You can talk about classified things. Yeah, no problem. But um but I think actually the real problem in disseminating the truth was you the truth in in America which is the

00:50defining place for AI which is also super interesting because normally you have a revolution and like I spent half my life in Germany you'd expect a German version a French version an American version but really you have America and China and and this is already completely unique um because yeah I mean if you think of other revol olution's philosophical industrial if you look at the history of Asia for example um um when Deng Shoping took over China he he immediately got industrial mines from America Chinese Americans um and then Japan's rise visav v China and but these things were pretty intertwined so Japan rose because it it took over Chinese culture and then it took over western culture of how to do industrial this is

01:38a revolution that is basically At least the original thinking and research is completely done here and then derivative advancement fast follower or China. So that's already super unique. Then it that leads to a lot of confusion because the first wave of AI which was you give us your data we put it in a model. The model kind >> Are we still in the first phase or we're moving to >> we're we're in the I think we're both in the first and second phase. So a lot of people are in a phase of you give me mo I take your model I run it I chillax on paying you a lot of money I get very I get I get value on very easy use cases it doesn't work on complex use cases and I may or may not be in danger of outsourcing my primitives my knowledge

The first and second phases of AI

02:31that the my alpha what makes my business special to a model company that will compete with me effectively and put me out of business Um, and then but it's this unique phase. So like if you just may take a neutral example, why is Europe not scared of AI? Why aren't they adopting it on the battlefield even though they have all this information of it's the thing that works? It's because you can still look at the things that don't work and then say, okay, well this doesn't work. It's not worth it. But then so this let me just get to your question which is this is all backdrop for what are what is actually working and then the what what does it mean to work? The first big difference like it used to be that the delta between a good

Why Europe isn't scared of AI

03:14software implementation and one that was second rate was real but it wasn't determinative for your life. And so when we built Foundry, which is a data integration platform, which we're using with you, that would save you time and money and make you more efficient, it would make you A+ but B minus, which would take longer, was still in a range of if you were better at marketing and you had B minus or you owned your you you had a high moat, like just say you're you're building a certain kind of missile, right? Okay. Yeah, maybe you could build them faster with foundry, but you have a moat because it's hard to build this missile. And also, say you're a French company or a German company, they're German government's going to buy

03:56your missile even if it's B minus because it flies and it's German, right? Okay. So and we do the same thing but um so it okay so what happened what's happening in the second phase of AI was yeah but your missile isn't going on target against the Russians or I if if you do not have a way of putting it there and that way is going to be an application layer with a model and without that your interestingly your B minus missile is a triple plus with this and your A+ missile is useless >> so that's just binary and that this the profoundity of that minimal example is something that we don't have that you know in philosophical systems I I I was going to be an academic and I my form of >> law degree from Stanford you have your

The missile example: software as binary advantage

04:44PhD in philosophy >> your family's got to be somewhat proud of your academic accomplishments >> no my they we can go back to that but it's like and my dad's like you could have been something with that [laughter] literally it's like But uh I'm not joking. It sounds like um uh >> you you rarely have when these massive things happen it it from modernity to postmodernity to or or the or what is it let's just take an American example. What does it mean to internalize the first, second and fourth amendment? The it it the first amendment being our right of free expression which is first of all the constitution was given to us by a higher being. So it's already different than Europe. But then first,

05:32second and fourth amendment, right of first expression, right to defend to first expression which is really what the second amendment is about and the right of privacy. It these are concepts that are so radical it takes generation. So here you have something that's very s like you've now have a thing a technology that will make a rocket to use a very simple example that was third rate first rate and deadly or you have something without it that will take the first rate rocket and it's not useful. So this is true in every single enterprise commercial enterprise and and and and what is it that is special about AI. So just take an example you have unique ways of understanding how a market would react

06:15to a message. >> Yes. >> Okay. You are going to how do you share those insights with a with a partner where you only share the insights that you want to share and they want to share with you and how do you augment them where both the data the insights and the models you run against them are sovereign meaning unique to them or could be white labeled and sold to others. Well, and that's one of the things we love so much about Foundry. Our integration of our data cloud on Foundry gave us far more ability to do those things. And we're now able to onboard clients in hours that used to take months, but give them control. I'm I'm going to pivot us back and bring us a little lower to the ground so most of

