Nu Holdings Ltd. (NYSE: NU) — The World’s Best-Run Bank, Marked Down on a Credit Scare and a Tax-Rate Asterisk

Report date: 2026-06-13 Approach: Competitive-advantage and capital-allocation fundamental analysis (Greenwald, Competition Demystified; Chancellor/Marathon, Capital Returns) Price reference: $12.19 (NYSE close, 2026-06-12) · 52-wk range $11.20–$18.98 · ~4.91B diluted shares · market cap ~$60B · total equity $12.6B (Q1-2026)

The analysis below (Sections 1–15) is presented as independent fundamental research and contains no buy/sell recommendation and no price target. The single exception is the author’s-opinion block immediately below.

⚡ Claude’s Take

This block is the author’s own subjective opinion and general information only — not investment advice. The rest of this article is presented as position-free analysis.

Verdict: BUY / accumulate-on-weakness — a genuine quality-compounder that has, for the first time since its IPO, derated to a defensible price. Conviction: medium-high. Entry zone: attractive $10–13 (roughly 19–24x reported / ~22–28x tax-normalized forward earnings, ~4.0–4.8x book against a ~30% ROE); genuinely cheap below ~$10; I would stop adding above ~$16 (~30x normalized) pending proof the US/Mexico optionality is converting.

Nubank is, on the numbers, one of the best-run banks on the planet: ~135 million customers, a ~30% ROE, a ~75% five-year revenue CAGR, a structurally low cost base (efficiency ratio in the high-teens vs. incumbents’ 40%+), and a deposit franchise ($42B) that funds its loan book at a fraction of peers’ cost. It has compounded earnings while shrinking its tax-and-dilution drag. And after two quarters the sell-side branded “disappointing,” the stock now trades at the 1.4th percentile of its own 10-year P/E history — the market is pricing a cyclical wobble (first-quarter seasonal provisioning, a 100bp sequential dip in risk-adjusted NIM) as if it were a structural break. It is not: 90+ NPLs have been range-bound at 6–7% for three years and coverage is rising. This is the classic setup where a great business gets cheap because growth merely decelerated from extraordinary to merely excellent.

The reason this is not a slam-dunk — and why I size it as medium-high, not high, conviction — is honest earnings quality. The Q1-2026 headline of $871M net income was helped by an 8.7% IFRS effective tax rate; normalize to the ~30% rate management itself calls “economically meaningful” and the multiple is closer to ~26x than the screen-cheap ~19x. The story also leans on a founder who controls 74% of the votes (a real governance discount), a Brazilian consumer-credit cycle that is mid-to-late, and a US expansion that is still a call option, not a business. Framing: quality-compounder-at-a-reasonable-price (GARP), with a contrarian overlay — you are buying a structurally advantaged franchise during a sentiment air-pocket. Bull trigger: risk-adjusted NIM re-expands toward 10.5%+ in 2H-2026 while Mexico keeps inflecting (confirming the credit scare was seasonal). Bear trigger: 90+ NPLs break above ~7.5% and the tax rate normalizes simultaneously — that would expose the earnings as both cyclically and structurally inflated, and 4.8x book would not survive it. Tag: “Latin America’s best bank, finally on sale — if you can stomach the tax-line asterisk and a founder who owns the ballot box.”

1. Executive Summary

Nu Holdings Ltd. (“Nubank”) is the largest digital bank in Latin America and, by most measures, the most successful consumer-fintech franchise ever built outside China. From a standing start in 2013, it has reached ~135 million customers (Q1-2026) — ~115M in Brazil (the largest private financial institution in the country by customers, with ~57% of the adult population), ~15M in Mexico (third-largest FI), and ~5M in Colombia. It monetizes that base through a credit-card-led, increasingly diversified product shelf (cards, unsecured and secured lending, deposits, investments, insurance, marketplace), at a monthly average revenue per active customer (“ARPAC”) of ~$16 and an industry-leading efficiency ratio.

The financial record is exceptional. Revenue compounded from $1.7B (FY2021) to $15.8B (FY2025), a ~75% CAGR; the company turned IFRS-profitable in FY2023 and earned $2.87B of net income in FY2025 at a ~30% ROE — a return on equity that exceeds JPMorgan, the best-run US money-center bank. Q1-2026 set fresh records: $5.0B revenue, $3.25B net interest income, a 21.1% NIM, and $871M net income. The moat is a genuine, financially-anchored low-cost-producer advantage: a cloud-native, branchless stack lets Nubank serve mass-market and underbanked customers profitably at a cost-to-serve (~$0.80/active customer/month) that has stayed flat while ARPAC tripled — the textbook signature of operating leverage on a scale platform, reinforced by deposit-funding economics, customer captivity (primary-account principality, sector-leading NPS) and a data/AI underwriting flywheel.

