In 2025 Nubank earned $16.3B of revenue (+45%) with $2.9B of net income, closed the year with 131M customers and a record 33% ROE[1], and passed Bradesco to become Brazil's second-largest financial institution by customers[4]. Then, on record Q1'26 results, the stock fell up to ~10% and is down nearly 30% on the year[31].

The genuinely open question is not whether Nubank is a real franchise — it plainly is, with one of the lowest reported costs-to-serve in banking and a #1 NPS / fewest-complaints reading among its Brazilian peers. It is whether a continental-scale, unsecured-heavy lending book built almost entirely in the past few years can absorb a full Brazilian credit downturn while it also funds Mexico and a new US bank. The evidence cuts both ways on every question below. This study lays out both cases; the verdict is yours.

The climb that frames the debate

Annual revenue, US$B. FY2025 = $16.3B is the sourced focus; prior years are trend context. The speed of the climb is both the bull case and, for skeptics, the reason the credit book is so young and untested. Hover a year.

The decisive questions

Each links to the section that lays out the evidence on both sides.

How to read this

Eleven sections, each built the same way: a neutral synthesis, a two-sided case-for / case-against ledger, sourced data and charts, and dated facts. Brazilian (Portuguese) sources carry the domestic debate — analyst calls, consumer-protection rulings — alongside the English filings and press. Start with the question that interests you, or read in order from the Overview.

In twelve years Nubank went from one no-fee credit card to 135M customers across Brazil, Mexico and Colombia[30] — Brazil's second-largest financial institution by customers[4] — by attacking a concentrated, high-fee banking oligopoly with a free, app-only model. The growth is undisputed; the debate (the rest of this study) is about what happens to its young credit book in a downturn.

What it is

Nu Holdings Ltd. (NYSE: NU; Brazilian BDR ROXO34) is the holding company for Nubank, a digital bank founded in São Paulo in 2013[2]. It started with a single product — a no-fee, app-controlled purple Mastercard — and expanded into deposits (NuConta), personal and secured loans, investments, insurance and a marketplace, all inside one app[2]. It operates in three countries: Brazil (its core), Mexico and Colombia[1]. Co-founder David Vélez, a Stanford-trained former Sequoia investor, is CEO; co-founder Cristina Junqueira now leads the nascent US business[46].

The scale today

Total customers, millions, year-end (plus the Q1'26 reading). By Q1'26 Nubank served 115M in Brazil, 15M in Mexico and 5M in Colombia[30]. In Brazil that is roughly 62% of the adult population[1]. Hover a point.

How it got here

Latin America's ~300M+ unbanked or under-banked adults[5]were the opening Nubank ran through. But in its core market that opening is closing: Brazil's five largest neobanks already reach ~80% of the population[7], so future growth depends more on monetizing existing customers and on Mexico/Colombia than on signing up the unbanked.

The size of the prize

Latin America is one of the most dynamic fintech regions in the world, with 300M+ unbanked or under-banked adults, historically high bank fees, a young median population and smartphone penetration above 72%[5]. Third-party estimates put the LatAm neobank / digital-banking market at roughly $18.4B in 2025, growing ~19.6% a year toward ~$98.7B by 2034[5] (a market-research projection — treat as directional, not precise). Brazil is the single largest national market at ~38.7% of regional revenue, Mexico ~22.4%, Colombia ~11.2%[6].

Where the money is — and the regulation

Neobank economics in Brazil are shaped by two forces. First, Pix, the central bank's instant- payment system, made transfers free for everyone — eroding a former neobank edge and pushing monetization toward credit. Second, the policy rate (Selic) is high, which lifts the yield on lending and on Nubank's bond-invested deposits, but also raises funding cost and the macro risk of delinquency. The IMF noted Brazilian credit growth stayed strong through 2025 despite high rates[8] — a tailwind for origination and a warning sign on leverage at once.