Data sovereignty and protecting your alpha

06:57us here can understand the vast majority of what you say. We talk about data sovereignty a lot. We have a lot of CEOs in the room, a lot of CMOs in the room. Alex, what does that really mean? >> It means that well, first of all, and this is an very advanced thing, and we're going to be able to augment this and then put models on it and then train the models so that the models reflect just what you know and do not expose it to the outside. So basically, okay, every business, what is a business? A business a business meaning something that at some point in the future will provide high quality revenue meaning free cash flow. Okay. And what is why is free cash flow what everyone thinks about analysts? Um why is this an

07:42important concept? Because it basically tells you you have something unique in your business that is valuable enough where you have pricing parity so that you can charge more than the cost of the goods. Okay, that is a business. Every business in the end of the day comes down to that. Why can you charge more than it cost you produce it? Because you have some unique way of doing something, whether that's how you build it, how you produce it, how you distribute it, and how you market it. Okay? And preferably, it's unique along the whole stack. >> And the more unique it is along the stack, the better your financials at some point in the future will be. I mean, there's always a lapse of time, but I basically

08:23I do I think the free market will get you to pricing parody over a 18-month period, >> right? Or so uh and okay. So in every one of those things, the way you produce it, so supply chain, how you how you how you how you integrate it, what the final product is, what the aesthetics of the product and how you explain it to the market. You need to advance your unique capability, learn from the world, and never ever expose to anyone else how you're doing it exactly. And the minute you ex someone else can do any part of those things as well as you can your pricing parody goes down and at some point your value goes down. I mean these are like a lot of times is okay. So normal software is built to to force you

Why VCs didn't get Palantir's non-parasitic model

09:14to adopt something that's parasitic. So you they take every one of those things turn it into a commodity and sell it to everyone else. That's how you We couldn't get financing because I kept telling these VCs, "Yeah, but we're going to make we're going to make our partners and clients happy." They're like, "Why would I?" >> Wait, what? Happy clients? >> No. Wait. When you want to start a a company and you talk to VCs, what they wanted to hear was clients will be unhappy, but they can't get rid of you. Like they want to hear bad marriage, >> aka sticky, right? >> Yeah. No, it has to be sticky is a metaphor for parasetic. >> Correct. >> I get your data. I get it into something you can't get rid of. The thing turns

09:52into something I can sell to everybody else who then can't get rid of it. And de facto, your multiple goes down and my multiple goes up and all of us are happy and hope at some point you realize you're sad. Hopefully after it's too sticky and okay, so that's how you get we couldn't get finance because I kept telling people no, we're going to make we're going to make spec special operators more lethal and we're going to bring them home and they're like but where's the parasite? I'm like, "No, we're not doing the parasitic thing." And uh and then if you look at the stack of what we're doing, Foundry will bring all your data together and protect it granularly. So you can share the data and not share what you don't.

10:31>> Well, you get to decide what you share and what you don't share with Foundry. >> Yeah. And then the the the application layer on top of it will allow you to do the same thing with openweight models or closed weight models. I mean, close weight models, you need to get them to post train, but either way, you are not going to be seeping your alpha to everyone else. And so, at every stack, every part of the stack, you augment what's particular about your business without losing that particularity by giving it to somebody else who's anti-polled with that. And that that for those of you who are like VCs and others who >> bismerch us, whatever. Any case, I get along with them now that I'm standing on

11:12their head. But um it's uh who are basically they just want to finance. They've pass someone comes to pound you for two days, they'll give you like $10 million, start a business like planning plots. But um uh but any case, the the the radical thing here is the obvious nature of this, >> right? >> And and by the way, just to why did we do this? We did it because we knew we could not win on popularity. So that >> if you can't win on popularity, you better win on alignment. >> And and what I was literally just going to take exactly what you're saying to the next level. One of the things I love so much about the elegance. Can >> you come over to my family? I I'm not used to this love. >> I'm sorry.

Forward deployed engineers and talent

11:51>> I'm not used to When did I get all this love? >> I can I can give you later. I promise. >> Yeah, please. Please do. I'm not I won't believe you if it's positive. >> Exactly. the this really sits inside of your entire concept of forward deployed engineers from the way I see it like we have your FTEES in our office every day working with our FTEs. Well, there's this concept. I mean, another reason we couldn't get financed and every single person made fun of us is we were like, okay, to extend the unique nature of the enterprise into our product and into the enterprise itself. We have to send in people who are highly technical and understand our product and then understand your enterprise so that we

12:33can augment our product specifically for you. But that's so different for a software company. Like we pivoted a huge part of our business to follow you guys on this because it's so freaking efficient. >> Yeah. It it Well, it also gets you again it allows you to basically bespoke build what we have to exactly who you are. And and then over time we're training your people. Like we're not a body shop. So we want to train your people to be your own FDS. And we do that. But we have 4,500 people. We're not going to have 50,000 people. So, we we need this is where we need we need partners because we have to work through you because we have we're an artist shop. We we manage a certain kind of person very well. I

