Klarna CEO Sebastian Siemiatkowski: SaaS Is Dead, Systems of Record Are Next

Software costs are going to zero. That’s the thesis. And Sebastian Siemiatkowski isn’t just saying it from a conference stage. He’s executing it inside a public company with 114 million users and 3.4 million daily transactions, cutting Klarna’s workforce in half while the business keeps growing.

Episode Overview

This is a wide-ranging conversation between Sebastian Siemiatkowski and Harry Stebbings on 20VC. Siemiatkowski, who has run Klarna for 20 years, is one of the few public company CEOs who has aggressively restructured around AI rather than just talking about it. The conversation covers the collapse of SaaS valuations, Klarna’s internal AI transformation, the competitive landscape in fintech, and the founder psychology of building through crises. What makes it valuable is the specificity: actual headcount numbers, real decisions about building vs. buying, and candid reflections on what CEOs say in private versus in public.

The End of Software as We Know It

Siemiatkowski’s core argument is straightforward: if the cost of creating software approaches zero, then the value of owning software approaches zero. SaaS companies that charge subscription fees for code will see their multiples compress dramatically.

“The cost of creating software is going down to zero. That’s it. Everyone will be able to generate software at any point of time.”

He draws a distinction between two types of value in enterprise software. The first is the software itself, the code, the features, the UX. This is what’s being commoditized. The second is the data, the customer records, transaction history, behavioral patterns accumulated over years. This retains value, and possibly increases in value.

His prediction: revenue multiples for SaaS companies will converge toward 1-3x, down from the 20-50x premiums of the recent era. The argument is that if any company can generate equivalent software on demand, the switching cost collapses, and with it, the ability to charge premium subscriptions.

The Real Threat: Data Switching Costs Are Dissolving Too

But Siemiatkowski goes further. Even the data moat is under threat, just on a different timeline.

The “next thing that’s going to hit everyone bad is the switching cost of data.” Historically, enterprise customers stayed with vendors because migrating data was painful and risky. But AI is making data migration tractable. If an AI system can understand, map, and transfer data between systems, then the last defensible moat for SaaS companies erodes.

This leads to what he calls “enterprise compression”: the idea that AI won’t just replace individual software tools but will consolidate entire categories. Organizations today run dozens of overlapping systems because different teams bought different tools at different times. AI can identify redundancy, merge systems, and reduce the total software footprint of an enterprise. He draws an analogy to Wikipedia’s approach: you can’t just create a new article, you first have to search for whether it already exists. Companies don’t work that way today, but AI will push them toward consolidation.

Klarna’s AI Transformation: 7,000 to Under 3,000

The most concrete data point in the conversation: Klarna has gone from approximately 7,000 employees to below 3,000 through natural attrition (no layoffs needed), while the business continued to grow.

“We’ve gone from 7,000 people, we’re now below 3,000. We’ve shrunk 50%. And I didn’t ask for a single dime to do all this.”

When asked about 2030 headcount, he suggests it could be “even less than” 2,000.

The transformation started with customer service. Klarna built its own AI customer service system rather than buying from a vendor. The reasoning was that Klarna handles millions of customer interactions daily; the data from those interactions is more valuable than any off-the-shelf solution. By building internally, they could train on their own data and iterate faster.

The key insight: for functions that are core to your business and generate proprietary data, building your own AI is a competitive advantage, not a cost. For commodity functions, it doesn’t matter.

Build vs. Buy: Why Klarna Chose to Build Its Own AI

Siemiatkowski is emphatic about building customer service AI internally. His reasoning has several layers:

1. Data advantage: Klarna processes 3.4 million transactions daily. That volume of interaction data, when used to train and fine-tune models, creates a system that understands Klarna’s specific customers better than any generic solution.

2. Speed of iteration: With an internal team, they can ship improvements daily. With a vendor, they’re waiting on the vendor’s roadmap.

3. Cost structure: The AI customer service system is dramatically cheaper per interaction than human agents, and the cost continues to decline as models improve.

4. Competitive moat: If every fintech company buys the same customer service AI from the same vendor, there’s no differentiation. Building your own creates asymmetry.

He acknowledged that this logic doesn’t apply to everything. For back-office functions that don’t generate strategic data, off-the-shelf solutions are fine. The principle is: build where data creates compounding advantage, buy where it doesn’t.

The Nubank Lesson: A Billion-Dollar Miss

Siemiatkowski reveals that he had the opportunity to invest in Nubank early and passed, a decision he estimates cost him roughly a billion dollars.

What he admires about Nubank is their execution discipline and their focus on the customer. He believes Nubank is more likely to succeed in the US market than Revolut because Nubank has built from scratch in Brazil, a massive and complex market, and demonstrated that they can win against incumbent banks.

His take on Revolut: they have twice fewer customers than Klarna in terms of base, and he’s confident Klarna will outcompete them. But Nubank is a different caliber of competitor.

How Sequoia and Michael Moritz Joined the Board

The story of how Klarna got Sequoia to invest is revealing about founder persistence:

Siemiatkowski visited Sequoia’s offices four years in a row, each time pitching the vision of Klarna as a digital financial services platform. The first three visits ended with polite rejections. On the fourth visit, Michael Moritz agreed to join the board.

