# Europe should choose the AI layers it can build competitively Markdown copy of https://realalpha.app/s/europe-should-choose-the-ai-layers-it-can-build/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy. - Speaker: [Alex Karp](https://realalpha.app/builders/alex-karp/) · Alex Karp is CEO of Palantir. · Co-founded Palantir with Peter Thiel and has run it as CEO, selling data and AI software to governments. - Host: [Mathias Döpfner](https://realalpha.app/people/mathias-dopfner/) · CEO, Axel Springer - Show: MD MEETS - Episode: [Palantir CEO Speaks Out on the Dangers of AI](https://realalpha.app/v/xm0rGvvxVOM/), Jul 13, 2026 - Moment: 29:55 to 35:25 of the episode (6 min), playing on https://realalpha.app/s/europe-should-choose-the-ai-layers-it-can-build/; the episode from there: https://realalpha.app/v/xm0rGvvxVOM/?t=1795 ## Quote > “That's not realistic. So you're going to have to figure out which components are actually the ones that could **advance European industry**, European sovereignty.” > — [Alex Karp](https://realalpha.app/builders/alex-karp/) at [34:31](https://realalpha.app/v/xm0rGvvxVOM/?t=2071) · MD MEETS ## Takeaways 1. He separates the stack into chips, data integration, applications and models. 2. Karp says Europe cannot realistically lead every layer. 3. He frames sovereignty around choosing where Europe can build strong components. ## About this segment Alex Karp breaks AI into chips, data integration, applications and models, arguing Europe must choose which components it can make world-leading. He treats technological sovereignty as a technical, not just political, question. From “Palantir CEO Speaks Out on the Dangers of AI”, Jul 13, 2026. [Watch the full episode](https://realalpha.app/v/xm0rGvvxVOM/) ## Transcript Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there. [30:00](https://realalpha.app/v/xm0rGvvxVOM/?t=1800) I think we could turn it around and we could say, this is not an opportunity because I see a lot of AI skepticism in America. I see in America almost more AI skepticism at this moment than in Europe among normal people. If we have an all hands in America, people are raising posters and saying AI is killing our jobs. If we do the same in Germany or in Europe, people are saying, well, we want to learn, we want to understand, we want to take advantage of it. [30:26](https://realalpha.app/v/xm0rGvvxVOM/?t=1826) Do you think that there could be... Well, yeah, but not without acknowledging that there's a crisis for the very simple reason. First of all, the debate is different. Europe is partly not skeptical of AI because you are not, like, for example, take the Pallinger example. We are the only product available that makes AI useful on the battlefield and safe and useful and safe and commercial. It's very hard to discuss AI when what they mean are largely companies being slightly unfair that would not be able to compete in America. [30:59](https://realalpha.app/v/xm0rGvvxVOM/?t=1859) So you're talking about AI light. But the thing is… But let's talk also about the opportunity. I mean, I get it and we agree that Germany has not delivered, Europe has not delivered. What can we do to turn the world? Let me just finish this. nothing's going to work before the chancellor gets up and says, we have a crisis. Because to fix the problem, again, who's going to fix the problem? The problem is you are going to have to bring in a Tobias or a Schindler or a Thiel as in a very significant role. You can't rebuild with people who've never done this before. And then who's going to build under what conditions? [31:40](https://realalpha.app/v/xm0rGvvxVOM/?t=1900) what are the working conditions? How do you bring the whole society along? How much can someone work per day? Tech works because people work are dedicated, like doggedly dedicated, around the clock for years. And by the way, if somebody doesn't belong on the team, you can move them out of the team or out of the company. By the way, the... For tech people, again, I'm wildly, the people I got along with best when I was on boards in Germany were the unions. So I'm not saying, but you don't need worker protection for someone who isn't a software engineer that could work at an LLM. But to get this to work, you have to have someone saying, this is a crisis. I think you're also going to have to say, this is an economic zone. We are going to create an economic zone in Germany. And in that economic zone, [32:30](https://realalpha.app/v/xm0rGvvxVOM/?t=1950) competition is going to look exactly like it looks in America, or does in China, or does in Israel. Why are people in America more afraid of India than in Europe at the moment? Do they know more? Well, first of all, yeah, I mean, really, honestly, yes. And so, like, again... So you think the fear is appropriate? it? Of course. The simple metaphor that is simple but works is the large language models are like uranium. Okay. You can use it for a good cause or for a bad cause? No, no. They have to be processed. And in the processing, they become, first of all, just touching them probably is not a good idea. But so using them, like again... Yeah, but if you use the metaphor of uranium, I think it's very appropriate because you can use it for bad or for good causes. [33:18](https://realalpha.app/v/xm0rGvvxVOM/?t=1998) And here again, people start to have a moral discussion or discuss whether it's good or bad instead of saying, let's make sure that we use it for good causes. I mean, again, there is this comparison to a knife. You can use a knife in order to prepare food and you can use a knife in order to kill people. So we should not basically forbid knives. reason why, but it's both true. And again, this is why I think these things are actually economic and technical first. It is, you have to understand what the technologies do. You have to have people who are working natively with these technologies and those people have to be involved. And again, here it's like, you basically have different parts of the stack. You have the GPU part of the stack. So that's chips. You have essentially in an enterprise, you're going to have a data [34:08](https://realalpha.app/v/xm0rGvvxVOM/?t=2048) integration part of the stack. You can do that by hand. If you want civil liberties, you're going to, we do that. You have an application part of the stat. So ontology, what does it mean? How does it mean? How's it deployed? How is it safely deployed? How do you process, how to use? And you have the large language models, both closed and open models. And the closed and open models have different advantages to the world here. To have a tech scene in Europe, you need world leading components in all those parts. That's not realistic. So you're going to have to figure out which components are actually the ones that could advance European industry, European sovereignty. The sovereignty issue is a completely technical issue. And it's not just Germany, honestly. What's really happening in Europe is you have bureaucrats whose primary idea is, [34:58](https://realalpha.app/v/xm0rGvvxVOM/?t=2098) how do I protect the market against American companies? And everything else is downstream. Let me again try to twist it towards an opportunity. If this kind of anti-AI sentiment is ongoing in America, and I fear that America could see an election campaign where the two parties are basically competing in order to explain that their party is protecting their people better against AI.