Klarna CEO：SaaS 已死：为什么记录系统将在智能体时代消亡

We've gone from 7,000 people, we're now below 3,000. We've shrank 50%. And I didn't ask for a single dime to do all this. And the reason for that is because I've seen the acceleration of AI and I know we can ship all these things on the existing organization. It's 2030. How many employees do you have then? 2,000?

No. It may very well be even less than that.

No, but listen now. We have an incredible episode today. Seb from CL is probably one of the leading figures in how to implement and use AI effectively to shrink headcount and make your business way more efficient. This was one of the most wide-ranging conversations we've had. This is what I signed up for. It is stressful. It was hard as hell, but this is what I wanted. The next thing that's going to hit everyone bad is the switching cost of data because ready to go.

Sebastian, it is so good to have you in the studio, dude. We've done this before, but to have you here in person is fantastic. Thank you for joining me.

I am so happy to be here. This is going to be a lot of fun.

Dude, this is going to be great. This is also going to be the best show you've ever done. You've done a lot of shows, I'm telling you already. But I'm just freewheeling. I had these brilliant notes, dude. Where the fuck is value in a world of Anthropic and Claude Code wiping billions of dollars off the stock market? How should I think about that?

You should think that the cost of creating software is going down to zero. That's it. That means that everyone will be able to generate software at any point in time. It is a massive change, and I was 100% convinced about this already when I saw it 1 or 2 years ago. That's been very, very clear to me.

If the cost of software creation is going down, how do we determine which businesses have sustaining value versus which do not?

The key thing right now is that, so far, the only thing that's gone down to—or not to zero yet, but become extremely much cheaper—is the generation of software. The next thing that's going to hit everyone badly is the switching cost of data, because so far what you're seeing is that you have proprietary data stuck in, for example, the CRM vendor or the other software-as-a-service tools that you're currently using.

You may replicate and build the same dashboard or build the same processes in your own tool, but all your data is in there according to their data model and according to their setup. What's going to happen is that people are going to start solving that problem: How do I get all of my data from the existing vendor and move it to the new vendor with the help of AI through one click? That brings down the switching cost, and that's when the real threat to SaaS comes.

We had Anish from Andreessen Horowitz, one of their GPs, on the show, and he said agents in particular will dramatically reduce the friction of switching.

Yes.

Is that the method you're talking about that will allow for this migration to happen?

Exactly. That's exactly what's going to happen. It's happening already.

If that is the case, shouldn't ERPs, ServiceNow, and Salesforce be dramatically threatened?

I think the stock market woke up to that in the last few weeks, right? The question is just—it's not like any business is going to disappear overnight, because people tend to stick with things they've used for a long period of time. They like them, et cetera. The question is, what multiples should they trade at?

If you look at historical software, it could trade at a price-to-sales multiple. I'm not going to talk about price-to-earnings, because some of them aren't profitable, so it's not an easy way to compare. But if you do price-to-sales, they've been trading at 20 or 30, and now they're down at 5 to 10.

If you look at utilities—normal companies that are more utility-like—they may trade at 1 to 2. From that perspective, you would argue there is still, unfortunately, potential to come down even further.

Look at Chegg in the US. They're now trading at 0.2. ChatGPT was seen as basically wiping out their business a few years ago, and now they're trading at a depressed value. The revenue is also coming down, actually, 30% to 40% the last time I checked. Is that going to happen? I don't think so. That's probably too extreme. 0.2 would be very extreme for some of these companies. But is it likely they could come down to 1 or 2? Yes, I think so.

My question to you is that there's this kind of consensus from all investors, which is always a worrying thing: If you think that we're really going to vibe-code a lot of these tools internally, you've never worked in a big organization. The permissions and the hierarchy that ensue with the implementation of these tools mean we are not going to see large companies vibe-code mission-critical systems, and they will keep the largest systems of record. How do you think about that when David Sacks says that?

I understand that some people are of that opinion. I am not, because I think that currently, the way AI is set up—and I've already started seeing people doing this differently—AI is allowing us to reinvent the wheel all the time, right?

If you come in and say, "I want to write this piece of code," somebody else prompted the same AI with the exact same thing somewhere else. We're still using tons of server power to generate the exact same code. What people are going to start realizing is, why don't I cache these things? If I'm getting the same question, why don't I use existing open-source components and reuse software?

If software becomes more like Lego pieces that you put together, that's perfect. It's going to be more and more efficient to bundle things together. This also means that things like what you're talking about—production-ready, security-assured, and all these things—will become more standardized building blocks.

I think in the future, I'm not sure AI will even code that much. It's just going to pick some pieces together and stitch them together to come to what you really need, which also actually means less need for compute.

The other argument—and I totally hear that, but the other argument—is that enterprise software spends about 8% to 12% of company budgets, and you look at that and go, "Well, hang on a minute. Our core business is X—"

Mhm.

Why the fuck are we building a Monday replica, or a you-name-it replica? That's not our core business. Why spend the internal resources on it? How do you justify that?

I think that's, to some degree, correct. But, funny enough, the same weekend that this whole Claude Code thing exploded on X, I was actually sitting myself and playing around with a project that I was just calling "Company in a Box," right?

The idea was just—I wanted to test it a little bit. The idea was to do something very similar to what Claude did, but more for a company, like a small company. I put a small workspace in there with accounting, and in that I put open-source accounting software. Then I put a CRM, and I put an open-source CRM. Then I put a Claude agent on top of that, and I told my Claude agent, "Hey, can you bookkeep this invoice for me?" or "Hey, can you set up this customer account for me on top of that software?" It worked really, really nicely.

I just wanted to test the idea, because that's actually where I see the risks of even more jobs being threatened. To some degree, if I'm a small company today, I may have an accounting firm that's helping with accounting. Those are the people I would email: "Hey, can you fix this invoice?" or "How much money do I have in cash? What's the current P&L look like?"

But now I have Claude as an accountant on top of the open-source accounting software, and I'm just asking, "Hey, bookkeep this invoice," or "Check my balance," and it works really, really well.

I'm not saying—I don't think the plumbing firm or the electrician of the future will vibe-code this themselves, definitely not. They will buy off-the-shelf products for this. But the question is, most of our ERP systems that we see today, or software-as-a-service tools, because coding was so difficult and hard, are still fairly siloed, right? They are not broad in the spectrum of what they cover.

The kind of winner of the future is much more likely to be extremely broad, kind of coming with a Claude bot for companies like those services. That's how I think the future of that kind of thing is.

It's different for a company like Klarna, because in our case, to some degree, this is the operating system of the company. What we realized when we looked at SaaS and all these things 2 years ago, which is why we started closing down SaaS for us, was that we need to provide our AI with the best context.

We need to provide as good a context as possible to be able to perform a job. If your data is separated into these silos—a little bit in this SaaS, a little bit in that SaaS, a little bit here, here's all the project-management stuff, here's all the product definitions, here is the accounting stuff, here is this, here's that—it's just harder to provide it with the appropriate context, right?

To us, it was like, no, we need to reimagine the tech stack with AI first, being AI-native, and incorporate AI, deterministic code, and probabilistic code into one tech stack that becomes the operating system of the bank. I think that's the future of larger enterprise, and that's why, to us, we're very mindful.

We do still use some SaaS, for sure. We use Slack, as an example, today, which is a Salesforce company, right? So that—

You should use Dashworks. It's one of our competitors.

Yeah.

Happy to try it.

We incubated it.

Happy to hear that. So, you see what I mean? For a large company, obviously not everyone needs to reinvent everything. I don't think the plumbing firm will reinvent everything. I don't think they're going to vibe-code it. That's not the point. I think they will buy something that looks like Claude or a company-in-a-box kind of thing.

Totally get you. It's the idea of the compound startup and the benefits that come from not having to integrate with 50 different providers. So, totally get that. Don't agents just make that easy, though? Don't agents just make the data migration between different tools from third-party providers way easier? And actually, won't we be looking at this going, "It was a ridiculous idea to ever think that we needed to own every part of every element"?

Yep. That's exactly right. That's what's going to happen.

So, if that's the case, why do you need to vibe-code it all yourself and build it all yourself if you're going to have agents that are able to move data between different products much more easily?

It depends on what kind of company you are. As I said, if you are a plumbing company, maybe Claude—Anthropic—will offer a solution for all of this. Or maybe there will be somebody else who uses Claude and empowers this kind of company-in-a-box experience, which I think is still the unknown answer. We don't know what's going to happen.

Customer support is one I just cannot get as a category, because there have been 14 players funded with over $100 million in the last 15 months. Then you have all the existing incumbents, and I speak to Ariel at Navan and Jack at Airwallex, and they're building their own. Are you building your own customer support? You are?

I mean, we were one of the early ones, right? This is actually one of the things we were surprised by. I think I announced already in 2023 that our AI customer service had done the equivalent of 600 agents' jobs, and it caught a lot of attention at the time. The media always tends to simplify these stories a little bit. The truth was that, at that point in time, our customer service was handling very simple questions: "Hey, did I pay Klarna?" "Yes, you did." "Okay, thank you." Obviously, that wasn't that hard to do.