The maturation question

The bull framing is penetration: hundreds of millions still under-served, and Nubank monetizing each customer lightly (low ARPAC) with room to grow. The bear framing is saturation: the five biggest Brazilian neobanks — Nubank, Mercado Pago, PicPay, C6 and Inter — already account for ~75% of active clients, a base of ~184M people, ~80% of Brazilians[7]. Both can be true: the account land-grab is largely done in Brazil, while the credit and cross-sell opportunity — and the Mexico/Colombia land-grab — is not.

Nubank serves a customer for ~$0.80 a month, acquires them for ~$5 (mostly organic), and earns a 33% ROE[9][10]. That cost advantage is real and structural. But ~83% of revenue is interest- and credit-linked[11], so the same engine that makes the margins makes them sensitive to Brazil's rate cycle and to how its loans perform.

How it earns

Revenue comes from four main places: interest on credit-card revolving balances and personal loans; interchange on card transactions; financial income on deposits invested largely in government bonds; and a growing tail of fees from investments, insurance and the marketplace. By 2024, interest income and fair-value gains were ~83% of total revenue[11] — net interest income (NII) is the engine. NII reached an all-time-high $2.8B in Q4'25 and $3.25B in Q1'26[44].

Approximate revenue mix, 2024. The model is overwhelmingly interest-driven.

Why the unit economics work

The cost side is where Nubank's reported metrics most diverge from incumbents'. Monthly average cost to serve was ~$0.80 per active customer in Q4'25[9]; the company designs products, automates service and runs risk models in-house, so it scales without an incumbent's branch-and-headcount base[23]. Customer acquisition cost is ~$5 with ~80% of customers arriving organically and contribution-margin payback under 12 months[10] (these last figures are analyst estimates, not company-disclosed). The result: deposits of $41.9B at ~87% of interbank rates[12], an 83% monthly activity rate, and ARPAC rising to ~$15-16[9][30] — with mature cohorts well above that.

The catch

Because the model monetizes through lending, two macro variables drive it. A high Selic widens the spread on credit and lifts bond income, flattering margins; but Brazilian coverage notes a Selic cut is double-edged — it lowers funding cost yet also shrinks floating income on deposits, offsetting part of the gain[52]. And growth in NII has increasingly come from more lending to riskiersegments — XP characterized the story as "the pains of growth," flagging portfolio expansion, delinquency and model-dependence as attention-points[48]. So the "magic margin" that bulls cite and bears question is partly a structural cost edge and partly a bet on Brazil's rate and credit environment.

FY2025 ended with delinquency falling — 15-90 NPLs down to 4.1%[13]. Then Q1'26 reversed: short-term NPLs jumped to 5.0%, credit-card Stage 2 hit a company-record 11.5%, and risk-adjusted NIM fell ~100bps to 9.5%[14][16]. Whether that is a young book finally seasoning into trouble, or conservative models flagging risk early, is the single most contested question about Nubank — and analysts genuinely disagree.

The leading indicator turned

Nubank's most-watched metric is its short-term (15-90 day) NPL ratio — the leading edge of delinquency. It improved through 2025, falling to 4.1% in Q4'25 with 90+ NPLs also down to 6.6%[13]. In Q1'26 it rose ~89bps to 5.0%, while 90+ NPLs held at 6.5%[14]. Hover a point.

The profit gauge compressed

Risk-adjusted NIM — net interest margin after credit losses — is the number that turns lending into profit. It fell from 11.5% (Q3'25) to 10.5% (Q4'25) to 9.5% (Q1'26)[14], pressured by both rising provisions and high deposit-acquisition costs in Mexico.

Why provisions surged

The proximate trigger for the 2026 sell-off was provisioning. The credit-loss allowance reached $1.79B with total coverage at 16.2% of the portfolio — about 2.5x the 90+ delinquency balance[14]. Provisions rose ~75.7% YoY to ~$1.7B on a 40% larger book[31]. Management's explanation: new exposure is tilting hard toward the riskiest products — credit cards and unsecured loans were ~98%of Q1'26 net new exposure, up from 88% a year earlier[19] — which mechanically lifts marginal provisioning even if no loan has gone bad yet.

