The Purple Card That Ate a Continent
In June 2021,
Warren Buffett's Berkshire Hathaway wired $500 million to a company in São Paulo that had no branches, no teller windows, and no legacy of the kind Buffett had spent six decades celebrating. The check went to Nubank — a digital bank born in 2013 from the fury of a Colombian-born, Stanford-educated engineer who could not open a simple checking account in Brazil without enduring five hours of bureaucratic theater. By the time Berkshire invested, Nubank had amassed 40 million customers. By the time it went public on the New York Stock Exchange six months later, the count had crossed 48 million. As of the first quarter of 2025, the number stands north of 114 million — making Nubank, by customer count, the fifth-largest financial institution in the Western Hemisphere and the largest digital bank on Earth.
That Buffett wrote the check is less interesting than why. The Oracle of Omaha had famously avoided technology companies for decades, only to discover — through Apple, through his lieutenants Todd Combs and Ted Weschler — that certain technology businesses function not as speculative bets but as toll bridges with compounding economics. Nubank, in Buffett's framing, was not a tech company that happened to do banking. It was a banking franchise — vast, sticky, structurally advantaged — that happened to run on code. The distinction matters enormously, because it captures the essential paradox at the heart of Nu Holdings: a company that built one of the most capital-efficient financial platforms in history inside economies where the cost of capital has historically been ruinous, that amassed a customer base the size of Japan's population while spending a fraction of what traditional banks spend on acquisition, and that turned profitable — genuinely, sustainably profitable — in a macroeconomic environment that has destroyed lesser fintechs by the dozen.
This is the story of how that happened, and what it costs to keep it going.
By the Numbers
Nu Holdings — Q1 2025
114.2MCustomers across Brazil, Mexico, and Colombia
$11.5BAnnualized revenue run rate (Q1 2025)
$557MNet income, Q1 2025
$2.8BTrailing twelve-month net income
30%ROE (annualized, Q1 2025)
$49BApproximate market capitalization
$3.60Average monthly revenue per active customer (ARPAC)
83%Monthly activity rate among customers
Five Hours in a Bank Lobby
David Vélez grew up in Medellín, Colombia, during the years when the city's name was still more closely associated with Pablo Escobar than with innovation hubs and urban renewal. His father was an entrepreneur; the family had means but not insulation from Latin America's chronic institutional dysfunction. Vélez studied engineering at Stanford, then worked at General Atlantic and Sequoia Capital — the venture firm's Latin America operation, where he was tasked with finding deals in Brazil's overheated consumer economy. He moved to São Paulo in 2012, already fluent in the paradox that would become his founding thesis: Brazil had the world's most profitable banking system and the world's most underserved banking customers, simultaneously.
The top five Brazilian banks — Itaú Unibanco, Bradesco, Banco do Brasil, Santander Brasil, and Caixa Econômica — controlled roughly 80% of deposits and 85% of credit. They operated more than 20,000 branches each. Their return on equity regularly exceeded 20%, and in some years approached 30% — figures that would make a Jamie Dimon weep with envy. And they achieved this not through operational brilliance but through what Vélez later described, with barely contained contempt, as a cartel-like oligopoly that extracted rents from a captive population. Annual credit card interest rates exceeded 300% in some products. Monthly bank fees consumed meaningful portions of minimum-wage earners' income. The customer experience was, by any global standard, abysmal.
Vélez's personal inflection point — the founding myth, polished by now through hundreds of retellings — involved attempting to open a bank account at a large Brazilian bank. The process consumed an entire afternoon: waiting in line, producing stacks of documentation, enduring the condescension of a system that treated customers as supplicants. He emerged enraged and certain. "I walked out of that bank branch and thought, there has to be a better way," he has said in various formulations. The anger was personal. The opportunity was structural.
Brazil has some of the most profitable banks in the world, and some of the least satisfied customers in the world. That gap is the opportunity.
He recruited two co-founders whose complementary skills would prove critical. Cristina Junqueira, a Brazilian who had run the credit card division at Itaú — meaning she understood precisely how the incumbents' economics worked, from the inside. And Edward Wible, an American computer scientist from Princeton who would architect the technology stack. In May 2013, the three of them set up in a small house in the Brooklin neighborhood of São Paulo. The company was called Nu Pagamentos — "Nu" being Portuguese slang for "naked," as in stripped down, transparent, no hidden fees. The first product was a credit card. Purple, because no Brazilian bank used purple, and because Vélez wanted the card to be physically recognizable — a tribal marker, a status symbol for a generation that felt exploited by its own financial system.