But to some degree, as a large company like ours, what do we do? We try to improve our product, partially. We try to do that so fewer people contact us and ask, "Hey, it isn't working," or, "I'm not sure I understand what I'm supposed to do," right? That's part of it. You just want your product to be so good that people don't feel they have to do that. So, we've always tried to reduce customer service calls.

The only, to me, shocking experience back then was that we rolled this thing out, and we rolled a lot of product improvements out. I've never rolled a product improvement out that instantaneously took away the equivalent of 600 agents' worth of work. These people, because we don't hire them ourselves—they work for customer service companies—just shifted and started working on something else. Fortunately, in that situation, nobody lost their job, but it was still an eye-opener for us. We thought, "Wow."

The point is, what you realize when you're early on that journey is, again, for customer service agents—whether it's AI or humans, for that matter—to be able to answer questions really well, they need as much context as possible. Where is that context? It's in the source code of your software. How does Klarna calculate interest? We can have documentation of that, but the truth is in our source code. It's somewhere deep in our source code where that interest calculation is actually explained, right?

So, even if the documentation may be inaccurate, what you realize when you pursue this is that customer service isn't just, "Hey, I need an agent that answers questions." Sooner or later, you want it to read the source code and explain to the customer how it works. You want it to provide as much context as possible to be able to give the right answers. That's when you start realizing that it's not something you can buy off the shelf. In our case, at least, we came to the conclusion that we cannot buy it off the shelf because it actually becomes part of our tech stack.

Will every large technology-first company build its own customer support system?

I'm not sure. I think, obviously, the right thing for Klarna was to be early, to try to find what we can do with this technology and where it can bring us. I think it's going to be a competitive advantage over time compared with incumbents that haven't done that. But a lot of incumbents will obviously procure fantastic AI customer service solutions in order to try to reduce the gap between what we're doing and what they're doing. Who knows? We'll see what happens. But in our case, it was very evident that we needed to do this ourselves.

When you said this, it was a brilliant headline. It was like, "Seb from Klarna replacing 700 people." I can't remember—I think it was 600 or 700. And my question to you was: Do you think that did more to harm or to help you? Because, for me as a marketer, I actually thought it helped you, because it put you in an AI-first CEO camp that very few public-company CEOs are in.

Yes. It's a good, very valid question. One of the things we also realized is that, obviously, as a Polish person, I think when things are about to change, I look at them very cynically, right? I'm just like, "Okay, this is what's happening." I'm not that kind of person who's going to gloom things over. I'm like, "Okay, this is happening. It's going to be a big change to the world. How do I adapt? What do we do to make the best out of it?" I'm sure there's going to come a lot of positive things and a lot of negative things.

In this case, when we announced this, we obviously had some people being very frustrated with us: "Oh, you're laying people off because of AI." People were angry to some degree as well.

And I respect that. I understand why, right? That's why, a few months later, we also tried to go out and tell a different story. You walked it back.

Yeah, we don't think so. I think Bloomberg changed the headline, and then that got misinterpreted. What we were trying to say a little bit later on is that, to be honest, if AI can do customer service, it's going to be the cheap customer service. It's going to be the one that everyone gets because it's cheap and simple.

But as has always happened historically, when people started in factories making cheap clothing or cheap furniture, we started appreciating artisan things. We started appreciating an artisan coffee shop and artisan furniture that was done by artisans.

So, we said the future of the VIP experience will be the human connection and the relationship. We genuinely believe that. We said we need to transform our customer service from thinking about it as, "Okay, yes, it's obviously just good customer service." But to some degree, when I challenged it internally, I said, "Look, what I've seen has happened is there's been too much focus on cost, right? There's been too much focus on that."

We have to rethink this and make customer service into this human part of what Klarna is, and make sure that we offer everyone who wants a human connection what VIP service looks like in the future. "Oh, I'm not dealing only with machines. I'm dealing with a human." That's what we think, at least. So, that's the message we tried to get out.

I completely get you, Seb, but I'm sorry, dude, for being so blunt. That sounds a bit Silicon Valley idealistic. What I mean by that is most people are not actually as good as we think they will be. Often, customer support is a role where you're in there for a year or 2 and then you move on.

What you need is, Seb, I know that you have X number of kids, and you often like to travel here. I thought about these 3 restaurants for your romantic Valentine's Day trip with your wife. That's a really thoughtful agent or person.

The trouble is, most people are not that, and it takes training and development to get there. The low-skilled labor—all social media marketing and low-level content creation—is going to be eroded faster than ever before.

No, but you're right, and that's exactly why we started changing. One thing that we've done that has worked fantastically well, and we're rolling this out, is that we started saying, "Okay, to your point, if we look at the agents that we were hiring previously—even through other companies—we didn't really have a relationship with them. We don't know who these people were." Some of them were great, but it was, to your point, a mix, right? We said, "How do we create something very different?"

What we started since then, which has worked amazingly well, and we're just starting to ramp it up, is that we actually built our own Uber model. Today, we recruit our own customers—the most passionate customers who live in rural areas and so forth—and we say, "Hey, do you want to work extra, or do you want to work part-time in our customer service?"

Just like somebody can go and drive an Uber for a while, they can actually jump on and work for Klarna's customer service. These are our most passionate customers. They love our product. They love how it works.

They know Klarna in and out. And now they earn extra money by actually working on our customer service. Obviously, the NPS and customer satisfaction of those interactions with our customers are through the roof. So, to your point, we needed to change, and that was what we were trying to do.

But it’s very hard. In the Bloomberg article, I’m not going to blame the journalist. I think, actually, if you read the original article, it’s fairly balanced and kind of tries to describe this, but then the headline is like, “They’re rolling AI back,” and then the whole media circus goes on, like, “Oh, Klarna has just announced that they’re rolling it back.” It’s like, no, that’s not at all the truth.

The truth is, no one reads the article anymore. They just read the headline: “CEO predicts end of finance.”

I was just laughing with them because yesterday on X, somebody wrote, “Klarna—obviously Klarna—the company is going to offer buy now, pay later on rent.” It was actually another company, but people didn’t see that. They just said, “Klarna is going to offer buy now, pay later on rent,” and then suddenly we’re all over the press. People are calling us, and it’s like, what are we going to do? It’s just not true.

I remember when I was called in an article “former teenager.”

Yeah.

I was like, I think this is a broad spectrum of people right now: former teenager. Me and Warren Buffett both have that in common, but I’ll take it. I sat down with the team the other day. Our job is to find value and invest in amazing companies. I was like, if they sell per seat, we’re not doing it. We need to replace jobs. Labor displacement is what we’re investing into, and that’s the only way that we can actually return the amount of money that we need to now to make funds worth it.

Yeah.

Do you think that’s fair?

Well, I think, unfortunately, I think that is going to happen. I don’t think it’s a question of if, and I feel like Dario is maybe one of the few who’s actually willing to speak up and say that officially. I feel a lot of the other executives of these big tech companies are getting nervous. They don’t want to—they’re seeing the negative backlash of talking about this, and then they’re trying to portray something different.

But I don’t want to be one of them. I’m more in Dario’s camp. I want to be honest about the fact that I do think there’s going to be a very big shift. In addition to that, I think, more like Elon, that it might lead to a golden age of humanity, where AI does more of the jobs and more people can enjoy themselves and do other things, and we can have a richer society. That is not an unthinkable outcome. It could be a positive outcome. It could happen, and I think it’s not unlikely that it could happen.

So I’m still an optimist at heart, but I also want to be realistic about what’s going to happen in the shorter term, and it’s going to be a lot of turmoil in this.

Can I ask, when you said the statement about replacing 700 customer service agents at Klarna, what did you not know then that you wish you had known, knowing all that you know now?

Nothing.

You haven’t had anything that you’ve seen that has changed your mind?

You know what, Harry? In 2015, I sat down with my management team. At that point in time, we had been trying to compete with Stripe and Adyen for 5 years, and we were losing it. They were just crushing us, and we were like, this is it. Adyen had just signed with Daniel Ek, my neighbor, and I was like, this is over, man. We’re not going to win in these checkout wars, these payments wars. Forget it.

So we were like, what are we going to do as a company? What are we going to do? And then we sat down as a team, and we were like, what’s the future of banking?

The future of banking is going to be some kind of digital financial assistant that wakes you up in the morning and says, “I checked your mortgage. You’re overpaying like hell, and I have renegotiated it for you. I can do all the paperwork. You just need to say yes to save 50 quid.” That’s the future of retail banking.

And we said in 2015, “Okay, so what does that mean for Klarna?” We were like, “Hey, let’s become your digital financial assistant. Let’s be that assistant that saves you time and money.” Ever since then, it’s been, where are we going?

Obviously, I didn’t predict ChatGPT. I didn’t predict all this stuff happening, but to me, the fact that we were going to become a digital financial assistant, and that we needed the kind of technology that AI is to accomplish that, was crystal clear for me for the last 10 years. I’ve just been running down that path continuously, and to me, it’s like self-driving cars. We all know it’s going to happen.