The two readings of "Stage 2"

Bears point to Stage 2loans — performing loans flagged as higher-risk. In Q1'26 the credit-card Stage 2 ratio hit 11.5%, the highest in Nubank's history, and personal-loan Stage 2 reached 15.4%[16]. JP Morgan's reading is the opposite: the jump came not from an actual default spike but from more conservative internal probability-of-default models — Nubank identifying risk proactively, before borrowers miss payments[17]. Citi is less sanguine, warning that 2026 will be "a demanding year for asset quality" with cost of risk rising[18]. Brazilian research house Nord framed the moment plainly as an important "resistance test" for fintechs like Nubank as the cycle turns[51].

Nubank added ~11.2MBrazilian customers in 2025, the most of any institution, ahead of Mercado Pago's ~8.7M[20]. But incumbents defend deeper, higher-monetization relationships[21], Mercado Pago is following Nubank into Mexico[45], and Pix has commoditized the free-transfer hook — so leadership in reach does not settle leadership in profit per customer.

Who Nubank competes with

Two fronts. Against incumbents — Itaú Unibanco, Bradesco, Santander, state-owned Caixa and Banco do Brasil — Nubank wins on cost, brand and reach; Bradesco, Santander and Itaú each lost ~500k active users to the neobanks recently[21]. Against fellow challengers — Mercado Pago, PicPay, C6, Inter — the contest is for the same young, digitally-native customers; Nubank and Mercado Pago are pulling ahead while C6 and Inter lose relative share[20].

Net Brazilian customer adds in 2025, millions. Nubank and Mercado Pago led.

Positioning: reach vs. proven credit

Vertical axis = proven through-cycle credit resilience; horizontal = scale/reach. Nubank sits far right on scale but mid-vertical: its credit book is enormous but young. Incumbents sit higher on seasoned underwriting; fellow neobanks share Nubank's youth-of-book risk. Click a company.

Five forces

Click a force for the evidence behind its pressure rating.

The stated moat is a tech-driven low cost-to-serve plus a cross-sell flywheel: acquire cheaply, deepen with more products, fund with cheap deposits[23]. Management frames 2026 as an "inflection year" — win Brazil and Mexico, build secured/payroll lending, lay foundations for the US, and lean on AI[24]. The revealed tension: the safer-credit pivot keeps hitting regulatory snags while growth concentrates in riskier loans.

Stated strategy

Management describes three 2026 pillars: win core markets (Brazil and Mexico get most of the ~R$45B / ~$8.2B Brazil capex), strengthen foundations for international (US) expansion, and build secured and payroll lending[24]. The throughline is to keep the low-cost engine running — Nubank designs products, automates service and runs risk models in-house, so it grows without an incumbent's fixed-cost base[23] — while raising ARPAC through cross-sell, since its monetization per customer is still well below incumbents.

Revealed strategy

Two things the numbers reveal. First, the safer-credit pivot is harder than planned: changes to FGTS (severance-fund) rules cut new FGTS-backed secured-loan originations by more than half, only partly offset by public-payroll lending[25] — so the book has stayed unsecured-heavy (~98% of new exposure, per the Credit-Cycle section)[19]. Second, AI is now a real cost lever: Nubank reports near-100% internal AI adoption, ~+50% YoY engineering throughput, and proprietary models in real-time credit decisioning in Brazil and Mexico[40].

Where the moat could erode

The cost-and-data advantage is real, but it is not unique to Nubank — Mercado Pago and others run similar low-cost, data-driven models. Pix removed a payments differentiator. And the deepest moat question is underwriting: a low cost-to-serve means little if credit losses run ahead of the spread. That is why the safer-lending pivot and the AI-underwriting claims matter so much to the thesis — and why the FGTS setback is more than a footnote.

Mexico and Colombia are pitched as a "shield" against Brazil concentration[27]. Mexico hit its first quarter of IFRS profitability in Q1'26 and won a banking license[29][28]; Colombia deposits grew 841% YoY[27]. The cost: a ~$4.2B Mexico bet whose high deposit-acquisition cost is already dragging the consolidated margin[29].

