From “Interesting Experiment” to “Most Profitable Digital Bank on Earth”
When Nubank launched its purple credit card in Brazil in 2013, the dominant narrative from incumbents was that serving the country’s underbanked population at scale — people who had been rejected by fee-heavy banks, who distrust financial institutions, and who live in markets where credit risk is structurally elevated — was not a viable business model. It might work as a social enterprise. It could not possibly generate the margins that institutional banking requires.
The full-year 2025 results, reported on 25 February 2026, close that debate. Nubank generated $16.3 billion in revenue, up 45% year-on-year from 2024. Net income was $2.9 billion, compared to $2.0 billion the prior year. Return on equity reached 33% — a record high and one that exceeds the median ROE of Brazil’s incumbent banks. The efficiency ratio — the ratio of operating expenses to net revenues — dropped to 19.9% in Q4 2025, falling below 20% for the first time in company history. For context, JPMorgan Chase operates at an efficiency ratio of approximately 55–60%; Bank of America at 65–70%. A ratio below 20% is not merely best-in-class for a digital bank: it is in a category of its own.
The company now serves 131 million customers globally, adding 17 million net new customers in 2025 alone. In Brazil, where it serves 113 million customers, Nubank has reached approximately 62% of the adult population — making it effectively a mainstream financial institution, not a niche disruptor.
Advertisement
The Four Pillars of the Nubank Model
1. An almost-free customer acquisition machine built on word-of-mouth
Nubank’s monthly cost to serve per active customer was $0.80 in Q4 2025 — below one dollar. This is the figure that most surprises banking industry analysts encountering the model for the first time, because traditional retail banking typically spends $200–$400 to acquire a customer and $15–$40 per month to serve them.
The unit economics work because Nubank spent years building a product that customers genuinely recommend. The original waiting-list model — where approved customers could invite friends — created organic viral growth that avoided the customer acquisition cost spiral that kills most challenger bank business models. As of 2025, the majority of Nubank’s customer additions in Brazil still come through referrals, not paid acquisition. The $0.80 monthly cost per active customer is partly a technology efficiency story (no branch network, cloud-native architecture) and partly a product-market fit story: when customers love a product, they sell it for you.
Monthly Active Rate is 83% — meaning 83% of Nubank’s 131 million customers engage with the product at least once a month. This is an extraordinarily high engagement rate for a financial services product, where dormant accounts often represent 30–50% of a customer base.
2. Revenue per customer growing faster than the customer base itself
Average Revenue Per Active Customer (ARPAC) reached $15 per month in Q4 2025 — a 27% year-on-year increase. The customer base grew 15% over the same period. Revenue per customer is growing nearly twice as fast as the customer base, which is the signature of a platform that is successfully cross-selling: moving customers from single products (a no-fee credit card) into a broader financial relationship (savings accounts, personal loans, insurance, telecom, investments).
Nubank launched NuInsurance (insurtech products) and NuCel (a mobile virtual network operator) in Brazil during 2025. These non-banking products contributed to the shift in revenue mix: 55% net interest income (loans, credit), 45% non-interest income (fees, insurance, services). A revenue mix that is nearly half non-interest income is exceptionally resilient — it reduces the business’s exposure to interest rate cycles and credit loss provisioning swings that have historically made banking earnings volatile.
3. Geographic expansion using the Brazil playbook — not ignoring the complexity
Mexico and Colombia are Nubank’s growth frontiers. In Mexico, it received a full banking licence in 2025 — unlocking deposit-taking capability that transforms the business from a credit card issuer into a full deposit-gathering bank, materially expanding the lending base and therefore the interest income runway. Mexico’s adult population of approximately 93 million represents a market of similar scale to Brazil; Nubank already serves approximately 14 million Mexican customers (roughly 15% of adults) with limited products.
The Mexico banking licence is the pivotal event of 2026 for Nubank’s medium-term model. Deposit-funded lending at scale is dramatically more capital-efficient than relying on wholesale funding markets, reducing funding costs and improving spreads. The same licensing milestone in Brazil — crossed in 2013 — was the structural moment that enabled the Brazilian profitability curve Nubank has now achieved.