At first, it was a huge hype, and every day we were reading the paper: “Oh my God, it’s happening tomorrow.” Tomorrow, everyone’s investing, and it’s crazy. Then the hype dies away. But I was always saying, when people asked me, “What do you think about this stuff?” I was like, “Look, it’s going to happen. My daughter is not going to get a driver’s license.”

Do you think you’re best positioned to do that? If you look at the entry point and where you sit in the stack, is Revolut not in a more strategic, better position than you to be that digital financial assistant?

I mean, I love Revolut. I think Nick is a fantastic guy. I think it’s an amazing company.

Nick’s the greatest CEO I’ve ever interviewed. I’m also terrified he’s going to kill me. I love Nick. He’s amazing. I’ve always felt very competitive with Nick, and I think he feels competitive with me. We have a common friend, Oleg, who we talk to a lot about these things, which is funny.

But the point is that, if you look at it, Klarna has 110 million customers worldwide. Revolut now has 65 million, right? So I’m twice the size in number of customers. The engagement that I have is not as high yet as it is with Revolut, so people may use us more or less frequently and for other things, right?

What I’m doing right now—what we’re doing at Klarna—is moving from being your infrequent payments solution to being your high-engagement banking provider, right? That transition is going extremely well. It’s going extremely well, and it’s accelerating. People are adopting our banking services at a very rapid pace.

Do you look at Robinhood? I’ve had Vlad on the show multiple times, and I think Vlad is astonishing. He’s got 11 lines of business that produce over $100 million in revenue. I think it’s an incredible model for moving from an entry product—BNPL, or for him, kind of free trading—into the full-stack banking provider. Do you look at him, and do you take lessons from that?

Yeah, absolutely. But we all have different entry points. I mean, Revolut, to some degree, was like an early-adopter thing: “Hey, I’m an expat. I travel a lot in Europe, and this is better for my currency,” or “I trade in crypto.” I think Robinhood was like, “I’m trading. I love trading.”

Klarna is very different. Our customer is like, “I shop online. I do shopping.” We skew more female than male. We have a very different brand. I would think of us more like a lifestyle brand, like a digital version of American Express. That’s how I think about Klarna.

I think all of these growing fintechs—I know David from Nubank from the early days, right? I have a funny story where I was down in Brazil meeting David when he was just leaving Sequoia to start the company. We had a fantastic conversation. We talked about the future of banking, and then he sent me an email which was like, “Hey, Sebastian, would you like to advise me a little bit? I’m starting this new company.”

And I was like, “I’m sorry, I don’t have the time.” So I lost out on a lot. I would have been a great angel investor there. But that’s how life goes. Yeah, you know, that’s how life goes.

I’m going to be honest. I’m not crying for you.

No. Okay, okay, okay. Yeah, I don’t think anyone will cry for me over that. I will cry over myself. That’s okay. I can fall asleep crying over that.

But the funny thing is that you have these fintechs now, right, who are actually starting to become really big. I mean, Revolut is really big, Klarna’s really big, Nubank, et cetera. We’re all coming from slightly different angles.

And now, the people who are going to be threatened by this are the incumbents, primarily. Of course, I’m going to partly compete with Revolut. I’m partly doing that already. But if you look at it, they’re very big in Romania. Romania is not a big market for me. I think it’s the second-largest market or something for Revolut.

If I look at the number of users, Romania is not a big market for me. I’m very big in other markets. So that’s going to differ, right? But whose market share are we eating? Barclays, Wells Fargo, Capital One—those are the companies that we are going after, right? So I don’t really see that there’s a big conflict between us. It’s more the incumbents that are going to lose customers.

You spoke about Nubank and David Vélez. I love David Vélez. Legend. Anyway, obviously, they got a bank license in the US, and they’re very aggressively planning the—Is the US the main goal for you?

Yes.

So, going back, sitting in London, I just feel so small in the US.

No, but listen, it’s like this. It comes back to what I said in 2015. The future of financial services is going to be this digital financial assistant.

So then the next question is: Who is going to participate in that challenge, and why would Klarna stand a chance? Those are the 2 questions we need to answer, right?

We said, well, first and foremost, there are going to be 3 types of companies participating: the tech companies—Google, Amazon, Apple—there are going to be fintechs like Revolut, which hadn't even started then, but we knew they were going to become entrants, and then banks. That was kind of our view.

So what's going to be critical for us to win in that big transformation globally? First, if we're just big at that point in time—we weren't even in the UK yet—if we're only in the Nordics and Germany, we're not going to have the scale to be able to win this big transformation that's coming. We need to be global, and global means the US.

If you're not in the US, if you're not big there, you're just not going to be big enough. The risk is that you're going to get acquired by somebody in the US. So, to us, nailing the US was a super-high priority. Super-high priority.

Second, we realized that the more I understand about you as a customer, the more likely I am to give you the advice that you said: “Hey, you should pick up the flowers when you go there,” or whatever. The key thing we saw, which was different from Revolut and all the others, is that we have our own payments network, just like Amex. We have our own rails.

Every time you shop with Klarna, the information that flows on the rails isn't just the amount that you purchased for. It's the exact products. We have the full digital receipt, so we know you shopped at Sephora, but also what cosmetics you bought at Sephora.

The benefit of that is, if I'm then supposed to advise you on your purchases or your day-to-day finances, I have much richer information that allows me to provide you with good advice: “Those contact lenses were really expensive. You can get them cheaper,” et cetera. If I want to help people with their everyday spending, I have more information.

So we said understanding the customer in depth is very, very critical, and data was going to be very, very important. It was about being global, having a good understanding of our customer, and having a lot of trust and brand. The other thing was brand: Build a brand that people relate to, that people feel emotionally connected to, and not just like a utility. Create something different.

I'm sorry. I struggle—and I'm a proud European—but I struggle when I look at the European neobanks and I look at Chime, Dave, and all the others, Current and—

I mean, they're all a fart compared to the market cap of Revolut right now, or whatever the private latest valuation is. Market cap and private valuation are very different things, as you know, as a public company CEO. Why, if the US is such a focus, has it been such a lackluster performance from their domestic participants?

Partially because competition is better. If you take your Amex app in the US and compare it to one here, it is significantly better there. Your JPMorgan app is significantly better. So the financial institutions in the US are just better.

The problem is that people then try to find an entry point that is probably, for example, very big in lending. Then they walk into a lot of subprime, make huge losses, or there are other things. People struggle to find that entry point.

But we have 30 million users in the US—almost, I think, 28 million or something, but soon it's going to be 30 million. Our card is growing at a very rapid pace in the US.

So your core focus is: Can I turn those 30 million users from buy now, pay later into core customer accounts?

Yes. We launched a card in the US, and I have to be careful now because we haven't released the new earnings yet—they're coming next week—so I have to use the Q3 numbers. But if I remember correctly, we're at 2 million or 3 million active cardholders in just a few months in the US.

We are transitioning these buy now, pay later customers into full banking relationship customers at a very, very high pace.

I totally get it. When you look at Nubank and Revolut moving aggressively into the US, if you were to put money on who's going to do better, who would do better?

That's a good question.

I'm a good interviewer.

Exactly. Who is going to do better? Klarna's going to do better. That's my answer. I can't tell you between those 2. It's very, very interesting.

I think—I have to go with David.

Wow. Why?

Because I think Nik's challenge right now might be that he is so distributed. He's trying to go for the whole world, right? He's launching in Dubai, he's launching in India, he's launching everywhere.

Banking is not your standard tech company. It's difficult, and so I just wonder if he's running out of bandwidth. I think it's more risky.

David at least has slightly more focus. He has a very solid base in Brazil, making a lot of money from there. He has his Mexican business and the other things that are growing pretty well, but then he does the US. That's just 1 more big market. It's very different. Nik is so thinly spread right now.

A question for you: You mentioned Stripe earlier and the competition there. You're a public company CEO now, so I'm going to go there, and you can give me a no comment.

I have so many public company CEOs on the show. I've never met a happy one. Never. Are you happy as a public company CEO?

“Happy” is a strong word. No, but I think, look, to us, we've had so many shareholders and so many employees over the years. At some point in time, it's actually easier to be public for us.

We were already reporting earnings on a quarterly basis. We're a bank, so I'm not sure that the difference is that big, to be honest, from our perspective. It is what it is.

Of course, if I could own the company 100% and be private, I would prefer it, but that's not reality.

Do Patrick and Stripe have advantages that you don't have as a private company in their ability to invest long-term in R&D and not think about next quarter quite so religiously?

Historically, yes, but this has changed due to AI. One of the things I remember vividly is that we came to the board and said that the same plan that I mentioned in 2015 was now going to happen: Grow globally, get customers across the board in all countries, and then, once you have the love and affection of these customers and the relationship with them for their day-to-day purchases, basically go deeper with them, offer them more banking-like services, and increase the revenue per customer.

That was the plan. We had this interesting board meeting where we said, “Okay, now it's really going to happen, guys. We are now going to truly start focusing on this transition from just a single-payments customer to banking. We're going to launch peer-to-peer, and we're going to launch potentially—it's also been leaked to the press already—trading. We're going to do a lot of the other banking services that we didn't have at that point in time.”