The concentration problem

Nubank is overwhelmingly a Brazilian company: ~113-115M of its ~131-135M customers are in Brazil[1][30], and the bulk of revenue and credit risk sits there. That is the case for international diversification — geographic spread "helps soften the cycle when one market declines," as the company frames it[27], and as Brazilian coverage echoes in calling Mexico and Colombia a "shield" that reduces Brazil exposure as customers and deposits grow[53].

Customers by country, year-end 2025 (millions). Brazil dominates.

Mexico: inflection, at a price

Mexico is the most important second market. Nubank surpassed 12M customers and $6.7B in deposits[27], won CNBV banking-license approval in April 2025 (the first Sofipo to do so), unlocking higher deposit limits, payroll accounts and 16x deposit insurance[28], and reached its first quarter of IFRS profitability in Q1'26 with ARPAC nearly doubling[29]. The cost: bears flag the ~$4.2B Mexico investment as a high-stakes bet, and high deposit-acquisition costs in a market dominated by BBVA and Banorte weighed on the consolidated risk-adjusted NIM[29]. The open question is converting Mexican savers into borrowers without that becoming an earnings drag.

Colombia: small but fast

Colombia is earlier and smaller — ~3.4-5M customers and $2.1B deposits — but growing very fast after the mid-2024 launch of the Cuenta Nu savings account nearly tripled the base, with deposits up 841% YoY in constant currency[27][43].

The US: a different league

In January 2026 Nubank won conditional OCC approval to establish a US national bank, Nubank N.A., led by co-founder Cristina Junqueira (who relocated to Miami) and chaired by former Brazilian central-bank head Roberto Campos Neto[26][46]. It must fully capitalize within 12 months and open within 18. The US is a far larger market — and a far more competitive one — which is why analysts are split on whether it is valuable optionality or a distraction from the Brazilian credit phase[37].

FY2025: revenue $16.3B (+45%), net income $2.9B, ROE 33%[1]. Q1'26: a record $5B revenue and $871M profit (+41% YoY)[30]. Yet NU fell up to ~10% on the print and is down nearly 30%on the year, to a ~$58B market cap, with BofA cutting it to "sell"[31][32]. The market is repricing a hyper-growth story as a bank.

The growth is real

Annual revenue, US$B — FY2025 $16.3B (+45%) is sourced; prior years are trend context. Profit grew from $2.0B (2024) to $2.9B (2025)[1], and Q1'26 set fresh records on revenue and quarterly profit[30].

Why the stock fell anyway

Three things converged in 2026. First, provisions: up ~75.7% YoY to ~$1.7B on a 40%-larger book, the rise the market read as "the cracks"[31]. Second, margin: risk-adjusted NIM fell ~100bps QoQ (see Credit-Cycle). Third, the CFO departure: Guilherme Lago's June 1 exit, amid heavy senior turnover, raised governance and execution questions just as credit tightened and the US push began[37]. The reaction was sharp: BofA downgraded NU from neutral to sell on June 2 and cut its target from $16 to $10[32].

The valuation reset

After the drawdown NU trades at a trailing P/E ~22x and forward P/E ~15x[34] — still a premium to LatAm incumbents, but sharply reset. Sell-side targets diverge unusually widely: a consensus near $20 (range ~$16.9-$22) against BofA's $10[33][32]. That spread is the debate priced into the stock: bulls see a company still compounding 40%+ at a reset multiple; bears see decelerating margins and an untested credit book.

The debate, in one chart: where 22 covering analysts put NU's fair value (US$ per share, early June 2026), against the ~$13 the stock actually traded at. The spread from BofA's $10 "sell" to the $22 high is the bull/bear divergence the rest of this study argues out.

Consensus mean ($20.48), range low ($16.90) and high ($22.00) across 22 analysts and the ~$13 trading level per VCP Scanner[33]; BofA's $10 "sell" target per NeoFeed[32]. The unusually wide spread is the credit-cycle debate priced into the stock.