4. A technology cost structure that compounds efficiency gains over time
Nubank operates with a cloud-native, microservices architecture built in-house. It has no legacy core banking system, no branch network, and no paper-based operational processes. The efficiency gains from this architecture are not linear — they compound as the customer base scales, because the marginal cost of adding a customer to a cloud-native platform is a fraction of the cost of adding a customer to a branch-based bank.
The declining efficiency ratio — from above 30% three years ago to 19.9% in Q4 2025 — is the quantified expression of this compounding. With each additional million customers, the fixed-cost base is spread more thinly, and the operating leverage grows. Nubank’s conditional approval from the US Office of the Comptroller of the Currency for a national bank charter in January 2026 signals that its next compounding opportunity may be the US market — a detail that will become increasingly significant across 2026 and 2027.
The Broader Lesson: Profitability Is Not the Enemy of Financial Inclusion
The Nubank results challenge a narrative that has persisted in development finance circles since the mobile money revolution of the early 2010s: that financial inclusion and commercial returns are in inherent tension. The argument goes that serving low-income, underbanked populations requires cross-subsidies, donor capital, or below-market returns because the economics of serving high-risk, low-balance customers cannot support profitability at scale.
Nubank’s 33% ROE, earned primarily from a customer base that was structurally excluded from formal financial services a decade ago, demonstrates that the tension is not inherent — it is a product of legacy cost structures, exclusionary product design, and risk assessment methodologies calibrated to reject rather than include.
The Nubank model works because it eliminated the cost structure (branches, legacy systems) that made inclusion unprofitable, used data and technology to price credit risk more accurately at individual level rather than relying on demographic proxies, and built products that customers with prior negative banking experiences actually want to use. The 83% monthly active rate is the empirical refutation of the “underbanked customers are too risky and too disengaged to serve profitably” premise.
For the cohort of digital banks now operating in sub-Saharan Africa, Southeast Asia, North Africa and the Gulf with 1–5 million customers and pressure from investors on the path to profitability, the Nubank data provides both a map and a timeline: the inflection from “interesting experiment” to “profitable platform” took Nubank approximately 8–9 years from founding. The structural drivers that produced the inflection — product-led growth, cross-sell depth, technology leverage — are replicable in any market where mobile penetration and a significant underbanked population coexist.
Frequently Asked Questions
How did Nubank achieve a $0.80 monthly cost per customer at scale?
Nubank eliminated branch infrastructure entirely, building 100% on cloud-native systems from founding. All customer service is digital-first with AI-assisted support. The company processes credit decisions algorithmically using behavioural data rather than manual review. At 131 million customers, fixed costs are amortized across an enormous base — but the cost efficiency started from architectural decisions made before the company had any customers.
Why is a 19.9% efficiency ratio significant?
The efficiency ratio (operating expenses divided by revenue) is the standard banking profitability metric. Traditional banks in Brazil and globally typically operate at 50–65% efficiency ratios. Legacy infrastructure, branch networks, and manual processes create fixed cost floors. Nubank’s sub-20% ratio — achieved at 131 million customers — is comparable to technology companies, not banks. It is the proof point that the digital-native architecture delivers structural cost advantages that compound over time.
What would an Algerian digital bank need to replicate this model?
Three prerequisites: a regulatory license from the Banque d’Algérie (EME or full banking license), a mobile payment rail with merchant acceptance (extending BaridiMob or building on Satim infrastructure), and a credit scoring capability using alternative data (mobile payment history, utility payments, behavioural data). The technology is available; the regulatory pathway is the primary dependency.
Sources & Further Reading
- Nu Holdings FY 2025 Financial Results — Nu International
- Nu Holdings Q4 2025 Earnings — BusinessWire
- Nubank, WeBank and MYBank Top the 2026 World’s Best Digital Banks Ranking — The Asian Banker
- Nubank Reports 45% Revenue Increase in Latest Earnings — American Banker
- Nubank Statistics 2026 — Business of Apps
- —