We're going to increase the card, increase balances and deposits, and do international remittances and all these things. Then the board looks at this and says, “These kinds of transitions can be hard. It changes what the bank is.” We had this discussion for an hour, I thought.

Thank you, board.

Yeah, this can be hard.

Well, no shit.

Yeah, exactly. But the funny thing is, then they said yes, and I went out and thought, “Why was that so easy? How come that went so easily?”

Then I realized why. Usually, when a CEO comes to a board and says, “We're going to do all of these new services. We're going to launch all of this new stuff. Hence, I need to increase my investment and my costs by $100 million, if it's the size of our company, to be able to do all of this.”

But at the same point in time, I was showing them a budget. Klarna has been shrinking. We used to be 6,000 or over 7,000 people, and we're now less than 3,000. I didn't ask for a single dime to do all this.

The reason for that is because I've seen the acceleration of AI, and I know we can ship all these things with the existing organization. We've gone from 7,000 people to below 3,000. We've shrunk by 50%, and the majority of that is just through normal attrition.

Initially, in 2020, we did a little bit of layoffs, but that was not that big a number compared to what has happened since then. That was the reason it was easy for the board to make those decisions. I didn't ask for a single dime in investments.

It's 2030. How many employees do you have then?

Less, for sure.

2,000?

No. I think it may very well be even less than that.

No.

Yes. But listen, again, relationships cannot be replaced. One of the—

We're going to do a show in 2035, and Seb's going to be the only—

Yep.

I think the thing that's important for us is relationships. We have relationships with merchants, as an example—retailers—and we have those relationships locally.

So I have people in Portland talking to Nike. I have people in China talking to SHEIN. I have people in Amsterdam talking to Adyen, et cetera, et cetera. We have over 50 locations where there are people. AI is not going to move those jobs like that.

What we need is the relationships with our partners, which are very important, and the same with customer service.

I'm still going to argue that it's going to be vital to offer a human connection there.

Yeah. So those jobs will remain, but for the rest, it's definitely going to be smaller. We're shrinking through natural attrition by about 20% per year. It's just people leaving; they stay about 5 years and then move on, which is natural. What we've said very clearly is that we're not going to recruit.

So we're recruiting a little bit. People came again on X saying, “Oh, it's not true. Look, they're recruiting.” Come on, guys. Yes, occasionally we hire somebody here and there, but if you go to LinkedIn and look at the insights, you're going to see how the company is shrinking. The point is that we've shrunk 50%.

But we also promised our employees, which is very important. We said, “Guys, this is going to mean we're going to do much more with much fewer people. This is going to make more profit for us, and you're going to share in that profit.” Our employee compensation has grown almost 50% per head during that time. We have given a lot of that money back, and that creates safety for our employees. They know that through this AI transformation, using these utilities, they're getting some of the benefit of that.

Weird question: how do you think about the current state of SBC—stock-based compensation, for anyone who's not familiar, which is obviously how we divide or give stock to employees? You see it aggressively from OpenAI, with the argument being, why does Sam give a shit if he doesn't have any stock anyway? Dilution doesn't matter to him. Evan Spiegel is the godfather of SBC—astonishingly high amounts. How do you feel about the state of SBC?

I think it's very clear when we compare American companies to European companies that there's a huge difference. I think American companies give 5 to 10 times more than European companies do. Klarna, from that perspective, is a European company. We have come from very low levels, and we have increased because we need to stay competitive for talent. Talent today can move between the US and the EU pretty easily with the help of companies.

I still think that the thing that's going to happen is that there have been these industries called tech and fintech—financial services—and they've all had this amazing thing, which is that you create this service and there's a huge switching cost, so your customers can't really switch that easily. Hence, you create this money-printing machine, and then life is sweet, right? You build these campuses, play volleyball in your office, go and get free lunches, and live off the spoils of this money-printing machine in your basement. That's not how normal businesses work.

If you're in retail or you run a restaurant, you freaking wake up every morning and ask yourself, “How do I put the right product in front of the right customer, bring them into my store so that I can actually sell to them and make them happy?” You have to wake up every day and care about that, right? The point is that this is going to be a brutal awakening for fintech and tech. That's what's going to happen to all of us: we're going to have to wake up every morning, make sure that we serve our customers, and work really hard and relentlessly to make them happy with what we're offering them. It's not going to be what it used to be, right?

I think share-based compensation, as an example—was it really because it made sense, or to what degree was it just a sport? To some degree, it's a sport. To some degree, it's relevant, because some people can make a huge difference. But there's a balance between the two.

There are an incredible number of incredible CEOs attacking the incumbent banks. What happens to BNP?

I think that again comes back to what we talked about in 2015. Our conclusion was that it was going to be different. What I thought was interesting was that David Solomon also recognized this at Goldman Sachs. He created Marcus.

Talking about public companies, the challenge was that when fintech was everything in 2021, valuations were up, and Marcus was the best thing Goldman had ever done. Then suddenly what ends up happening is Marcus turned, and it became very hard for David Solomon to defend Marcus. But the problem is that Marcus would have needed 5 or 10 years to fully mature, and when you're a public company, that's hard to defend. I don't know—maybe David Solomon doesn't agree, and he's like, “I did the right thing, and I changed it.” But in my opinion, he should have stuck to his guns.

Jamie Dimon is now doing neobanking. JPMorgan is going in, but it's going in a bit later now. You'll see different banks. I think some of them will reinvent themselves, become neobanks, become tech-led, and use AI to reinvent themselves. Some of them will not, and they will wither away, which is always what's happening with the disruption of an industry. It will depend on the leadership of those financial institutions.

You said about 2021 and the high valuations, high prices. I remember the round where you were done at $45 billion—who was it? SoftBank, no?

It's not entirely true. That's also a little bit of media, but yes, there were some shares bought at $45 billion as well.

Yes, that's right. Knowing what you know now, are you happy you did that? Is there anything you would have done differently?

It's a good question. I think that when you're in that kind of high-growth phase, one thing to keep a close eye on is your multiple expansion going faster than your revenue is growing. If your revenue is growing faster than your multiples, then you're probably fine. But if you start seeing the opposite, where multiples are expanding faster than revenue growth, that may potentially be a problem longer term, right?

I think today I could have been more careful specifically on the hiring side, because it was very sad and difficult for me to be hiring at a high pace a few quarters earlier and then, a few quarters later, have to announce layoffs. I felt that I should have predicted that and been more cautious about it.

Was that the hardest board meeting? You mentioned the board meeting earlier where I was like, “Oh, that was kind of nice and easy with the expansion.” You've got an amazing board. And we're not talking about the board in any other capacity; I'm just talking about a hard board meeting. Does Mike ever get angry, by the way?

No, he doesn't. Angry is not the word for Mike.

I would be fucking scared if he was.

No. I mean, it's actually less scary because he doesn't get angry. What he's like is disappointed.

Disengaged is the worst. The worst thing you don't want to get with Michael is disengaged. That's the big warning sign.

He's never lain on the floor and pretended to sleep.

No, he hasn't. But, you know, he's so amazing. I love him so much. I think he's fantastic. You obviously care a lot to make sure that you want to see that he's engaged.

Can I ask you what's been your biggest lesson from working with him? You're like his chosen child, in the nicest way, which is amazing, right? Well done.

Look, I've worked with him for so many years. It's a funny story how we ended up working with him. At that point in time, at Sequoia, there was a guy called Chris Olsen, who's amazing and still a friend of mine. He was at Sequoia, and he kind of founded Drive.

Drive, no?

Yeah, exactly. He did the Klarna investment.

There was this story where I was talking to a guy, saying, “Hey, do you think Sequoia could be interested in investing in Klarna?” There was a guy in Stockholm who knew Sequoia a little bit, so I was talking to him: “Do you think Sequoia could be interested in investing in Klarna?” He was like, “No, that would never happen.”

I was like, “Okay, fine.” So we just talked to the European funds. We were just like, “It's not even worth it, right?” But I was still looking at Google Maps. I was looking at Sand Hill Road, and I was dreaming: could I have the chance to work with these guys?

Then suddenly my phone started buzzing, and I was like, “Oh, I had a message.” I was like, “This is Chris Olsen from Sequoia.” I was like, “Oh my God. Uh-oh.”

What I did was—I knew this was like dating, right? I couldn't call back in 3 days. I had to wait 3 days before I called back. So I was just sitting there, counting the hours. I didn't want to look too interested.

Yeah, exactly.

I called Chris, and we took a meeting. They got super excited, and we instantly wanted to work with Sequoia. We were like, “They're the best.” That's how we saw it. We saw that this was going to be a huge brand uplift for the company.

At this point in time, Daniel Ek at Spotify was getting all the tech credibility. All the engineers in Stockholm wanted to work with Spotify. Klarna was some boring invoicing company. People were like, “What the hell is that?” So we were like, “We need tech credibility. Sequoia is going to give us that.”

Anyways, what happens is that a little bit later on, Sequoia decides to make the investment. Chris flies to Stockholm, and we met Michael Moritz at a hotel breakfast here in London. Then Chris comes, and he makes this beautiful presentation, right?