13:20that's what I do for a living. >> It's not just a certain kind of person. You manage an extraordinary group of people. I I mean, the more time I have spent with people like Elias and and and all of the other people on your team, the more impressive it is the people who are there. >> Yeah. We we had long dark period by most standards 17 years uh and uh and in the desert of everyone making fun of us every day. Um and and especially people who are the most knowledgeable. But um the one thing everyone would always say is yes, but they have the best people. And it's still true. In fact, it's even more true now. We've had our our recruiting for whatever reason. We were just getting >> Well, I I think your success is probably

14:08helping. >> Um, you know, it this obviously success always helps, but it's you we had it's interesting. It's it really is that it's more and more clear to engineers why you would want to power a sovereign stack, why that's good for clients, why it's good for society, why it's what is it's the thing that will transform America and hopefully the rest of the world. And so at like every part of our stack, it has become much more charismatic. And so we've had we've had I mean we've always been working at the way way and high end of talent, but even for like we have different we have different vintages. It's like your wine like we vintage like vintage 2008 is different than vintage 2012. But >> eight is much better in burgundy but not

14:54as good in California. >> Yeah. No. And it's very similar interestingly because for we have different roles and like but right now we are crushing it on and by the way anyone here has a precocious person they you want to like recruiting we manage uniquely shaped I'm dyslexic which I I next you're you're stealing my next topic. >> So well any case >> because I am too. Well, I I was it's the only thing I was ever in the closet about and I was super and then any case but we manage highly uneven essentially brilliant people and then we slot them and grow them and Palanteer is not everyone's cup of tea but for some people and ex-punterians who are missing the Palunteer vibe uh they'll tell you. In fact, the best references for us are

15:45tend to be people who stormed out the door. They're like, "Another crazy lecture from Karp stormed out the door and now they're unhappy with the shagging of their current business and they want to come back, but it's hard to come. I mean, it's emotionally hard to go back, but they're our best reference. We're we're not for everybody, but for a certain kind of person, we are the only place." And we I spend a lot of time actually my favorite times besides of course hanging out with you. Oh, there we go. >> Hanging out on the factory floor with they all you know these >> By the way, you live that like your office is in the middle of the factory floor. >> Yeah. No, I I but I It's what I enjoy. It's like, you know, it's funny. One of

Dyslexia, decisions and the Athena story

16:24the great luxuries I have is I mostly get to do not always, but I really do get to do mostly what I enjoy. And what the thing I like most about my thing is just hanging out with my fellow palunteerians and working on things and fixing things and being frustrated and you know and then and they say the cra you know I just I really enjoy it. >> I always look back on our first conversation. So we're both very dyslexic. We both from come from similar backgrounds. We both, I think, have a big chip on our shoulder from not getting financing and not being seen. At least for me, it was as a kid, not being seen. And then we couldn't raise money for our business either. Everybody thought we were nuts >> until a few big part platforms like, you

17:11know, Bob House. >> And actually, you're a funny guy. I think in private comedy gets me. >> It's I think I'm not the least unfunny guy either. >> I think you're very funny. >> So, it it goes like in private. You're funny here too, but you guys have no idea what we're dealing with here. This [laughter] is like in private is a different level. And you're laughing proformer some of these jokes and I am too, but in private you really laugh. By the way, put put that on my tombstone. He said I'm funnier in public. In private. No, but in all seriousness, we had our first conversation. We were supposed to be together 10 minutes. 40 minutes later, they're pulling us out of the room. We start with philosophy. We move on to

17:52wine and watches. And then you looked at me and said, "So why are we doing this?" And I said, "Alex, because you've spent 20 years solving the world's biggest problems, and we can help our clients to solve one of their biggest problems that you don't have a current solution set for." And you just looked up, you're like, "Oh, great. Let's do it." >> It's true. Yeah. >> Is that how you make all decisions? Like I mean, >> um, well, I would say if I I I asked I I sometimes ask people like because I it's very hard to know. It's very hard to perceive yourself. So I ask people I work with like, well, what is it like the people, you know, I've worked with them a long time and we're we have pretty we have a very honest

18:32relationship. So it's not like they're telling me things I want to hear. In fact, I often tell them, could you tell me something I want to hear, >> please? [laughter] >> By the way, it sounds like my conversations with my kids, >> like please something. Uh but um uh um and I I a very precocious person and he's like look we prepare for hours and then you kind of have a bias pretty quickly minutes and I would say I have a kind of I probably think they have this for my mom but maybe in a different way from my dad. I have an aesthetic idea of what could work and what won't work and and then I just that's what I do all day. It's like, okay, well, this could work, that couldn't. This everyone does this,