Yet the stock sits near 52-week lows, having derated to the cheap end of its own history. Three things explain the gap between the business and the tape, and they are the crux of the analysis. First, credit normalization: Q1-2026 saw seasonal provisioning lift the credit-loss build and compress risk-adjusted NIM by 100bps (to 9.5%), which the sell-side (BofA, UBS) read as a “second consecutive disappointing quarter” and met with estimate and price-target cuts. Second, earnings quality: the Q1 IFRS effective tax rate was just 8.7%, materially flattering headline net income versus the ~30–35% “managerial” rate management itself regards as economically meaningful; cash taxes paid ($1.53B in Q1) dwarf the booked tax expense ($83M). Third, structural overhangs: a controlled-company governance structure (founder David Vélez holds ~74% of votes), an intensely contested competitive arena (Mercado Pago, incumbent digital arms, C6, Inter, PicPay), and a Brazilian consumer-credit cycle that is mid-to-late.

The right lens is embedded expectations. At ~$12, NU trades at ~19x reported / ~22–26x tax-normalized forward earnings and ~4.7x book. For a franchise growing earnings 40%+ at a 30% ROE, that is not demanding — provided the ROE is durable and the credit cycle behaves. The bear case is that the ROE is part-flattered by an unsustainable tax rate and a benign credit window that is closing. The bull case is that a structurally advantaged, deposit-funded, AI-accelerating compounder with a decade-plus runway in Brazil and a second “Nubank” inflecting in Mexico is being handed to patient capital at a sentiment trough. This memo argues the business is real and the moat durable; the legitimate debate is about price for cyclicality and earnings quality, not about business quality.

2. Business Overview

What Nubank is. Nu Holdings is a Cayman Islands-incorporated holding company (NYSE: NU; also a Brazilian Depositary Receipt, “NU Units”, on B3) that operates regulated financial subsidiaries in Brazil (Nu Pagamentos / Nu Financeira), Mexico (Nu México, converting to a full multiple-banking license) and Colombia (Nu Colombia), plus emerging operations in the US. It is a digital-only, app-native bank — no branches — built on a proprietary cloud core-banking stack.

How it makes money. Nubank’s revenue is dominated by net interest income (NII) from its credit portfolio and, increasingly, by fee and commission income:

• Interest income & gains, net was $13.4B in FY2025 (85% of revenue) and a record $3.25B in Q1-2026 — generated by credit-card balances, unsecured personal loans, and secured loans (payroll, FGTS, collateralized).
• Fee & commission income was $2.34B in FY2025 (15% of revenue) — interchange, marketplace (“Nu Shopping”), insurance, investments (NuInvest), and float/transactional fees.

Management frames the economics as a flywheel: a growing, engaged customer base × rising ARPAC × a flat, scalable cost base = compounding earnings. ARPAC has expanded sequentially every quarter the company has reported, from ~$4.50/month (FY2021) to ~$13.30 (FY2025) to ~$16 (Q1-2026), while cost-to-serve held at ~$0.80/active customer/month — an ARPAC-to-cost ratio that has widened from ~5.6x to ~16.6x.

Segmentation. Nubank reports geographically (Brazil/Mexico/Colombia) and by product. Brazil is the profit engine (the vast majority of revenue and essentially all of group profit today); Mexico reached its first quarter of IFRS profitability in Q1-2026, ahead of internal plan; Colombia remains an investment market. The product mix is anchored by the credit card (the original wedge product and still the largest receivable, ~$20.2B net at Q1-2026), with unsecured lending ($10B portfolio, growing 53% YoY FX-neutral) the fastest-growing line and secured lending (~8% of the book) the strategic diversifier. Deposits ($42.4B) and an expanding suite of investments, insurance, and a nascent SME franchise (~5M small-business customers built at near-zero acquisition cost) round out the shelf.

Recurring vs. non-recurring. The revenue base is highly recurring and granular: tens of millions of monthly-active customers transacting daily, an 83% consolidated monthly activity rate, and a deposit franchise that re-prices but rarely leaves. There is no project/lumpy revenue. The principal variability is not churn — it is the credit cycle (provisioning) and interest rates (Brazil’s Selic, currently 14.50%), both of which swing NII and the cost of risk.

Verdict. A genuine, recurring-revenue consumer bank with a coherent, demonstrated monetization model — not a pre-profit growth story. The business is straightforward to understand at the unit level (customers × ARPAC × cost-to-serve, minus cost of risk and funding), which is itself a point in its favor.