The presentation is: “These are the Apple guys in the garage, these are the Google guys in the garage, and these are the Klarna guys in the garage. You're going to be the next Google.”

We're just sitting there like, “Oh my God, we're going to be Google.” We were just buying it all.

And then I feel, after a while, that this is not good. We have to do something cocky here. We can't just swallow this and be like, “Okay, please, can we work together?”

So, when Chris exits the room and is just about to leave, I say, “Hey, Chris, just one thing. If we are genuinely the next Google, how come you're the only one from the Sequoia partnership who's here in Stockholm today?”

Chris looks at me and says, “Oh, I'm so sorry. The other guys couldn't make it.” You have to remember, they invested—they took a 25% stake at a $100 million valuation, right?

They took a 25% stake.

Yeah, at a $100 million valuation. That was the deal. They eventually did that.

So Chris is like, “I'm so sorry. I could make it.” He goes into the elevator and, literally, 20 seconds later—this is so impressive—my phone starts buzzing, and it's Michael Moritz. Michael Moritz is like, “Hey, I'm so sorry I couldn't make the meeting. If we get to do this investment, I'll join the board.”

So, thanks to me saying that and being a little bit cocky and not being so freaking Swedish, we actually got Mike on the board. Since then, I've worked with Mike and gotten to know him.

The thing people don't understand about Mike is that I always think that, for you to be really good and really understand the topic, you need to be in every freaking detail. You need to read up a lot, and he does. But the point is, it's almost like sometimes he can just take this huge mass of information and, without a second of thought, say, “That is important,” and just get it at a level I don't see in other people I interact with. He's just brilliant at that. It's amazing.

He called me in the summer of 2019. I don't know why he called me, but this was when we'd been in the US for a few years. We weren't getting any traction. The business wasn't doing well. Then Nick from Afterpay in Australia was starting to get traction in the US with buy now, pay later.

Michael just calls me out of the blue and says, “Sebastian, I think it's now or never. If we don't do the US now, we're never going to do it.” I was just like, “I don't know how the hell he knew that,” but he was so spot-on. It was exactly the thing.

I dropped everything and was just like, “We have to win the US.” Then I spent the next 2 years focusing 100% on that.

Does Sequoia move the needle for a company?

In my opinion, they do. Yeah, I think so. I think they do. I've been very impressed with all the people I work with there. I think they're amazing.

There's a new generation now with Sonya and Andrew, and the new guys coming. It's really cool. Pat and—what's his name from Zappos? Sorry. Alfred.

When you talk about the expansion of products, and you said obviously about Afterpay and BNPL—God, I'm going to get in trouble for this—I'm pleased to hear about the movement away from just pure BNPL, because is that not just evidence that, for all the people who said lending is a shitty business to be in, they were right?

Lending? What does that mean?

Well, BNPL is a shitty business.

Why would it be a shitty business?

Oh, it's really hard to build a $20–30 billion business on BNPL. Consumer lending is a hard business to make a lot of money in.

It's a hard business. Yeah. When I get a startup pitching me consumer lending, I'm like, “Sorry.”

Look, you're like, “Wow, he's honest.”

I think the way I thought about this is that, when we started Klarna 20 years back, we were just like, “Okay, why are we not just making these banking products better and working online?” You have to remember, in 2005, the banks' internet offerings were shit.

Dude, 2026. That's all shit.

So it was just like, they're all shit. We were just like, “Okay, we're going to take some of the stuff that the banks do and we're just going to do it better online.” We did that, and then that grew and we were successful.

People don't know this, but we raised $60,000 in our first angel investment. $30,000 of that was spent, and then we became profitable. We were running this as a profitable company from 2005 to 2019. We had almost 10 consecutive years of high growth and profitability, which actually got us this award, because we were, I think, one of only 2 companies in Sweden that had ever had such a long streak of high growth and profitability at the same point in time.

What we realized over time was that suddenly I'm sitting one evening and looking at my P&L, and I'm like, “Wow, what is that line? Oh, shit, that's late fees, right? That's a big revenue line.” I was like, “That's not going to be long-term sustainable.”

At some point, I started thinking, “Wow, you know what? We're actually doing lending. What does that mean? What does it mean for consumers? What does it mean for their financial lives? What are the implications of this?”

At that point, I thought to myself, “There are 2 things I can do here. I can either sell this business and say, ‘Oh, shit, we're making a little bit too much on interest and late fees. Let's go and do something else,’ or I can try to change this.”

One of my co-founders left at that point, but I said, “No, I'm going to stay and I'm going to make the change.” I realized that the buy now, pay later credit offering is healthier than your credit card.

On your credit card, you put all your spending for a full month on that, and then the bank tries to push you to revolve. You build up a balance of a few thousand dollars, and then you pay very high interest.

I remember when I worked at Burger King, it used to be like, “Press 1 for debit, press 2 for credit,” when I would swipe my card. I was like, “Where did that go?” Banks didn't like it, because the problem was that if you pressed debit, then your bill at the end of the month was much smaller and you were less likely to revolve and borrow money, so they would make less money. They removed the debit button.

I was like, “No, no, no. Let's bring that back. Let's make sure Klarna offers ‘Press 1 for debit.’” Twenty percent of our transactions are debit, and the rest is credit, but the credit is interest-free, fixed installments. No revolving.

We removed revolving. It cost us—we gave up $100 million of revenue when we removed it, because we used to do it in the Nordics. We took it away. We even didn't have late fees in the UK for a period of time at all, but it turned out that wasn't great either, because then people were, to some degree, overextending themselves.

It's good to have a little bit of a fee, some kind of consequence for not paying on time. So we got that back, but you have to be mindful of not making too much money on it, because people will tend to use it in a way that's not good for them. You just have to find the balance.

Over the years, we iterated on a model that we feel is a better alternative to credit cards. If, 10 years from now, fewer people have credit cards and more people have debit cards, and then use buy now, pay later occasionally, I would argue it's a better society. That's my belief, right? That's what we've been pushing.

The consumers now—we see that they love that. They reward us for that. They agree with that, right? That doesn't mean you're not still using credit. You're still going to have, unfortunately, occasional people who overextend themselves. You have to be mindful about that. You have to think about how you help them when they're distressed and so forth.

It's a difficult business in that sense. It's a bank, right? So you have to be mindful about these things. But generally speaking, the type of product we're offering is better than the traditional products of the banks.

You mentioned starting with Spotify and Klarna in Stockholm together. Obviously, Stockholm has been the birthplace of great AI companies in the last year, with Legora and Lovable, to name a few.

I'm a young, 18-year-old Stockholm entrepreneur. Seb, you've got this incredible experience and you've been to the US—oh, how wonderful. Do I have to be in the US if I want to build a big startup today?

No.

Huh?

Do you disagree?

No shit. I wouldn't be doing this project. But every US-experienced founder tells you, “Yes, you do.”

Yeah. I think, to some degree, from a European perspective, it's obviously partially sad to see that a lot of the successful AI companies are in the US, where the founders are actually European.

100%.

Right. It's a bit sad, but that's just how it is. I remember Truecaller. The founders are friends of mine, and it's also a fantastic European company.

I was there when they did that.

Yeah, Nami and Alan. They were told by the VCs—the American VCs—“You have to move your engineering center to Silicon Valley, because otherwise everything's going to go to shit.”

They did, and it was a disaster. They had it there for a year. They tried to recruit, but obviously nobody in Silicon Valley knew what Truecaller was. They were already a pretty grown company, so they couldn't attract people as a startup in that sense. They had a hard time recruiting and whatever.

They were just like, “Why are we doing this?” Then they shut it down.

They lost a year on that—lost so much traction. Then some of these VCs were like, “Oh, you’re not doing well, so now we’re not going to put our partner on your board anymore. We’ll put some junior guy on your board because we don’t care, because you’re not our top priority of companies anymore.”

It’s the most savage indictment, isn’t it? When you get the associate joining the board and you’re like, “Oh, fuck, I’ve been relegated.”

Exactly. So first they’re advising them to do this thing, and then when they actually go and execute it and it turns out to be a disaster, they’re like, “I’m sorry, your company is not that good.” I mean, it’s just terrible, right? It was terrible. And I think that—

Do you think the state of VC is very good today, actually?

I think, again, it’s changing so fast. I think the challenge right now is that a lot of them are piling money into AI that they don’t necessarily fully understand. How good is it? How differentiating is it? Does it really have a real moat?

Well, what should I know, then, as one of these VCs piling money into AI?

I think what you should be doing is coding with Cursor and building things yourself. If you do that—I think you do that, if I remember correctly.

Yeah, but Lovable.

Yeah, yeah, yeah. Well, only Lovable, or have you tried Cursor as well?

I’ve tried Claude Code.

Okay, good. The point is, if I meet investors today who haven’t actually downloaded these tools and tried to build something themselves, I think they don’t have the skill set to evaluate the company they’re looking at. I think it’s so critical to actually understand how powerful these tools are today before you make those decisions. If you have that insight, if you understand that, and then you make your decisions, fine. There are going to be opportunities.

I think Cursor will lose half of its revenue in 2026.

Is that your prediction?

Yeah.

Why is that?