19:17it won't work. No one does this, it could. And then about once a month, I have some idea that I think is important. And then I know it's important when everyone tells me we could never do that. >> By the way, >> like DPO like DPO or or sovereign or AIP or a lot of other things. Foundry. Foundry famously Ted, who's a wonderful person who runs US Commercial, famously said, "Please stop talking to your Palunteerians about how big Foundry is going to be. It's depressing and everybody knows you're out of touch." [laughter] >> Well, I would tell you not. >> And he was right. I had to stop talking to people for like six months cuz But it worked. >> I know. I know. I I when when I first sort of came up with the idea for

19:59Athena, normally I go into Steve Gerber's office, who's my partner, and he looks at me and he says, "All right, David, that's great. Let's keep that ju just between us. We're not going to go talk to other people about that because you're going to freak everybody out." >> And I brought him Athena. I had been I had decided to take >> my three older kids. We went over to Africa for the weekend because I'm an idiot. And we did like a safari and I was trying to squeeze everything in and nobody slept for days and I came back and I go into a very manic state >> which I did a lot more when I was younger than I do now that I'm old >> but >> and I sort of came out with Athena and I went in and Steve was like you know

20:37that's actually a really good idea. It was the first one I'd gotten in many many years but let's once again I like you're so smart and you're so sort of open. I I want though to go a little myopic for just one second. If you're talking to the CEOs in this room, of which there are many, other than signing up with Palunteer today, which is we believe the absolute best decision they could make today, what's the one thing that every CEO in this room, what's the one decision they should make today before they go home? Um, in your mind's eye, what is the thing in your business that makes it special that is not quite understood? And how do you get complete vision from the factory floor, prefactory floor

Advice for CEOs: speed and pentesting your stack

21:33distribution all the way up to cost of goods in front of you immediately. and and every [clears throat] single person who's crucial at each one of those points it that like in total granularity where you impose this thing up and down the stack and make every single shirt and there's a tech version that you would need to do this but that's the business thing and then then how do you get AI to enhance that without it a model company or your competitor understanding exactly what that is and that's what I It's what I focus on. But but it the thing is the thing about the shortest version of AI is it just makes it it makes the real things 10 times more real. So you have to get them right. There's no room for up

22:23the way there was. And this is an advantage for America because corporate leaders in America are on the front line and there's no chillaxing vibe like you know basically >> this is not a chillax week. No, this is so like um I I I had a of a very a really talented um somebody I admire a lot, a European CEO, and he he's like he was saying, and I'm going to provide references for him, but he's like, "Look, no one will hire me in America." The head hunter said, "There's no point." And that's a problem, but it it it really is because we are at because more like it's like a war envir like so you you have to get that right. And if your person down the street gets it right, even if they're inferior to you,

23:08just like the missile, they will they will they will outperform you on not just on margin. It's not just it's not this like the classic of private equities. You strip out the costs. This is enhanced product and strip out the costs and that's a rare thing. And also the timeline is different. So but the way to begin >> everything's speeding up, right? the yeah the nature of time is it's one of the things I feel like I I feel very proud of the things I got right the one of the things I didn't understand is and I it's just time is excel like time isn't time in this country in business is like just it's like 5x what it was 5 years ago you it's everything is and and you can make these things the other

23:52thing I would say which is a derivative of it the tactics of this are shorten the timeline of execution Anything that's taking too long, it's because there's something in the stack that's parasitic. We say time is the enemy of everything at this point. It just you have to move faster >> and and and pentest. It's a very important way to pentest because there are complicated things that take time obviously. So maybe there's some important regulatory thing. But if you if you just say if you make a list of on that stack of things, what is taking too long? you'll pretty much articulate wrong person, wrong part of the stack, wrong technology with a high degree of re reliability. And why is this important? Because most people running

24:37companies are not tech people or don't inherently trust their tech intuition. So, but this will give you a way to pentest your actual intuition, which is business intuition against a tech problem because if you can't understand what the tech is doing and it's taking too long, there's definitely something wrong. Ladies and gentlemen, Dr. Alex Karp. >> That was the keynote. That was um [clears throat] part of their first hour. It didn't get live streamed, so it got posted later, but um I don't know. I I think I'd have to watch it a few more times to understand where Zeta fits into that. But obviously still pretty exciting for uh a Palunteer investor.