The Credit Card as Trojan Horse
The decision to start with a credit card rather than a bank account was not aesthetic. It was strategic, and the reasoning reveals the architecture of everything Nubank would become.
In Brazil, credit card issuance required a payment institution license — not a full banking license. This meant Nubank could enter the market with lower regulatory capital requirements and faster approval timelines. A full banking license would have taken years and required the kind of capitalization that a seed-stage startup could not command. The credit card was a regulatory arbitrage play wrapped in a consumer product.
But it was also a customer acquisition play of extraordinary elegance. Brazilian consumers hated their banks, but they loved credit cards — Brazil was already the fourth-largest credit card market in the world. The problem was not demand but supply: incumbents limited card issuance to existing customers, charged extortionate annual fees, and made the application process punitive. Nubank offered a Mastercard-branded credit card with no annual fee, managed entirely through a mobile app, with real-time transaction notifications and a simple, clean interface. The application took minutes. Approval came in days. There was no branch visit, no paper, no condescension.
The waitlist launched in September 2013. Within months, it had grown to hundreds of thousands of names. The viral mechanics were organic and ferocious: in a country where complaining about your bank was a national pastime, receiving a purple Nubank card became a social signal — proof that you had escaped the system. Customers photographed their cards and posted them on Instagram. They invited friends. The referral dynamics were self-reinforcing in a way that no acquisition campaign could have manufactured.
By mid-2014, Nubank had raised a $14.3 million Series A from Sequoia Capital and Kaszek Ventures — the latter co-founded by MercadoLibre veterans who understood Latin American consumer technology better than perhaps anyone alive. The money funded credit card growth. But the card was never the point. The card was the wedge — the initial surface area through which Nubank would insert itself into the financial lives of tens of millions of Brazilians, then expand into deposits, personal loans, insurance, investments, and eventually full banking.
The Cost Structure as Competitive Weapon
What made Nubank's expansion possible — what made it lethal to incumbents — was not the product. Products can be copied. It was the cost structure.
Traditional Brazilian banks operate branch networks of staggering scale. Itaú Unibanco alone maintained approximately 4,700 branches in 2018, each requiring real estate, security, tellers, middle managers, and the full apparatus of physical retail distribution. The cost-to-serve a single customer at a traditional Brazilian bank was estimated at $15–25 per month. Nubank's cost-to-serve, by contrast, was approximately $0.70–$1.00 per customer per month in its mature Brazilian operation — a 15-to-20x structural advantage.
This was not merely a function of having no branches. It was the compound result of several architectural decisions made in the company's first two years:
Cloud-native infrastructure. Nubank built its entire stack on public cloud — initially AWS — at a time when every Brazilian bank ran proprietary on-premise systems designed in the 1990s. This meant Nubank's marginal cost of adding a customer was approximately zero in infrastructure terms, and its ability to deploy new features was measured in days rather than the 6–18 month cycles typical of incumbents' IT organizations.
Clojure and functional programming. Wible made the unusual decision to build the core banking system in Clojure, a functional programming language that ran on the Java Virtual Machine. This was not fashionable and made hiring harder — but it produced a codebase that was more reliable, more testable, and more amenable to the kind of rapid iteration that a financial services company attempting to scale from zero to fifty million customers would require. The choice was a tax on recruitment and a subsidy on resilience.
Customer service as product. Nubank invested heavily in in-app customer service — chat-based, responsive, empathetic in tone — and made it the primary support channel. The resolution rate within the app was extraordinary, reducing call center costs and creating a feedback loop: good service generated word-of-mouth, which reduced customer acquisition cost, which freed capital for further service investment. By 2018, Nubank's Net Promoter Score exceeded 87 — in an industry where the Brazilian average hovered around 12.
The cumulative effect of these decisions was a cost structure that allowed Nubank to offer products at prices incumbents literally could not match without cannibalizing their own branch economics. A bank with 4,700 branches cannot suddenly charge zero annual fees. The fixed costs demand extraction. Nubank had no fixed costs to amortize. It could price at the margin — and the margin, for a cloud-native digital bank, approaches zero for basic products.