What changed wasn’t the pitch. It was the evidence. After three years of consistent execution, the numbers spoke for themselves. Siemiatkowski’s takeaway: the best investor relationships are built through repeated exposure and demonstrated execution, not through a single brilliant pitch meeting.

Once on the board, Moritz became a genuine strategic partner. Siemiatkowski describes him as someone who challenges thinking at a fundamental level and calls out “a piece of advice that I’ve carried with me”: stick to the message, focus on the core story, don’t try to explain everything.

What CEOs Really Think About AI (But Won’t Say Publicly)

Siemiatkowski reports that in private, most CEOs he talks to, including conversations with Dario Amodei, recognize the transformative potential of AI far more than they let on publicly. The gap between private acknowledgment and public statements is significant.

He observes that this generation of AI is turning every CEO, even public company CEOs, into builders again. He mentions having dinner with Mike Cannon-Brookes (Atlassian), who told him he’s now “up at 5 a.m. coding,” something that wasn’t the case before.

For Siemiatkowski himself, who wasn’t a coder, AI tools like Claude have been transformative. He describes a specific experience where he used Claude to create an animated visualization explaining a complex financial accounting concept. The result combined skills that would typically require an animator, a designer, an accountant, and a finance specialist. No single human would have all those skills; AI was the first time he experienced something exceeding human capability.

“This is actually the first time I felt that AI could do something that humans couldn’t.”

Investors Who Don’t Build Will Lose

Siemiatkowski makes a pointed argument about the investment landscape: VCs and board members who don’t personally use AI tools will lose the ability to evaluate AI-native companies.

The reasoning: if you haven’t experienced what AI can do, you can’t assess whether a startup’s claims about AI-driven efficiency are real or inflated. You’ll either miss genuine AI-native companies or overpay for AI-washed ones.

This extends to his view on enterprise AI adoption. He’s revised his estimate of how quickly the transformation will happen, now believing it will be slower than he initially expected. The technology capabilities are advancing rapidly, but human adoption, especially in enterprises, lags significantly behind consumer adoption.

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

The Valuation Trap

Siemiatkowski speaks candidly about Klarna’s own valuation journey, from being Europe’s most highly valued fintech to a painful markdown to $6.5 billion, and the layoffs that accompanied it.

His framework for thinking about when high valuations become dangerous: the risk isn’t the number on paper. It’s the behavior it induces. High valuations can make founders complacent, make hiring harder (because equity expectations are inflated), and create a gap between external perception and internal reality.

He shares a specific example: when Klarna was at peak valuation, everything he said was received as brilliant. When the valuation cratered, the same type of decisions were seen as disastrous. The substance hadn’t changed; only the frame.

Under Pressure: Building Through Crisis

The most personal segment of the conversation reveals Siemiatkowski’s psychological framework for handling adversity.

He describes an MSNBC interview during Klarna’s crisis where a hostile commentator essentially declared the company dead on air. After the interview, he drove to the office, put on Queen’s “Under Pressure” at maximum volume, and reframed the moment through the lens of his idol Zlatan Ibrahimovic (born on the same day, October 3, 1981).

His reframe: professional athletes dream their entire lives about playing in the Champions League final. The pressure, the scrutiny, the stakes, that’s the dream, not the obstacle. Similarly, being a public company CEO facing intense scrutiny during a crisis is “what I signed up for.”

The deeper story is his immigrant background. His parents divorced when he was eight, his father started drinking. As a child, he concluded that money was the root cause of family dysfunction, and fixing the money would fix everything. That drove his entrepreneurial ambition. The painful coda: once he had money, he gave generously to his father, who used it to drink more and eventually drank himself to death. “There are some problems that money won’t solve.”

Dreams, Visions, and the Email from Day One

Six months into founding Klarna, Siemiatkowski sent a late-night email to his co-founders laying out everything that happened over the next 20 years: expanding from Sweden to Finland and Norway, then Germany, then globally, then attacking the banks with financial services.

He pushes back on Harry’s assertion that “visions are bullshit.” While acknowledging that you can’t know the specific path, he argues that the directional ambition matters. He knew he wanted to build a global bank from the beginning. What he didn’t know was how.

His disagreement with the common VC framing: asking pre-seed founders for a detailed vision is constraining, not enabling. You don’t unlock the next chapter until you’ve achieved the current one. But having the ambition to imagine those chapters is what separates founders who build something enduring from those who optimize for an exit.

A Few Observations

This conversation stands out because Siemiatkowski isn’t theorizing about AI disruption from the sidelines. He’s living it inside a 20-year-old, publicly traded company with real revenue, real customers, and real consequences for getting it wrong.

• The SaaS compression thesis is now supported by concrete evidence: a public company has cut its workforce in half while maintaining growth, and its CEO predicts even further reduction. This isn’t a prediction anymore; it’s an ongoing case study.
• The build vs. buy framework for AI is clearer than most: build where data compounds, buy where it doesn’t. Simple, but most companies haven’t internalized it.
• The adoption speed mismatch, consumers fast, enterprises slow, explains the current disconnect between AI capability and AI impact on employment. The technology is ready; organizations aren’t.
• The most underappreciated risk to SaaS companies isn’t AI replacing their software. It’s AI making data migration easy, which dissolves the last real lock-in.
• His personal story, from immigrant kid eating pancakes seven days straight to running a $6.5B+ fintech, is a reminder that founder drive often has roots in something that can’t be manufactured or taught.