Because Claude Code has just eaten their lunch. I don’t see any engineering team that’s still on Cursor, and that’s—

We love it, actually. We use it all the time.

Really?

Yeah.

Is that because you have an enterprise deployment and contract?

No, I don’t think so. I don’t know why. I kind of switch between Claude Code and Cursor. I’m a big Anthropic fan. I love Claude, the chat version of it, as well. I use that all the time.

I just find that it depends on the task. Sometimes I’m almost using them as I would use different people: I would go into Cursor, write some things, then go to Claude Code and ask Claude Code to do some other things. I still feel they have almost distinct personalities and skills, so I kind of enjoy still using them. And I need an IDE.

The problem is also that, because I wasn’t an engineer, I never used VS Code or any of these tools. I still need an IDE today, so Cursor is my standard IDE, just as a consequence of that. You could be right, but I’m actually more optimistic about its future than that.

You can invest in Anthropic at $360 billion or OpenAI at $500 billion. I’m giving the discount to make it easier.

Don’t make me answer this question, please.

I think they’re going in very different directions.

I think that, at least in my experience with OpenAI, it’s becoming a consumer company. If I’m building AI for a billion people, I would focus on making sure that, for example, there are going to be people who seek AI as a friend—as a friend in their day-to-day life, as somebody more like the AI from that movie. What’s the movie again?

The Scarlett Johansson one.

Yeah. What’s it called?

Her.

Exactly. So some people are going to look more for the Her experience. I think that’s ChatGPT to me, because if I’m so big—if I’m such a big consumer brand—I’m going to start looking at my KPIs and optimize for emotional connection. How much time are they spending with my product?

Yeah. Is that really going to be a ChatGPT, or is that going to be a companion product? And there are 15 providers that are so specialized—

It could be, but my point is that if I’m just looking at who their audience is today, it’s very likely that they will optimize for that relationship aspect of being your counselor, your provider, your play friend that you’re playing with or playing games with, or whatever.

Claude, to me, is very different. Claude is my intelligent advisor, and I say to Anthropic all the time, “I don’t want an AI that tells me, ‘You’re so great, man.’” I understand that that’s a nice experience and some people may enjoy that. I don’t actually enjoy somebody just telling me how awesome I am every day. Personally, I’m not that interested in that part.

What I’m interested in is somebody telling me, “Sebastian, that’s freaking stupid. Don’t do that. That makes zero sense.” I want an AI to tell me, “You’re wrong, man. Don’t do that. That makes no sense.”

I feel currently Claude is more likely to provide me with something less biased. It’s trying less to please me. I think if you’re building a product, OpenAI, the risk is that you become too pleasing, because I think a lot of people like that.

I mean, I understand: if I were to use AI for entertainment, then I would also want to be entertained. I want to feel pleased. I want to feel happy. It’s a different product. See what I mean? It’s just a different thing you’re looking for. I have less interest in that. I have more interest in somebody telling me, “You’re totally off. Don’t do that.”

So you’d rather be Anthropic.

Yes, from that perspective, yes.

Yeah, $150 or $149 billion.

I didn’t look at the valuation. I thought $150 billion was the revenue expectation by 2030.

Oh, really?

Yeah.

Which was phenomenal.

Yeah.

You invest a lot now as well through Flat Capital?

No.

What have you changed your mind on since also being an investor?

I have changed my mind on software. I think software is getting much more risky, and it’s much more unclear what the future of SaaS is. So generally, we, for example, made a big investment in Defensor, which is into military arms—you know, that type of thing. Not arms, but military defense.

I think data centers are one of the most underinvested categories today. When you look at inference needing to run 24 hours a day for most of the knowledge-worker population, while it’s running for only 1% of the knowledge-worker population today, I’m like, “Why the fuck is more money not going into data centers?” Do you agree with that? And do you think—

It’s funny you asked me about this. I’ve played around with Suno sometimes when I have time. Have you played around with Suno?

Dude, I love Suno. I actually sent Nik at Revolut a song to get him to come on the show most recently, and it helped get him on the show.

It helped get him on. Yeah, I love Suno. What I’ve done with Suno is—I was playing around, and I’ve been doing some songs. I even published them on Spotify for the fun of it. You can go to Klab; I have around 74 monthly listeners or something.

My kids get so annoyed when I publish songs in my name. It’s really funny. Two of those songs are called “Compression.” This came from a conversation I had with Claude, so I blame all the lyrics on Claude. Don’t blame them on me. I was only the producer. Claude wrote the lyrics. Then I had to give Claude some artistic freedom.

I’m very interested in this. I got this question at a conference in Yellowstone, where I was onstage, crazy enough, with Sam Altman and Eric Schmidt. Someone in the audience asked, “How is it possible that you can take the whole of ChatGPT-5, as an example, one of these models, once they’ve been trained, once the training is over and the whole thing is done, and fit it on a USB stick? How is it possible that it’s not bigger in size? It’s just a few hundred gigabytes, or whatever the size of the models are?”

They gave different answers, and I had an answer in my head, but I felt embarrassed in that setting to say it. I think what people underestimate with AI is that it’s a compression technology.

What that means is, if you historically put data in a database, you say, “Klarna has a customer called Sephora,” and then we write again, “Klarna has a customer called Sephora,” you create a tremendous amount of duplication. If you look at any large enterprise company, they will have the same information over and over and over again, right?

But if you look at Wikipedia, how many articles are there about Klarna? 1. Why aren’t there 15? What do they do so magically? How can it be that Klarna historically had a customer relationship with Sephora and we had information about that customer relationship in Slack, in Salesforce, in Google Docs, in Google Slides? Kind of the same information over and over again. But on Wikipedia, it’s just 1 article. How do they do that?

I realized that when you train the model, if I tell it that Harry not only runs a fantastic podcast but also runs a VC, and you tell it that once when you train it, it will forget it.

It will ignore that information. But if you tell it enough times, it will remember it, and then when you go and ask it, it will know that information. But it's not storing it twice, because if it's getting the same information that it already knows, it doesn't move the tokens. So it's automatically compressing all the information.

This is why you can take the whole freaking internet—all human knowledge—and compress it down to a few hundred gigabytes. That's a huge amount of information. Now, obviously, you lose precision. This is why it's very worthless to go and ask, “What are the opening hours of the Starbucks down on the corner?” The AI would be like, “I have no clue.”

If I asked this AI—AI is responsible if I'm wrong—but ChatGPT-5, as an example, the number of gigabytes that model is equivalent to about 2–3 days of weather data from the whole globe. That's it. So how can it then be so capable of answering all these questions? Because unfortunately, despite what we humans would like to say, the amount of true, novel information and knowledge in human society is quite limited. What we see is repetitions and variations on the same themes over and over and over again.

So, to answer your question, when we realize this, we also realize that we're compressing knowledge at an extreme level. That's why people are playing around doing this, like, “Oh, look at my Mac Mini. I was running my own model on it. It actually works. I can run this on a Raspberry Pi.” The thing is, unlike Romeo and Juliet, that story exists in 100 different variations. Unfortunately, when you compress it down with math, AI sees that not as Romeo and Juliet, and then another love story, and another love story. It is a love story, and then it knows it has slightly different names in different variations.

So it's a massive compression. And now comes the question: Do we need all that compute in the future? I happen to have had this amazing conversation with Michael J. Burry about this the other week.

Wow.

Because I was talking about this—he's making those bets that it's not right. I think there are 2 arguments, for and against. One is, if you look at enterprise, what does enterprise want? Enterprise wants the highest quality at the lowest cost.

Why would I recompute my information about Sephora over and over again? If somebody could help me compress that down to just one single source of truth and not have all that unnecessary information, I'll take it. I'll save a lot of money. I don't need all that compute. Why do I want to do that? So, in enterprise, you're going to see a dramatic squeeze and compression.

The counterargument to this is that you and I then go out and say, “Hey, we've had this amazing podcast. Now let's watch a movie together. We want to watch Star Wars, but with our faces. So you'll be Darth Vader and I'll be Luke.” Right now, that needs generation. That needs a data center to generate that for us. So the question is, which power will be greater: the compression of enterprise data or the generation of new stuff for entertainment and other things? I don't know the answer to that question. I think you can argue both ways.

But in enterprise data, there's going to be a massive compression that's going to come naturally through this technology. That's why I'm a little bit like—I don't want to be the guy who said there's only going to be 4 computers in the world. I don't want to be that guy. So I want to be a bit mindful—

Clips stay forever. Don't do this to yourself.

He said there was only going to be—

In 2040, they're going to play this clip. I was like, “Can you believe it?”

He was so stupid. No, but I'm just saying that it will. So I don't know which power is going to be greater, right? Which one?

If you run with that enterprise compression, what does that mean for data centers and for chip companies like Nvidia?

Well, that would mean you would need significantly less. That's the consequence. Why do we have so much big software? Why do we have so much data? It's all a mess because we had humans who were trying to do their best, but the right hand didn't know what the left hand was doing. They were overwriting that, and the code over there didn't connect with the code over here. Then people started doing transformations and stuff, and it all became a big mess.