3. Industry Dynamics

The Brazilian banking pool is large, profitable, and historically oligopolistic — which is exactly why a low-cost disruptor can thrive. Brazil’s banking system was, for decades, dominated by five incumbents — Itaú Unibanco, Bradesco, Santander Brasil, Banco do Brasil, and Caixa Econômica Federal — operating high-cost branch networks and earning some of the widest banking spreads in the world. Management sizes the addressable profit pool of the products Nubank serves today at >$100B of annual gross profit in Brazil alone, of which Nubank captures ~7% despite already being the largest private FI by customers. That gap — large customer share, small profit share — is the core domestic growth thesis: monetization, not customer acquisition, is the runway.

The structural tailwind is profit-pool redistribution plus pie expansion. Two dynamics operate simultaneously. (1) Share is shifting from high-cost-branch incumbents to low-cost digital challengers — a redistribution Nubank is the prime beneficiary of. (2) The formal-banking pie is growing as previously underbanked populations are brought into the system (acutely so in Mexico, where less than half of adults hold a formal credit product and cash still dominates). Management sizes the Mexican served-profit pool at >$40B of annual gross profit, growing faster than most major banking markets, with Nubank’s share still below 1% — “where Brazil was a decade ago.”

The regulatory backdrop is double-edged. Brazil’s central bank (BCB) has pursued an aggressively pro-competition, pro-consumer agenda that has helped challengers attack the incumbents but is now also commoditizing parts of the value chain:

• PIX, the BCB’s free instant-payment rail (launched 2020), now represents roughly half of all Brazilian financial transactions and has overtaken combined debit + credit card volume. This is unambiguously negative for card-acquiring economics (the StoneCo/PagSeguro problem) but more neutral-to-positive for a deposit-and-credit bank like Nubank, for which PIX is a cheap engagement and transactional rail rather than a monetized product under attack.
• Open Finance lowers switching friction (a modest negative for incumbent captivity, a positive for a challenger winning share).
• Consumer-credit and pricing interventions: the income-tax exemption raised to ~R$5,000/month (progressive to ~R$7,350), effective 2026, removes 15–16M Brazilians from the tax rolls and is a tailwind to the disposable income and debt-service capacity of Nubank’s core low-income customer. Desenrola 2.0 (household-debt renegotiation, effective May 2026) is double-edged — it cleans up future credit quality but signals current household-debt stress. There is also live discussion of pricing caps on payroll lending, which Nubank (a deliberately slow, low-priced entrant there) is positioned to weather better than aggressive competitors.

Interest-rate sensitivity. The entire Brazilian financial complex is geared to the Selic policy rate (14.50%, with the BCB having begun easing in 2026 after a tightening cycle that peaked at 15%). High rates widen NII and float income but pressure consumer debt-service and credit quality; an easing cycle is, on balance, supportive for credit quality and volume but compresses some spread. Nubank’s 21.1% NIM is rate-geared.

Verdict: structurally attractive — and a better pool than the one Brazilian payments/acquiring peers occupy. Applying the Marathon capital-cycle lens: the acquiring/payments sub-sector is a capital-cycle breakdown (the state, via PIX, is deliberately compressing the profit pool — a structurally deteriorating industry, the StoneCo/PagSeguro predicament). The consumer-banking pool Nubank occupies is healthier: a genuine redistribution from inefficient incumbents plus an expanding underbanked pie. The caveat is competitive intensity (below) and the position of consumer credit in the capital cycle — mid-to-late boom, with abundant fintech capital having chased the Brazilian consumer for years, which is the textbook condition for mean-reverting credit losses. Net: a good industry for the lowest-cost operator, with the cycle as the swing risk.

4. Competitive Position

The moat — named. In Greenwald’s taxonomy, Nubank’s primary competitive advantage is a low-cost-producer / economies-of-scale advantage, reinforced by customer captivity and a proprietary-data/technology edge. The mechanism is concrete and shows up in the financials:

• Cost structure. A digital-only, internally-built core-banking stack lets Nubank operate at an efficiency ratio in the high-teens-to-~20% versus incumbents’ 40%+. Management’s “20x–30x more efficient than incumbents” claim, while a round number, resolves to a real and measurable gap: cost-to-serve of ~$0.80/active customer/month against legacy banks carrying physical branch networks, large headcounts, and legacy IT. This is what lets Nubank profitably serve thin-margin, mass-market and underbanked customers the incumbents cannot serve at a profit — the definition of a cost-advantage moat.
• Scale + deposit funding. 135M customers and a $42.4B deposit base (gathered directly, no brokers, at ~88% of the interbank rate) fund the credit book cheaply — deposits substantially exceed the loan book (loan/deposit ratio ~74%), so Nubank carries excess low-cost funding. Scale also amortizes the fixed technology and risk infrastructure across an enormous base.
• Customer captivity. Primary-banking-relationship principality, an 83% monthly activity rate, sector-leading Net Promoter Scores, and a brand that management (credibly) describes as creating “fans, not customers.” Switching costs in primary banking — direct deposit, bill-pay, credit history, embedded card-on-file — are real, if not absolute.
• Data/AI flywheel. 135M customers transacting daily generate one of the largest, cleanest proprietary financial datasets in the region. Nubank’s “nuFormer” proprietary models are in production for credit-card decisioning (Brazil + Mexico) and unsecured lending (Brazil), enabling real-time, NPV-based individual loan pricing in under one second. This is a genuine underwriting edge — if it proves out through a full down-cycle.