The Funding Escalator
Between 2013 and its December 2021 IPO, Nubank raised approximately $4.2 billion in private capital across multiple rounds, at valuations that tracked the company's customer growth with an almost mechanical correlation. The funding history is itself a case study in how venture capital can be deployed as a strategic weapon in financial services — not merely to fund losses, but to build the balance sheet credibility required to issue credit at scale.
Key funding rounds from founding to IPO
2013Seed round from Sequoia Capital and Kaszek Ventures — $2M
2014Series A — $14.3M led by Sequoia Capital
2015Series B — $30M led by Tiger Global and Sequoia
2016Series C — $80M led by Founders Fund
2018Series E — $150M led by Tencent at ~$4B valuation. Nubank surpasses 5 million customers
2019Series F — $400M at $10B valuation. Nu becomes Latin America's most valuable fintech
2021Berkshire Hathaway invests $500M. Pre-IPO valuation reaches $30B
2021
The investor roster reads like a deliberate strategy: Sequoia for Silicon Valley credibility and operational expertise. Tiger Global for growth-stage aggression and capital velocity. Tencent for the imprimatur of the company that had built WeChat Pay — proof that a messaging-first financial platform could work at billion-user scale. Founders Fund for contrarian conviction. Berkshire Hathaway for the ultimate legitimacy signal in financial services. Each investor brought not just capital but a different form of institutional validation that mattered in Nubank's negotiations with Brazilian regulators, card networks, and funding partners.
The IPO itself, in December 2021, was the largest fintech IPO ever at the time and the largest listing of a Latin American company since 2004. Nubank sold shares at $9 per share, valuing the company at $41.5 billion — briefly making it more valuable than Itaú Unibanco, Brazil's largest private bank. The valuation was aggressive by any standard, and the stock promptly fell as the 2022 fintech correction swept through public markets. Shares touched $3.26 in June 2022. The bears declared the thesis broken.
They were wrong, though their timing was impeccable.
The Banking License and the Second Act
Nubank obtained a full banking license in Brazil in 2017, through the acquisition of Easynvest (later renamed) and the establishment of Nu Financeira. This was not a checkbox exercise. The banking license transformed Nubank from a credit card company into a deposit-taking institution — and deposit-taking, in Brazilian financial services, is the master key.
With deposits, Nubank could fund its own loan book rather than relying entirely on securitization markets and credit facilities from other banks. The cost of funding dropped precipitously. In a country where the Selic rate — Brazil's benchmark interest rate — fluctuated between 2% and 13.75% during Nubank's first decade, the ability to fund lending operations with low-cost or zero-cost demand deposits was an extraordinary competitive advantage. By Q1 2025, Nubank's total deposits exceeded $28 billion, with a cost of funding meaningfully below that of traditional banks — because Nubank's deposits were largely transactional, sitting in NuConta digital accounts used for everyday payments and Pix transfers.
The product expansion that followed the banking license was rapid and deliberate. NuConta — the digital account — launched in 2017 and grew to become the primary banking relationship for tens of millions of Brazilians. Personal loans followed, starting with small-ticket unsecured credit and expanding into payroll-deductible loans, auto loans, and buy-now-pay-later products. NuInvest (the rebranded Easynvest acquisition) offered investment products — CDBs, government bonds, equity trading. Insurance products — life, auto, home — arrived through partnerships and proprietary underwriting. A marketplace for third-party services began to take shape.
Each product extension followed the same logic: enter with a no-fee or low-fee version of a product that incumbents overcharged for, acquire customers at near-zero marginal cost through the existing base, then gradually monetize through higher-margin products as the relationship deepened. The credit card customer became a NuConta customer became a personal loan customer became an investment customer became an insurance customer. The average revenue per active customer (ARPAC) — Nubank's most closely watched operating metric — rose from $4.70 in Q4 2021 to $11.40 by Q1 2025 in the mature Brazilian cohort.
That ARPAC trajectory is the entire story, compressed into a single number.
Pix and the Infrastructure Gift
In November 2020, Brazil's central bank launched Pix — an instant payment system that allowed free, real-time transfers between any bank accounts, 24 hours a day, seven days a week. The incumbents hated it. Pix cannibalized their fee income from wire transfers (TEDs and DOCs), which had generated billions annually. Within 18 months of launch, Pix was processing more transactions per day than all debit and credit card payments combined.