Actually, Wikipedia is the most successful knowledge graph in the world, I would argue. But if you look at the standards that they apply, how do they do that? They're very disciplined. I'll give you a very simple example. I love this principle.

If you go to Wikipedia and try to create a new article about a new topic, it's very hard. If you go to Google Docs, you just click “New”—boom—you start writing a new document. If you go to Cursor, you start a new codebase. Wikipedia: Where is the “New” button? There is none. Do you know how to find a “New” button? No. You have to search for an article, and if you search for something that doesn't exist, then you're allowed to create new. See what I mean?

That's so different. Companies don't work like that. In companies, it's like, “Hey, I have an idea. I think Klarna should be doing this.” “Okay, let's just start coding.” But let's go and check. Maybe we already have code doing this. Maybe we already do that.

So the point is that the reason compression hasn't happened historically is just because so many people have been involved, with different experiences, knowledge, and so forth. It just creates this massive mess, and nobody has been as disciplined as Wikipedia has been in keeping it to one source of truth and so forth. They've really been amazing at that. You can read up a lot about how they've been doing it, but AI is going to help with that.

AI is going to help organizations say, “Hey, should we really be doing this thing, because we already have code for this?” and so forth. It's not there yet, but that's going to happen—not because AI gets smarter, but because it's economically sound in a business. It's not economically sound to duplicate. It's economically sound to reuse what you already have. That's the reason it's going to happen. It's not because this is some hypothesis. It just makes sense economically not to do what you've already done again, right?

Were there any other takeaways from your conversation with Michael J. Burry? He's a phenomenal thinker.

Yeah. No, we enjoyed it. I think, in this regard, we were aligned, right? But it was funny because we were talking about this: How much novelty is there really? It's an interesting concept because, as I said to him, there's other data that suggests otherwise.

I've heard data that suggests that 30% of the searches on Google every day are new. I don't know if that's true. It sounds crazy to me, but there is some data to suggest that. So maybe there is more novelty. I don't know the answer to these questions, but I think this is very, very fascinating—to see the countereffects of these things.

Before we go into a quickfire, it's been fascinating not knowing the answers to these questions. I'm sure you're in CEO groups, go to CEO events, sit in green rooms, and there are topics or themes that CEOs discuss about AI that they do not discuss publicly. What do you think they mostly are?

Well, as I said previously with Dario, I think most of them recognize—first of all, I don't go to that many of those, to be honest, because I'm hard-coding and working with my teams and trying to make sure that Klarna is as successful as possible through this transformation. So I spend very little time on those kinds of things.

Does this generation make every CEO, even public-company CEOs, a builder again?

I think it has to be. Yeah.

Doesn't it?

Well, I was with Mike from Atlassian last night, and he's like, “I'm up at 5 a.m. coding.” I'm like, “Was that the same before?” He's like, “No.”

No. Well, I think so. But it's also amazing, right? It's accelerating. I mean, I think for somebody like myself who didn't used to code or couldn't code, I think it's fantastic to be able to take my ideas and thoughts and turn them into something I can show others. It doesn't have to be production-ready, but I can articulate things at a very different level of quality than just trying to explain something on a whiteboard or whatever. Now I can actually bring things to people.

I had this very, very unique experience with Claude just last week. We were trying to communicate a very specific thing that was touching on accounting, finances, predictions, and stuff. It was a very, very complex thing, and we were just talking about it. This was the first time I went to Claude and said, “Hey, can we interact a few times? I want to see if we can explain this concept.”

After a few iterations, I got this beautiful animation. It wasn't even a slide because it was an HTML file, but it was just like—and I thought to myself, “Wow, you know what? This is actually the first time I felt that AI could do something that humans couldn't.” It was the first time I had that experience.

The reason was that if I had wanted to do the same animation and beautiful pedagogical explanation of a very complex technical accounting and finance thing, historically I would have brought in an animator, a designer, an accountant, a finance Google Sheets person, et cetera.

Each one of them may have been extremely skilled at what they do, but they didn't necessarily know what the other person needed in order to perfect that outcome. The animator would have been like, “Yeah, I can animate like this and that,” but they wouldn't really understand what they were animating—the financial concept. The financial guy may have been like, “I can do the numbers like this and that, but I don't really understand why we would need this animation, because I read the numbers and I get everything. I don't need this freaking visualization.”

But here, Claude had all the skills in one and created this, and I was like, “I could not have done that.” It was such an experience. There's a special moment when AI exceeds human capabilities.

I think in this experience, to me, it did something that I don't think a single person could have done—not each individual part, but all the different parts. How many people are you going to find in the world who are great animators, great visualization people, super-pedagogical, know everything about financial accounting, and understand financial services? Who are you going to find?

Rare Venn diagram.

Yeah, exactly. It's going to be hard to find the person who's good at all of these things in one, right?

Final one for a quick-fire: What happens to Elon with Grok? I would never dare to bet against him. When you have 2 people you would never dare to bet against in the same market—and most people would never dare to bet against Sam—then one of your prior assumptions has to be wrong.

Mm-hmm.

And so, if you'd never dare to bet against Sam, but you'd also never dare to bet against Elon, one of your prior assumptions has to be wrong.

Yeah, I think he's going to do really well. If you think about the fact that he could, in a few weeks, put together a frontier model at that quality level, you've got to give credit to the guy. That's just insane.

Do you use it ever?

Actually, you know what I love about it? I think he's—and this is funny, because this is exactly what Elon has been saying all the time, and nobody wants to give him credit for this—but on X, people spread all these rumors. Yesterday, there was a rumor that Klarna offers buy now, pay later for rent.

Yeah, yeah. You're doing rent.

Yeah, we're doing rent. We're like, “No, we're not doing rent.” Nobody cares, right? But one thing that makes a difference is that if it were established media, they would fix it. But these X posts and TikTok posts—nobody does anything. It just goes one way, right? Then people write, “@Grok, is this true?”

Grok actually almost always answers correctly, and I think that, to me, is very, very impressive. I think it speaks to Elon's vision of X: that over time, it can become the trusted source of information.

And what he's done with Grokipedia, where he's taking inspiration from Wikipedia, as I said, and so forth—I think there's something there. That's going to be critical for our societies, because right now we have these scams flowing all over the place, and AI is creating virtual versions of everything. To me, that's where the holy grail is, where there's something really, really exciting.

I agree. Do you know, I just wish you could hide the @Grok, because I'm too embarrassed to ask, “What is an ideal...?” You just explain it quietly; I don't want anyone to see this.

Dude, I want to do a quick-fire with you. What have you changed your mind on most in the last 12 months?

I've changed my mind most about the pace at which the transformation is happening. I thought it was going to happen faster than it did. My problem is that I overestimate how long it takes people to change habits and ways of working.

I actually think adoption is going to take a little bit longer. It's not necessarily the capabilities of the technology, but how fast people will adopt it and how it will change things.

Do you think that differs for enterprise versus consumer? I'm actually the opposite. I'm surprised by how quickly consumers adopted it when you look at the wow numbers for ChatGPT.

Yeah, but consumers are always going to be faster, right? To your point. Way faster.

Yeah, way faster.

I think we underestimate how fast consumers adopt it and overestimate how fast enterprises adopt it.

Yeah, for sure. In work life, it's definitely going to be slower.

What criticism about you stings because it's partly true?

No, I think the criticism that stings is the one that isn't true. People have sometimes said, “Oh, he's just trying to make an exit. He's just trying to make a big, fast buck.” After 20 years, I feel like, hello—haven't I proven now that I'm in it for the long term?

The other one is that I don't care about the consumer and just want to make money on interest rates—reckless lending and all that stuff. None of that is true. I deeply care about our customers. I deeply care about them being financially well-off, and I think I'm providing a product that's better than the alternatives.

That's the one that stings because it's just not true. I don't understand how you'd ever level that criticism against you. You're battling as a public company CEO every day. It's the opposite of a quick buck.

What would you do first if you were not a public company and had no scrutiny on you? If I said, “Here's an invisibility cloak. You can do anything, spend any money on anything internally,” what would you do?

The only thing I would do differently is spend less time communicating and talking to investors. That's it.

To be honest, when I look at the strategy and what I'm trying to do with the company, I honestly feel like, as I said, that 2015 vision—let's be that digital financial services assistant—is the one I'm executing on. We're executing on it. The company's executing on it. I feel we're having tremendous momentum on it. The adoption by our customers of our banking products is amazing.

Do you think you've told a good story around that?

No, I think communication is hard.

Because, respectfully, I didn't know this. I feel terrible.

But you and I talked about it even before this podcast started. I think my problem is—and this is definitely what Michael gives me as feedback—that I need to stick more to the message. I probably failed at this podcast again. I tried to talk about all kinds of things.

I love my work. I love what we're doing. I think so many things are interesting and complex, and sometimes I want to tell people everything. I want to explain everything to them, and I should be better at sticking to a clear message.

No, totally bad advice. People buy you, not what you sell.

Yeah.

We don't release 30% of shows because CEOs come on and just sell: “Let me tell you why our banking products...” It sucks.

People buy you, not what you sell. I hope so. I hope that, over the years, that will pay off.

