Klarna CEO Sebastian Siemiatkowski on 20 years building from Burger King to 100M users and the AI moment in fintech

Summary

Klarna CEO Sebastian Siemiatkowski has spent 20 years building the company from a niche European buy-now-pay-later product into a global payments network with 100 million users, $100 billion in annual payment volume, and roughly 500,000 merchants including Nike, Macy's, Sephora, and Walmart. The company employs around 3,000 people and holds a Swedish bank license that passports across EU markets.

Origin and US entry

Klarna's founding addressed a real friction in early e-commerce. Customers didn't trust unfamiliar merchants enough to pay upfront, so offering payment after delivery removed the barrier. This model mirrored 20th-century US mail-order credit, where cash-on-delivery histories built early credit bureau scores. Klarna expanded from Sweden into Finland, then Germany and the UK, always assuming the US was out of reach because of credit card saturation. The 2008 financial crisis changed that calculation. Post-crisis regulation, including a ban on credit card marketing at university campuses, compressed credit card growth while debit card volume grew 10x between 2010 and 2020, compared to roughly 2x for credit cards. The US consumer base converged toward the European model Klarna was already built for.

Bank license and balance sheet stability

Klarna obtained its bank license roughly 10 years ago. The primary advantage was not deposit float, which Siemiatkowski acknowledges Klarna has underutilized. Instead, the license provided balance sheet resilience. Many fintech lenders offloaded credit risk to third parties, a strategy that worked until underwriting models failed and counterparties pushed losses back. Klarna, as a regulated bank funded by government-insured deposits, holds capital against its own balance sheet and is not dependent on wholesale funding markets. That structural difference is why Klarna survived cycles that bankrupted or collapsed the valuations of competing fintechs.

Stablecoins

Siemiatkowski was a crypto skeptic for years, applying a simple test: does this make his mother's life better? About a year ago he revisited that position. He now sees genuine promise in next-generation stablecoin infrastructure, particularly for cross-border corridors with weaker currencies where traditional rails are expensive and slow. On higher-volume corridors like UK-to-US, the advantage is smaller. Payments is a $2 trillion global revenue pool heading toward compression. There is no fundamental reason sending money should cost more than sending an email, and stablecoins accelerate that race to the bottom.

AI and the workforce

Two years ago, after experimenting with early LLM tools, Siemiatkowski concluded that revenue per employee would become the defining competitive metric in an AI environment. Klarna stopped hiring. The company had 5,500 employees at the time. Through natural attrition roughly 20% annual voluntary turnover in tech it has shrunk to 3,000 without mass layoffs. Engineering headcount held flat at 1,500. All the reduction came from functions like marketing, analytics, and operations. Revenue per employee has moved from roughly $400,000 to over $1 million, which Siemiatkowski compares favorably to Goldman Sachs's current ratio, with Apple and Netflix cited as the next benchmark at $2 million.

The savings from lower labor costs are being recycled into accelerated cash and equity compensation for remaining employees, framing efficiency gains as a shared dividend rather than pure margin capture.

The one function exempt from reduction is B2B enterprise sales. Klarna runs sales staff in over 50 offices globally, sitting alongside merchants like Shein and TikTok in China and Nike in Portland. Human relationship-building in enterprise sales has no near-term AI substitute.

A separate workforce trend he flags is convergence. Several hundred Klarna business employees have spent the past two years learning to read, inspect, and interact with code using tools like Claude Code and Cursor. Combined with their commercial and customer knowledge, which engineers sometimes lack, they are becoming the company's most productive workers. The implication is that the most valuable new hires are not pure engineers or pure business generalists, but people who can operate credibly across both.

SaaS consolidation

Klarna's AI strategy required a unified data layer first. The company closed down Salesforce and Workday and is planning to remove Atlassian/Jira, along with more than 1,500 SaaS vendors in total over the past two years. Fragmented, inconsistent data across many tools is unusable as LLM input. Klarna retains a handful of vendors where the product is genuinely differentiated, citing ElevenLabs for voice and FullStory for user interaction analytics as examples. Everything else that was essentially a database with a UI and some process logic is being rebuilt internally on a consolidated data model.

Macro outlook

Siemiatkowski is less focused on tariffs. He says merchant behavior shifted sharply when the first announcements hit, with some large Chinese partners redirecting marketing spend from the US to Europe, but those signals have since faded as the policy environment kept changing. The risk he watches more closely is white-collar job displacement driven by AI. Consumer credit performance is more sensitive to job loss than to sentiment shifts. A material acceleration in white-collar unemployment would translate directly into reduced repayment capacity across Klarna's book.

GDPR

His critique of European cookie law is that GDPR's original intent was sound. Giving users ownership and portability of their data made sense. The implementation produced the opposite outcome. Friction-heavy cookie banners reduce conversion rates and discourage users from trying new services, which entrenches the large incumbents GDPR was meant to discipline. True data portability, the ability to extract your data from one platform and move it seamlessly to a competitor, would force companies to compete on actual user value rather than lock-in. That vision was never properly legislated into existence.