For Nubank, Pix was the equivalent of the U.S. interstate highway system for trucking companies — a public infrastructure investment that massively amplified the advantages of the most efficient operators. Nubank's app-first interface was perfectly suited to Pix. Sending money via Pix on Nubank was instant, intuitive, and required none of the branch visits or complex authentication protocols that incumbent banks imposed. Pix became the primary driver of NuConta adoption among lower-income Brazilians who had never had a bank account — because suddenly, you needed a bank account to receive Pix payments, and Nubank was the easiest account to open.
Pix is the best thing that has happened to financial inclusion in Brazil in decades. And it happens to be the best thing that has happened to Nubank.
The central bank's infrastructure investment solved Nubank's most expensive remaining problem: activating dormant accounts. Before Pix, a customer might download the app and receive a credit card but still maintain their primary banking relationship with Itaú or Bradesco. After Pix, that same customer began routing salary deposits, paying rent, splitting restaurant bills, and receiving government transfers through NuConta — because Pix made Nu the default payment rail. Monthly activity rates surged past 80%.
The Mexico Bet
In 2019, Nubank launched in Mexico — its first expansion outside Brazil, and the decision that will likely determine whether the company becomes a regional champion or merely a very large Brazilian bank.
Mexico presented a variation on the Brazilian thesis: five banks controlled roughly 70% of deposits; more than 60% of the adult population lacked a formal bank account; credit card penetration was under 30%; and the customer experience at incumbents — BBVA México, Banorte, Citibanamex, Santander — was widely despised. The regulatory environment was evolving favorably, with Mexico's 2018 fintech law creating a framework for digital-only banking operations.
The differences from Brazil were also significant. Mexico's credit bureau infrastructure was less developed. The population was more geographically dispersed and less smartphone-saturated than Brazil's urban centers. The competitive landscape included well-funded local fintechs — Stori, Klar, Fondeadora — and the looming presence of MercadoLibre's Mercado Pago, which had ambitions of its own.
Nubank entered Mexico with the same playbook: a no-annual-fee credit card, managed entirely through a mobile app, with an emphasis on customer experience and viral referral dynamics. The growth curve tracked Brazil's early trajectory — then accelerated. By Q1 2025, Nubank Mexico had surpassed 10 million customers, making it one of the fastest-growing financial services operations in the country's history. The company obtained a Mexican banking license in 2024, enabling deposit-taking and lending — the same regulatory unlocking that had transformed the Brazilian business.
The economics of the Mexico operation remained pre-profitable as of early 2025, with Nubank investing heavily in customer acquisition and credit underwriting. But the path to profitability was visible — and the total addressable market, in a country of 130 million people with catastrophically low financial inclusion, was arguably larger on a per-customer basis than Brazil's.
Colombia, where Nubank launched in 2020, followed a similar pattern on a smaller scale, reaching approximately 2.5 million customers by early 2025.
The Credit Machine
There is a version of the Nubank story that focuses entirely on customer acquisition, viral growth, and the romance of disrupting entrenched oligopolies. That version is incomplete. The real story — the one that explains the $49 billion market capitalization and the 30% ROE — is a credit underwriting story.
Nubank is, at its core, a lending institution. In Q1 2025, interest income on the loan portfolio constituted roughly 70% of total revenue. The credit card portfolio, personal loans, payroll-deductible loans, and the emerging auto loan and BNPL books are the engine. Everything else — interchange fees, investment product commissions, insurance premiums, marketplace revenue — is either a complement to the lending business or a customer acquisition vehicle that feeds the lending business.
The critical question, then, is whether Nubank can underwrite credit profitably at scale in a macroeconomic environment characterized by volatile interest rates, high baseline default rates, and a customer base that skews younger and lower-income than traditional bank portfolios. The answer, as of 2025, appears to be yes — but the evidence deserves scrutiny.
Nubank's non-performing loan (NPL) ratios have tracked at or slightly above industry averages, which is expected given the demographic composition of its portfolio. The 90-day NPL ratio for the Brazilian consumer loan book was approximately 6.5% in Q1 2025 — elevated by absolute standards but manageable within the context of Brazilian consumer lending, where the central bank's NPL data shows system-wide consumer loan delinquency rates consistently in the 5–7% range. What Nubank has demonstrated is the ability to manage credit losses while maintaining net interest margins (NIMs) in the high teens — a function of both pricing power (Brazilian consumer credit rates remain extraordinarily high by global standards) and the low cost of funding enabled by the deposit franchise.
