David Vélez - Building Nubank and the Future of Global Banking - [Invest Like the Best, EP.495]

What It Covers

Nubank founder David Vélez explains why digital banking's disruption of incumbents is still in early innings, with fintech holding under 10% of a $10 trillion global market. He covers AI's role in democratizing financial intelligence, Nubank's expansion into the US and 35 countries via stablecoin accounts, and how the company rebuilt its organizational structure using AI workflows.

Key Questions Answered

• •Cost Structure Advantage: Nubank serves 140 million customers with 10,000 employees while competitors need 100,000 employees and 6,000 branches for similar scale. This produces a cost-to-serve roughly 3% of traditional banks, enabling zero fees and higher yields. Companies entering financial services should quantify this structural gap before competing, as it determines how much value can realistically be returned to customers.
• •Three-Bucket Framework for Banking Disruption: Vélez categorizes global banking problems as access, cost, and intelligence. Access: 3 billion people lack credit. Cost: US banks pay depositors 40 basis points on $15 trillion while rates sit at 4.5-5%, representing roughly $600 billion annually flowing to fund branches. Intelligence: AI can deliver private-banker-quality financial optimization to the bottom 99% of earners who currently receive no personalized guidance.
• •Proprietary Foundation Models as Competitive Moat: Nubank acquired Palo Alto-based Hyperplane to build a financial services foundation model trained on 20,000-30,000 data points per customer across 140 million customers over 13 years. This model now drives credit and fraud decisions across Mexico and Colombia, and will deploy day-one in new markets including the US, compressing the underwriting learning curve that typically takes years to develop.
• •Organizational Flattening via AI: After mapping 120 distinct workflows across the entire organization, Nubank reduced management layers from 10 to 5 and expanded the CEO's direct reports from 1 to 20. The trigger was recognizing in early 2025 that AI models had crossed a capability threshold for Clojure programming. Leaders should audit workflows systematically rather than waiting for organic adoption, as a 60-day lag behind peers produced measurable competitive disadvantage.
• •Cash Digitalization Unlocks Credit Markets: When informal cash transactions move onto digital ledgers through real-time payment systems like Brazil's PIX or India's UPI, previously unbanked small businesses gain documentable cash flow histories. A taco stand owner with recorded daily transactions can suddenly access credit to expand operations. This mechanism will unlock trillions in credit for billions of people across cash-heavy economies in Latin America, Africa, and Southeast Asia.
• •Agents as a Distinct Customer Segment: Vélez frames AI agents as a formal customer segment requiring dedicated product design, marketing, and SEO strategy—equivalent to how Nubank treats under-18 users or small businesses. Companies should assign teams specifically to understand agent behavior and optimize for agent-driven discovery. However, banking's trust requirements mean human-facing brand building remains essential alongside agent-facing infrastructure, requiring parallel investment in both channels simultaneously.