Pressure-testing the moat. The honest assessment is that the cost advantage is real but increasingly shared, not unique. Nubank is the lowest-cost operator versus the legacy branch banks, but it is no longer the only low-cost digital operator: Mercado Pago, Banco Inter, C6 Bank, PicPay, and the incumbents’ own digital arms (Itaú’s iti, etc.) all run digital-cost structures. The data/credit-underwriting advantage is the same claim Mercado Pago makes, and Greenwald’s warning applies — market growth is the enemy of a scale-based moat, because a fast-growing, contestable market lets sub-scale entrants reach minimum efficient scale. The credit-underwriting edge, the linchpin of the bull case, is unproven through a severe Brazilian down-cycle; rising NPLs are its first real test.

Direct competitive comparison. Against the Brazilian payments peers (StoneCo, PagSeguro), Nubank occupies a structurally superior pool and is winning the customer-growth race (BCB client rankings show Nubank and Mercado Pago leading net adds while StoneCo declined). Against Mercado Pago — its most formidable competitor — Nubank is larger in pure consumer banking (135M vs. Mercado Pago’s ~71M Brazilian MAU and $14.6B credit book) but Mercado Pago is gaining acquiring share fast and is investing heavily in Mexico ($4.6B planned 2026). Mexico is therefore a contested land-grab, not an uncontested runway. Against the incumbents, Nubank’s structural cost edge is decisive in mass-market consumer; the incumbents retain advantages in corporate/SME relationships, payroll-anchored data, and wealthy-segment wealth management — areas Nubank is now attacking (high-income “Ultraviolet”/UV cards, SME, private payroll).

Verdict: a real, wide-ish low-cost-producer + captivity moat — materially more durable than the narrow distribution niche of StoneCo/PagSeguro, but contested and not impregnable. The moat is genuine because it is anchored to a financial outcome (the efficiency ratio and cost-to-serve) that would deteriorate without it. The risk is that the digital-cost advantage diffuses across a crowded challenger cohort and the data edge fails its first cyclical stress test.

5. Growth History and Forward Opportunities

History — extraordinary, and high-quality. Nubank has compounded across every dimension that matters:

Revenue grew at a ~75% CAGR over five years; the business turned profitable in FY2023 and tripled net income over the subsequent two years. Critically, the growth is organic (customer acquisition, cross-sell, and ARPAC expansion — not acquisitions; M&A has been small tuck-ins). The flat cost-to-serve against a tripling ARPAC is the single most important growth-quality fact: this is genuine operating leverage on a fixed-cost platform, not growth bought with proportional cost.

Forward opportunities — four stacked vectors:

1. Brazil monetization (the base case). With ~7% of a >$100B profit pool and only ~8% market share in unsecured lending (but ~25–30% of new monthly originations), Nubank is taking disproportionate share. Deepening engagement (ARPAC), the high-income push (2 of 5 high-income Brazilians are already customers; high-income purchase-volume growing 40%+), secured lending (early innings — public/private payroll, a deliberately slow, disciplined entry), and the SME franchise (5M customers at near-zero CAC) all extend the domestic runway. Management calls Brazil “the first minute of the first half.”
2. Mexico (the second Nubank). 15M customers (3rd-largest FI), ARPAC nearly doubled, first IFRS-profitable quarter reached ahead of plan, full banking license advancing. Sub-1% of a >$40B profit pool. This is the highest-conviction expansion vector.
3. Colombia. ~5M customers, still an investment market — optionality.
4. AI-native banking + the US “call option.” Management’s three-phase AI transformation (assistance → workflow reinvention → AI-native bank) is operationalized (engineering throughput +50% YoY, AI Private Banker serving 15M+ MAUs, nuFormer in production). The US entry (OCC conditional national-bank charter, Jan 2026, led by co-founder Cristina Junqueira) is explicitly a bounded-downside call option: management caps the US OpEx drag at <100bps of the consolidated efficiency ratio in each of 2026 and 2027, inside the ~20% efficiency envelope. If product-market fit fails, the cost is small and absorbable; if it works, it is “a second Nu” in the world’s largest retail-banking market. The asymmetry is genuine — but the upside is unproven and US digital banking is a graveyard of failed challengers.