The company's credit models are built on data that incumbents do not possess. Every Pix transaction, every bill payment, every savings deposit pattern feeds a proprietary credit-scoring engine that supplements Brazil's traditional credit bureau data (which is limited, especially for the unbanked population Nubank serves). This data advantage compounds over time: more customers generate more transaction data, which improves credit models, which reduces loss rates, which allows expansion into riskier segments that incumbents avoid, which generates more data.
We have spent ten years building a credit underwriting machine that gets better every single day. Every transaction our customers make teaches the system something new.
Whether this virtuous cycle holds during a severe economic downturn remains untested in the truly extreme case. Brazil experienced a sharp recession in 2015–2016 and a pandemic-driven contraction in 2020, but Nubank's loan book was relatively small during both episodes. The company's first real stress test at scale will come during the next Brazilian recession — and the outcome will determine whether the market re-rates Nubank as a durable financial franchise or punishes it as a growth lender with untested risk management.
The Profitability Inflection
For years, the bear case on Nubank was simple: it's buying growth with venture capital. The customers are real, the revenues are growing, but the company cannot make money because it underprices products to acquire customers and will eventually face the same credit losses that plague every consumer lender in Latin America.
The 2023–2024 period demolished this narrative with a thoroughness that surprised even the bulls.
Nubank turned net-income-positive on a quarterly basis in Q4 2022, posting $58 million in profit. By Q4 2023, quarterly net income had grown to $360 million. By Q1 2025, the figure was $557 million — representing a 47% year-over-year increase. Full-year 2024 net income exceeded $2 billion. The return on equity, which had been negative as recently as 2021, reached 30% on an annualized basis in Q1 2025.
To put this in context: Itaú Unibanco, Brazil's most profitable private-sector bank, has generated ROEs in the 18–22% range in recent years. Nubank, at 30%, was not merely competitive with the incumbents. It was more profitable than all of them.
The profitability was driven by three converging forces: ARPAC expansion (the deepening monetization of existing customers through additional products), operating leverage (the fixed cost base growing far more slowly than revenue), and credit quality management (NPL ratios stabilizing even as the loan book grew). The cost-to-income ratio declined from above 80% in 2020 to approximately 30% by early 2025 — a figure that would be exceptional for any financial institution in any market.
Key financial metrics, 2021–2025
| Metric | FY 2021 | FY 2022 | FY 2023 | FY 2024 | Q1 2025 (Ann.) |
|---|
| Revenue ($B) | 1.7 | 4.8 | 7.7 | 10.1 | ~11.5 |
| Net Income ($M) | (165) | 11 | 1,030 | 2,100 | ~2,230 |
| Customers (M) | 48 | 70 |
The stock responded accordingly. From its June 2022 low of $3.26, NU shares climbed to above $10 by the end of 2024. As of mid-2025, the stock trades near $15, representing a market capitalization of approximately $49 billion — still below the IPO day peak but well above the depths of the 2022 correction, and supported now by actual profits rather than hope.
The Culture of the Purple Card
Nubank's brand power in Brazil is difficult to overstate for observers outside Latin America. The purple card is an icon — a consumer product that achieved cultural penetration on the level of the iPhone or the Nike swoosh within Brazilian popular culture. Nubank is the most downloaded financial app in Brazil, the most mentioned bank on social media, and consistently ranks as one of the most admired brands in the country across all categories, not merely financial services.
This brand loyalty is not accidental. It was engineered through a set of deliberate cultural choices that deserve enumeration:
Radical transparency on fees. Every fee, every interest rate, every charge is displayed prominently in the app. In a market where incumbents buried costs in footnotes and charged "tariffs" that consumers could not parse, this transparency was revolutionary. Customers trusted Nubank not because it was cheap — though it often was — but because it did not feel like it was trying to trick them.
Customer service as theater. Nubank's customer service team — called "Xpeers" — was trained to resolve complaints with personality, empathy, and occasionally absurdist generosity. Stories circulated on social media of Nubank sending dog toys to a customer whose pet had chewed their card, or writing handwritten notes to customers going through difficult times. These anecdotes were not scalable in a literal sense, but they were infinitely scalable as marketing — each one becoming a social media post seen by thousands.