What's funny is that when we were the highest-valued fintech in Europe, I could do no wrong. Whatever I said was like, “That's brilliant. He's so smart.” Then, when we came down to 6.5 and had to do layoffs, it was, “Everything he does is a disaster.”

Was that a brutal time for you?

Of course it was. It was intense.

How do you deal with that?

To me, it's also one story. I had this very crazy thing: I've been doing this for many years now, so I'm a little bit more thick-skinned. But what happened is that I was on an interview with MSNBC. I've been interviewed by them many times, and generally speaking, the questions had been balanced and fair.

Suddenly, on this interview, they brought in this new guy. This was exactly when all of this was happening, and I was sitting in my home because I was doing the interview from my house. The guy comes in and says, “This is a disaster. The company is going to shit.” I basically didn't say that word, but that was the gist of it: “This is the end of Klarna. It's all going down.”

He's putting this massive pressure on me, and I'm just sitting there. I'm almost starting to crack up and laugh because it was so insane. It's like, come on, the business is doing really well. Obviously, we need to make some changes, but it's going to be fine.

I'm trying to deal with it as professionally as I can, saying, “No, I don't think so,” as you try to do in those situations. But I come out of it thinking, “Oh my God, that was intense.” So I get in my car, drive to the office, and think, “There's only one song I can listen to now.” Obviously, I put on Queen's “Under Pressure.”

I put it on at maximum volume, and I'm just sitting there under pressure, and I crack up. The thing is, I said to myself, “Look, one of the biggest idols of mine is obviously Zlatan Ibrahimović. He was born in Sweden on the same day I was, 3 October 1981.” I think of him and say, “I wanted to play in the Champions League. How the hell does it feel to go into the finals and play in the Champions League?” He actually never won the finals, unfortunately for him.

But every soccer player dreams about being in the Champions League final, right? Or football player. Can you imagine the pressure when you're going into that stadium? Everyone's screaming. This is the height of your career. This is the chance you've got at winning this thing. That's what I signed up for. I signed up for being on that interview with MSNBC. This is what I signed up for. It is stressful. It was hard as hell, but this is what I wanted. I wanted to play at this level of the league.

I also have to cherish and be happy about the fact, and look at it with gratitude, that I get the fantastic opportunity in life to experience these things. It's amazing. I'm going through this amazing experience, and that's the only way to look at it. So as much as I can laugh and cry, and I can be sad, and I've had really tough times where I feel super depressed about stuff, at the same point in time, I'm always looking at it, putting on “Under Pressure,” and being like, “But this is what I signed up for.”

Yeah, I love that too.

Yeah.

Did you ever interview him?

Do you know what?

You should bring him here.

No, we should do that. There's a brilliant video where he's like, “When I step on the pitch, I think I am God.”

Yeah. I don't think I'm God, though. But did you always think you would succeed there?

No, but it's exactly the same. There's this funny email that I found that I wrote only 6 months into the company. I found this email just by coincidence, and it's to my co-founders. It basically goes like this. This is 6 months into the company. We just started, and we're getting our first customers. The email is written at about 11:30 p.m., and it goes, “I'm sitting here myself, and I started thinking, ‘What if we're actually successful with this thing? What if we start growing, maybe from Sweden to Finland and Norway, and then we go to Germany? What if we actually grow this globally? What if we actually go after the banks and start building financial services?’”

So basically, in that email, I write everything that's happened over the last 20 years. The point is, did I know that was going to happen? No. But just like Zlatan, when he was kicking a ball down in Malmö, on the street, did he dream about being in the finals? Of course he did. And of course I did as well. I dreamt about being where I am now, and even more so where I want to take the company in the next decades.

You dreamt about being here with me.

Of course.

Sweet Seb. I did see it.

This was a very vivid dream.

But I think visions are bullshit. All VCs say, “Hey, what's your vision for a pre-seed company?” Dude, if I told you, when you were starting your business 20 years ago, “We're going to be a fully fledged banking provider. We're going to have BNPL as the insertion point. We're going to own the US as well,” you would not have been like that. You unlock the next chapter with every achievement, in my mind. I think visions are actually the most constraining thing that we force founders to try and articulate. Am I wrong?

Well, I think you're wrong with me. Again, I'm not telling you that I knew how we were going to do it. I had no clue how, but I was dead dedicated to doing it. When I was a kid, I was an immigrant kid. My parents quarrelled a lot. They divorced when I was 8. My dad started drinking. It was quite a chaotic upbringing.

In my childish interpretation of what was happening in the family, they were always fighting about money. If I fixed money, then everything would end happily ever after. That was my childish interpretation. I also got very interested in Richard Branson for whatever reason. I read his book, and I've now met him, which was a fantastic experience for me. I read about him and was just like, “Wow, this guy built Virgin, did the records, and this and that.”

The other guy who really inspired me was Ingvar Kamprad, the founder of IKEA, who built this company and, for a period of time, was seen as the wealthiest man in the world. For whatever reason, I was always enthusiastic about building businesses. I remember we had this school night in middle school where the cafeteria was always the place that made the most money. There were tons of other businesses, but they all failed, and nobody was really making a lot of money. The point was to raise money to go on a class trip or whatever.

I started a pizzeria. I started selling pizzas, and we outcompeted the cafeteria and took all the money. My class could go on a nice school trip because we made all the money that the cafeteria was making. I've always wanted to drive a business. I've always had this. Eventually, I said to myself, “The coolest business to build must be a bank. That must be the ultimate business.” Banks always win. Banks always prevail. You see the JPMorgans of the world and all this. Sometimes they get financial crises, of course, but in the longer term, the bankers always win. That was always an inspiration to me.

Can I ask a hard question? I've invested in 13 unicorns. Well done me. VCs congratulating themselves. The single most common feature is a broken relationship with their father.

I, too, have that. My father had a drinking problem, or has had a drinking problem.

It's not “whatever,” though. Did he quit?

No.

Okay.

I don't really see him.

That's too bad. I'm sad for you.

Did that drive you in a way that you wouldn't have otherwise, if you don't mind me asking?

For sure. I think it's a combination of that and being an immigrant kid. Seeing all these other Swedish kids going to their summer holiday houses and having a surplus that looked great to me, and then us eating pancakes 7 days in a row because it was the cheapest food that Mom could put together because we were out of money—even though I love pancakes and thought they were really delicious—I think it created this feeling.

I felt my parents were smart. My dad deserved something better—but “better” is the wrong word. He started driving a cab, and he was very smart, very intellectual. That was not the right job for him. He was saddened when he saw drunk people and was driving them around and hearing what they were saying. I think it was just not his place to be in life. It broke him down.

I felt that if I was going to fix things, if I was going to get money, then I was going to fix this. The problem was, once I got money, it turned out that life isn't like that, because I gave a lot of money to my father, and he used it to drink more. He drank himself to death. It turned out that money was not going to solve those problems. There are some problems that money won't solve.

Again, I don't want to have that discussion about whether money makes you happy or not. Obviously, when you've been as fortunate as I have in life, and I can take my kids on an amazing vacation, I can do things, and I don't have to think every day, “Can I afford this? Can I do this?”—that's a tremendous luxury and privilege in life. At the same point in time, I've also experienced that it doesn't solve all problems. It's sad, because that was partially my aspiration with doing this.

Final one. I like to finish on a theme of positivity. What are you most excited for in the next 10 years? My mother's got MS, so I'm very excited for developments with diseases and treatments for diseases like MS. What are you most excited for?

I think with AI, obviously, the thing is that if you'd asked me 5 or 10 years ago, I could have said, “I kind of think I know what's going to happen in the future. It's going to continue like this,” or whatever. Then suddenly came COVID and the war in Russia, and the world just changed. Then suddenly came AI, and you sit there and think, “I have no idea how the world is going to be in 2 years. I have no clue.”

For myself, what I'm most excited about is that I'm still an optimist at heart. I do believe that these technologies will make life better for humans. I think they will actually lead to something positive. I'm in that camp. You can have an intellectual debate with me if you want to, but I think that's true.

What I'm most excited about for myself is that I want to realize the vision of Klarna. Against what you said about not having visions, I want to put that vision into reality. I want to finally bring a banking product that truly helps people save time, save money, and be in control of their finances. That excites me. It excites the hell out of me.

I think all these incumbents have been having all these excess profits. They've made so much money because people don't switch, and honestly because they didn't care enough about their customers. They didn't wake up like that restaurant or retailer every day and say, "What can I do to make my customer better off?" They didn't do that. I'm honestly excited about that journey.

The point is, I know that if I make more money, I'm not going to be happier because I have a bigger pile of money. That's not the point. But the journey of trying to make Klarna into that global retail bank, and the adventure it encompasses, excites me. Going through all these different challenges and opportunities and trying to make the best that I can of delivering on that thing just really excites me.

Now AI is enabling me to do things that I couldn't do with this company before. I can realize those visions faster and at a higher quality than was ever possible before. That's super exciting.

Dude, I've so enjoyed this. Do you see? It's so much nicer to do in person. Thank you so much for being so brilliant and being so open. It's been such a joy.

Thank you. It's been a joy to be here.