Verdict: high-quality growth. It is organic, monetization-driven, operating-leverage-fueled, and supported by a multi-vector runway with a decade-plus horizon. The honest qualifier is that the law of large numbers is beginning to bite — FX-neutral growth rates are decelerating from the triple-digits of 2022–23 toward the 30–40% range — and the credit-led nature of the growth means each incremental dollar of exposure carries upfront provisioning that depresses near-term reported profit (the Q1-2026 dynamic).

6. Financial Quality

The income statement (IFRS, USD):

Profitability — best-in-class. ROE was 28.1% (FY2024) and 30.3% (FY2025), with management reporting quarterly annualized ROE of 29–35% across 2025–Q1-2026. ROA is ~4.6% — an extraordinary figure for a bank (most large banks earn ~1.0–1.5%). NIM is ~21% and risk-adjusted NIM ~9.5–10.5%. These returns are the empirical proof of the cost-advantage moat: a bank cannot earn a 30% ROE and 4.6% ROA without either a genuine cost edge or reckless risk-taking, and the asset-quality data (below) argues against the latter.

Do economics improve with scale? Mostly yes, with a cyclical caveat. The flat cost-to-serve against tripling ARPAC, the falling efficiency ratio, and contained dilution all confirm positive operating leverage. But gross margin is not monotonically rising — it was 45.6% (FY2024) but fell ~3.6pp to 42.0% (FY2025) and 37.5% in Q1-2026, as funding costs (interest expense +62% YoY in FY2025) and ECL (+33% in FY2025, +76% YoY in Q1-2026) outpaced revenue. This is the rate-and-credit cycle, not a structural erosion — but it means the “economics improve with scale” thesis must be stated as operating-cost leverage (durable) layered over cyclically variable funding and credit costs (not improving in 2025–26).

Balance sheet — a genuine fortress. Total assets $77.5B (Q1-2026); total equity $12.6B; deposits $42.4B funding a credit book of ~$31B (net), leaving substantial excess liquidity in securities ($15.9B) and central-bank deposits ($9.2B). The credit-loss allowance is $6.1B (16.2% coverage of the portfolio, ~2.5x the 90+ NPL balance), and gross provisioning is running at 153.8% of new 90+ NPL formation — i.e., the bank is over-reserving relative to incoming bad loans. Regulatory capital (~$8.9B managerial vs. ~$3.6B minimum, ~2.5x) leaves ~$5.3B of excess capital. This is one of the strongest balance sheets in the sector.

Quality-of-earnings — three flags, in order of importance:

1. The IFRS effective tax rate flatters reported earnings (the single biggest QoE issue). The Q1-2026 IFRS ETR was 8.7% (vs. 29.9% a year earlier). The reconciliation starts from a combined ~40% Brazilian statutory rate and is reduced by income booked in low/zero-tax jurisdictions (notably the Cayman parent, +$167M), interest-on-capital (“JCP”) deductions (+$48M), and non-taxable sovereign-bond interest and R&D incentives (+$70M). Normalized at a 30% rate, Q1-2026 net income would be ~$668M, not $871M — the low rate added ~$203M, or ~23%. For FY2025 the effect is smaller (~$164M, ~6%, on a 25.8% full-year rate). Management itself guides to ~15–20% IFRS / ~30–35% “managerial” tax rates going forward, and legislated CSLL increases (rising to 17.5% in 2026–27, 20% in 2028+) push the rate up. Valuation should use a normalized ~30% tax rate, not capitalize the 8.7% quarterly print.

2. Cash taxes paid vastly exceed booked tax expense. Q1-2026 cash income tax paid was $1,527.6M against a booked tax expense of only $82.9M (the P&L netted $402.6M current expense against a $319.8M deferred-tax benefit); FY2025 cash tax paid was $1,641.7M vs. $996.7M booked. The deferred-tax asset has climbed to $3.0B. This is the cleanest evidence that the headline rate is partly a deferred-tax/timing artifact inflating IFRS net income relative to cash. Counter-point: ex-the-tax-line, multi-year operating cash generation ($3.5B in FY2025) exceeds net income, so the franchise is genuinely cash-generative — the issue is the tax line specifically, not wholesale earnings fabrication.