Nubank pairs incumbent-beating profitability (a 33% ROE) with neobank growth (45% revenue) and the lowest cost-to-serve in the group[35][9] — the bull case in one line. The bear case is the other columns: the most credit-levered, least cycle-tested book, at a premium multiple.

Profitability gap

Return on equity, %. Nubank's Q4'25 ROE of 33%[35]sits above large Brazilian incumbents (illustrative ranges; only Nubank's figure is point-sourced). The gap is the heart of the bull thesis — and what a credit downturn would most test.

Side by side

Peer ROE ranges are illustrative context from general coverage; only Nubank's 33% Q4'25 figure is point-sourced here[35]. Scale/positioning per cited Brazilian rankings[4][20].

Valuation

Post-2026 drawdown, NU trades at ~22x trailing / ~15x forward earnings at a ~$58B cap[34][31]— a premium to incumbents that reflects faster growth and higher ROE, but one the market has compressed sharply as it weighs the credit-cycle risk against the growth. Even when Nubank's ROE beat Itaú's, several Brazilian analysts held neutral ratings, arguing the stock already prices the growth and the risk-reward is unfavorable while the market pays incumbents a premium for predictability[49].

With customers, Nubank is the most-trusted name in its market — #1 on NPS and far fewer Central Bank complaints per million users than incumbents[38]. Inside the company, 2025-26 brought a striking run of senior exits, capped by the June 2026 CFO departure[36][37] — a governance question the market took seriously.

Customer sentiment: the brand moat

Nubank's strongest soft asset is trust. It ranks first on NPS among Brazilian peers with an ~86 customer-satisfaction score, and recorded only 1,222Central Bank complaints per million users in Q4'24 — versus Itaú's 4,127 and Santander's 3,365[38]. That gap underpins the ~$5, ~80%-organic acquisition engine. It is not spotless: consumer-protection bodies Procon-SP and Procon Carioca have notified Nubank over outages and fraud handling, and in one fraud case Procon found it had objective responsibility for failing to verify transaction regularity[39] — a reminder that scale brings service strain.

Leadership: founders back in the cockpit, bench thinning

Founder David Vélez resumed direct operating control in March 2025, with Brazil CEO Lívia Chanes and co-founder Cristina Junqueira reporting to him[36]. Around that, the senior bench thinned: 2025 saw the exits of the CPO, IR officer, chief legal officer and — most notably — President/COO Youssef Lahrech, who had built the high-growth credit portfolio[36]. Then CFO Guilherme Lago departed June 1, 2026 after seven years, replaced by ex-Visa/Capital One executive Rob Livingston[37]. JP Morgan counted four senior departures in 2025 alone and called the CFO change "negative"[37].

SWOT

An even-handed summary; each item is sourced in the sections above.

Method

This study was assembled from a fan-out of web research conducted during a single run, prioritizing primary sources — Nu Holdings' FY2025 (Q4'25) and Q1'26 results, press releases and earnings-call transcripts — then reputable Brazilian and international press and named analyst notes. As Nubank is a Brazilian company, a substantial share of the research was done in Portuguese (34% of cited sources), drawing on domestic financial media (Seu Dinheiro, NeoFeed, Bloomberg Línea Brasil, Finsiders, Poder360), analyst commentary (JP Morgan, Citi, BofA, BTG, Safra) and consumer-protection rulings (Procon). Every cited URL was fetched and read during the run; figures were reconciled against the company's own disclosures where possible.

Frameworks used

Pyramid-Principle executive summary; a market/TAM structure view; a unit-economics and revenue-mix breakdown; a dedicated credit-cycle analysis (the central debate); Porter's Five Forces; a 2×2 positioning map (scale vs. proven credit); a peer-comparables table; and an even-handed SWOT. Frameworks were applied only where data supported a real conclusion.

Source mix

Cross-checked at build time by an automated link checker. A few primary sources (SEC EDGAR filings and some publishers) bot-wall automated fetchers; those were verified against the equivalent Nu Holdings press release or transcript. See Methodology & Limits.