Internal culture. Inside the company, the culture reflected the engineering-first, customer-obsessive ethos of its founders. The workforce is heavily technical — more than 50% of employees are engineers or data scientists. Decision-making was decentralized through small, autonomous teams organized around specific product surfaces. The company maintained a flat hierarchy relative to Brazilian corporate norms, which tend toward rigid formality.
Cristina Junqueira, the co-founder who had left Itaú to join Vélez, became the public face of this culture in many ways — articulate, direct, and willing to criticize the incumbents with a specificity that carried the authority of someone who had seen their operations from the inside. Her presence signaled that Nubank was not merely an outsider throwing stones but a company built by people who understood exactly which stones to throw.
The Regulatory Tightrope
Operating a digital bank at Nubank's scale in Brazil requires navigating a regulatory environment that is simultaneously progressive and unpredictable. The Banco Central do Brasil (BCB) has, by global standards, been remarkably supportive of fintech innovation — Pix being the most dramatic example, but also through open banking mandates, sandbox programs, and the issuance of new license categories for payment institutions and digital banks.
But the relationship is not frictionless. Brazilian regulators have periodically signaled concerns about the concentration of consumer credit risk in fast-growing digital lenders. In 2022, the BCB implemented new capital requirements that specifically affected fintechs' ability to grow their credit card portfolios without commensurate increases in regulatory capital. Nubank's capital adequacy ratio has remained well above minimums — the company reported a Basel III capital ratio above 15% in Q1 2025 — but the regulatory environment demands continuous investment in compliance infrastructure and limits the pace at which the credit book can expand.
The political dimension is equally complex. Brazil's large banks are among the most politically influential institutions in the country, with deep relationships across the legislative and executive branches. Nubank has invested in government relations but remains, by definition, a disruptor in a market where the disrupted have enormous political capital. The risk of regulatory capture — where incumbents use their political influence to impose rules that disproportionately burden digital competitors — is not theoretical.
What the Machine Looks Like Now
As of mid-2025, Nubank operates across three countries with a total customer base exceeding 114 million. Brazil accounts for approximately 100 million of these customers, Mexico roughly 10 million, and Colombia 2.5 million. The product suite in Brazil encompasses credit cards, debit cards, digital accounts, personal loans, payroll loans, auto loans, BNPL, investment products (equities, fixed income, funds), life insurance, auto insurance, home insurance, a crypto trading platform, a marketplace for third-party financial products, and — as of 2024 — a nascent AI-powered personal finance assistant.
The company employs approximately 9,000 people, of whom more than half are based in Brazil and the remainder split between Mexico, Colombia, and satellite engineering offices. Revenue per employee exceeds $1.1 million annually — a figure more reminiscent of a technology company than a bank.
The competitive landscape has evolved. Traditional banks have launched their own digital platforms — Itaú's Iti, Bradesco's Next (later merged back into Bradesco), Banco do Brasil's various digital initiatives — with mixed results. MercadoLibre's Mercado Pago has emerged as a formidable competitor, particularly in payments and small-business lending, with a different strategic logic rooted in e-commerce adjacency. PicPay, Inter (Banco Inter), and C6 Bank compete for segments of the digital banking market. But none has matched Nubank's combination of scale, brand loyalty, and unit economics.
The question that now confronts Nubank is not whether the model works. The model works. The question is how large the model can become — whether ARPAC can continue to expand as the company moves into higher-margin products, whether Mexico and Colombia can replicate Brazil's economics, whether credit quality can be maintained through a full macroeconomic cycle, and whether the company can sustain a 30% ROE as the base grows and the easy gains from operating leverage diminish.
In Vélez's São Paulo headquarters — which is, naturally, branchless — there is a wall that tracks customer count in real time. The number ticks upward with the metronomic regularity of a population clock. Somewhere in the interior of Minas Gerais or the outskirts of Guadalajara, someone is downloading an app, uploading a selfie for identity verification, and receiving a purple card that will arrive in the mail within days.
The card costs Nubank roughly $5 to produce and ship. The customer it acquires will generate, on average, $137 in annual revenue within three years. The math is simple. The compounding is not.