3. The new non-IFRS “managerial P&L framework” (introduced Q4-2025). Management now anchors guidance (efficiency ratio, gross-profit components, the 30–35% managerial ETR) to a self-defined framework that “preserves net income, cash flow and capital” but re-cuts the line items. It is not an EPS-inflation device (the bottom line is unchanged), but it was introduced precisely as growth decelerated, complicates clean YoY/peer comparison, and lets management steer attention to favored metrics. Sell-side consensus now mixes IFRS and managerial figures — a recipe for confusion. Treat managerial figures as hypothesis and reconcile to IFRS.

Cash flow caveat. For a bank, “operating cash flow” is dominated by growth in loans/receivables (a use of cash) and deposits (a source) — it is not a clean profitability signal. Q1-2026 CFO was negative $1.21B despite $871M net income, purely because the credit book grew faster than deposits that quarter. Capex is trivial (~$300M/yr of capitalized software). Do not read a “free cash flow” number here as one would for an industrial.

Verdict: high-quality, genuinely profitable economics with operating leverage that is real — but reported earnings are flattered by an unsustainably low tax rate, and margins are cyclically pressured. Normalize the tax and the business still earns a ~22–25% ROE — excellent, but not the 30%+ the headline implies.

7. Capital Allocation

The record is, so far, disciplined and shareholder-aligned — and 2026 marks a maturation.

• No dividend, ever — and none planned. Appropriate for a company compounding equity at ~30% with a long reinvestment runway; paying out capital that earns 30% internally would be value-destructive.
• First-ever buyback (June 2026): a $1.0B authorization of Class A shares (12-month window). This is funded comfortably out of ~$5.3B of excess capital and signals that management views the ~$12 stock (near 52-week lows, 1.4th-percentile own-history P/E) as cheap and the balance sheet as over-capitalized. It is a meaningful capital-allocation signal: the company is now the marginal buyer of its own shares. (At ~$60B market cap, $1.0B is ~1.7% — modest, but a directional first.)
• Funding is deposit-led, not debt/equity-led. Growth is financed by a $42B deposit base (+45% YoY) plus retained earnings — a low-cost, self-reinforcing funding model, not dilutive equity raises or leverage.
• M&A has been small, equity-funded, capability-driven tuck-ins (Easynvest → NuInvest, Spin Pay, Cognitect, Hyperplane, Olivia AI) plus one EM-fintech minority stake (Tyme Group, Dec 2024). No large, debt-funded, integration-risky deals. This is the right approach for a company whose edge is organic.
• Capital position is conservative (~2.5x minimum regulatory capital), supporting both growth and the buyback.

Dilution and stock-based compensation. SBC was $359M in FY2025 — and, notably, declining both absolutely (from $408M in FY2024) and as a percentage of revenue. Share count has grown only ~1%/year. For a high-growth tech-adjacent franchise, this is unusually contained dilution. The legacy founder mega-grant — the “Group CEO Award” granted to Vélez at the 2021 IPO (up to 2% of fully-diluted shares, issued in two 1% tranches at 60-day-VWAP hurdles of $18.69 and $35.30, with a 5-year service condition maturing ~Nov-2026) — is the one large overhang: the lower hurdle was cleared in 2021–24 (so the first ~1% tranche was likely earned), but the upper $35.30 hurdle is far underwater at ~$12. This is aggressive but hurdle-gated, and the dilution is bounded and disclosed.

Incentive alignment. Aggregate FY2025 compensation for all directors and key management was $91.3M — substantial, but spread across a large senior team and weighted toward equity. The controlling founder’s interests are aligned with shareholders on the upside (he owns a vast economic stake), though the control structure (below) misaligns on governance.

Verdict: management has allocated capital intelligently. Retain-and-reinvest while ROE is ~30%, fund with cheap deposits, avoid dilutive raises and risky M&A, contain SBC, and initiate a buyback only once the stock is demonstrably cheap and capital demonstrably excess. This is close to a textbook capital-allocation record for a high-return compounder. The only blemishes are the aggressive (if gated) founder grant and the governance structure.

8. Changes and Headwinds — Last Two Years

Strategic and corporate changes (a notably eventful 18 months):

• Dec 2024: Minority investment in Tyme Group (South Africa/Philippines digital bank) — first international optionality bet outside owned markets.
• 2025: Mexico received CNBV approval to begin converting Nu México into a full multiple-banking institution (operational audit requested Dec-2025 — the final step before full authorization). Banking-license advancement materially expands the deposit-funding and product runway in Mexico.
• Q4-2025: Introduced the non-IFRS “managerial P&L framework” (a quality-of-earnings watch item — see the Financial Quality section).
• Jan 2026: OCC conditional approval to form a de novo US national bank, “Nubank, N.A.” — to offer deposits, credit cards, lending, and digital-asset custody. Led by co-founder Cristina Junqueira (relocated to the US); chaired by Roberto Campos Neto, former President of the Central Bank of Brazil — a notable credibility hire. This is the formal vehicle for the US “call option.”
• May 2026: Q1-2026 earnings — the “second consecutive disappointing quarter” per the sell-side; BofA cut its price target ($17→$16, Neutral) and UBS also cut, lowering 2026/27 net-income estimates.
• June 1, 2026: CFO transition — Rob Livingston (ex-CFO of Visa North America) appointed CFO effective July 2026, replacing Guilherme Lago (CFO since 2019, becoming Special Advisor). Management states the operating model, risk appetite, and strategy are unchanged. A long-tenured-CFO departure during a sentiment trough is worth monitoring, though the incoming hire is credible.
• June 4, 2026: Inaugural $1.0B share-repurchase authorization.

Headwinds:

• Credit normalization / NIM compression. Q1-2026 risk-adjusted NIM fell 100bps QoQ to 9.5%, and ECL expense rose 76% YoY — driven, per management, by first-quarter seasonality, portfolio growth (every dollar of new exposure carries upfront IFRS-9 provisioning), and mix (incremental exposure tilting 98% toward higher-loss cards/unsecured). Management is explicit that this is not asset-quality deterioration (90+ NPLs fell 10bps to 6.5%), and guides risk-adjusted NIM back toward 2H-2025 levels as seasonality normalizes. The market is skeptical — hence the derating.
• Brazilian household-debt-service stress (the backdrop to Desenrola 2.0) and a mid-to-late consumer-credit cycle.
• Competitive intensity from Mercado Pago and the digital-challenger cohort.
• FX. USD-reported results are exposed to BRL/MXN/COP translation; management reports growth “FX-neutral” because reported USD growth and underlying local-currency growth can diverge materially.

Verdict: on balance, the changes strengthen the long-term thesis (two transformative regulatory wins — US charter and Mexico license — plus capital-return maturation) while the headwinds are predominantly cyclical, not structural. The CFO change and the managerial-framework introduction are governance/disclosure items to watch, not thesis-breakers.

9. Risk Analysis

Catastrophic-loss assessment. The probability of a permanent capital impairment is low: Nubank is over-capitalized (~2.5x minimum), over-reserved (coverage running ahead of NPL formation), deposit-funded (no wholesale-funding run risk), and short-duration on the asset side (able to react within days to credit deterioration). The realistic downside is a valuation de-rating plus a cyclical earnings hit — painful but recoverable — not a zero. The principal tail risks are a severe Brazilian macro shock coinciding with the unsecured-book scaling, and (lower probability) a governance/key-person event given founder control.

10. Valuation Discussion (Embedded Expectations)

Current multiples (at $12.19; ~4.91B diluted shares; ~$60B market cap; $12.6B equity):

• P/E: ~18.8x TTM reported (EPS $0.649); ~22–26x on tax-normalized earnings (applying ~30% vs. the flattered TTM rate). On forward FY2026 estimates (consensus reflecting the recent cuts), roughly ~16–18x reported / ~22–24x normalized.
• P/B: ~4.7x (BVPS $2.56). This is the richest book multiple in the Brazilian-financials group — but it is justified arithmetically by a ~30% ROE (a 4.7x P/B at 30% ROE implies a ~6.4% earnings yield on book, reasonable for a grower).
• P/S: ~3.4x.
• Own-history context: the AZI own-history valuation index places NU at the 1.4th percentile of its 10-year P/E range, ~22nd percentile P/B, ~11.7th percentile composite — i.e., NU has never been this cheap on its own history.

Peer comparison:

The peer set bifurcates sharply: NU and MELI trade as growth compounders; the Brazilian payments peers (STNE/PAGS) trade as distressed-value banks at ~6–7x and below book. NU is simultaneously cheap versus its own history and rich versus the Brazilian-acquirer peer group — but the acquirers occupy a structurally deteriorating pool and earn lower, lower-quality returns, so the premium is warranted. Against the right comp (a high-ROE growth bank), NU’s ~4.7x book for a ~30% ROE is cheaper than it looks: JPMorgan trades at ~2.6–3x TBV for a ~20% ROTCE; on a ROE-adjusted basis (P/B ÷ ROE), NU at ~0.16 is in line with or below JPM at ~0.13–0.15 — and NU is growing 30–40% versus JPM’s low-single-digits.

Embedded-expectations analysis — what must be true at ~$12? At ~$60B market cap on ~$2.9B of FY2025 net income (~21x), the market is underwriting roughly: net income compounding ~20–25%/year for several years (well below the recent 40%+) and a normalization of the tax rate (so reported EPS grows slower than pre-tax income) and a stable-to-modestly-rising cost of risk. In other words, the current price already discounts both a tax-rate normalization and continued credit normalization — it does not require the 8.7% tax rate or the 2024-era credit benevolence to persist. That is the contrarian crux: the de-rating to the 1.4th percentile of own-history has, in my read, already priced the two biggest bear points.

Scenario analysis (illustrative, FY2027 net income, no price target):

• Bear: Credit cycle worsens (90+ NPL >7.5%), tax normalizes to 30%, growth slows to ~15%, multiple compresses to ~15x normalized → meaningful downside from here.
• Base: Credit normalizes as guided, tax to ~25–30%, earnings compound ~25–30%, multiple holds ~20x normalized → mid-teens-to-20%+ annualized return driven by earnings growth, not re-rating.
• Bull: Risk-adjusted NIM re-expands, Mexico inflects to a “second Nu,” US optionality begins converting, growth holds ~35%+, multiple re-rates toward its history → substantial upside.

Verdict (embedded expectations only — no recommendation): the market is pricing NU as a decelerating, tax-flattered grower whose credit window is closing. The valuation is demanding in absolute terms (4.7x book) but undemanding relative to a 30% ROE, 30–40% growth, and the stock’s own history — and it appears to already embed the two principal bear arguments. The debate reduces almost entirely to ROE durability (how much survives tax normalization and the credit cycle) and growth longevity (how long the Brazilian/Mexican runways stay open).

11. Variant Perception

Consensus view. The sell-side and market consensus has cooled: NU is a great franchise, but two “disappointing” quarters (NIM compression, rising provisions), an opaque tax rate, a new self-defined “managerial” P&L, decelerating growth, and a richly-valued book multiple justify a wait-and-see Neutral stance and lower estimates. Hence the de-rating to 52-week lows and the recent PT cuts.

Strongest bull case. Nubank is the lowest-cost, highest-return consumer bank in one of the world’s most profitable banking markets, with a decade-plus runway in Brazil (7% of a $100B+ profit pool), a “second Nubank” inflecting in Mexico, a free US call option, and an AI transformation that widens the cost moat further — and it is being handed to patient capital at the cheapest valuation in its public history because growth merely decelerated from extraordinary to excellent and a seasonal credit quarter spooked momentum investors. The $1.0B inaugural buyback says management agrees the stock is cheap.

Strongest bear case. The 30% ROE is partly an illusion — flattered by an 8.7% tax rate that will normalize (legislated CSLL increases guarantee it) and by a benign credit window that is closing as the bank scales a +40%-growing unsecured book into a mid-to-late Brazilian consumer-credit cycle. Normalize tax to 30% and the multiple is ~26x, not 19x; let cost of risk rise and ROE compresses toward the low-20s; add a controlled-company governance discount (founder owns 74% of votes), intensifying competition from Mercado Pago, and FX/macro risk, and 4.7x book is not cheap — it is a premium for a decelerating, cyclically-and-structurally-flattered grower.

The 3–5 assumptions that matter most, and what would falsify each:

1. Credit is normalizing, not deteriorating. Falsified if: 90+ NPLs break above ~7.5% and risk-adjusted NIM keeps falling through 2H-2026 instead of recovering toward 10.5%.
2. The cost-advantage moat is durable and not diffusing. Falsified if: the efficiency ratio stops improving (ex-investment) and Mercado Pago/challengers take share in NU’s core Brazilian consumer base.
3. ROE stays ≥~22% after tax normalization. Falsified if: normalized (30%-tax) ROE drops toward the mid-teens, making 4.7x book unjustifiable.
4. The Brazilian/Mexican monetization runway is real (ARPAC keeps rising). Falsified if: ARPAC growth stalls and customer growth alone can’t carry revenue.
5. The data/AI underwriting edge holds through a down-cycle. Falsified if: a credit downturn reveals the nuFormer-priced cohorts loss-experience worse than modeled.

My variant read (consistent with Claude’s Take): consensus is over-weighting the cyclical/tax-quality bear points (which the 1.4th-percentile valuation has largely priced) and under-weighting the durability of a genuine cost-and-scale moat plus the optionality of Mexico and the US. The mispricing is one of sentiment and earnings-quality optics, not of business quality.

12. Fact vs. Interpretation

13. Open Questions

1. Standalone Brazil (Nu Pagamentos) Basel/BIS capital ratio — disclosed only on a combined managerial basis (~2.5x); the standalone Brazilian ratio is not separately quantified in the FY2025 20-F.
2. CEO “Group CEO Award” status — how many shares of the up-to-2% grant have actually been issued to date (lower $18.69 hurdle likely met; quantify the dilution as it matures ~Nov-2026).
3. Durability of the tax-rate arbitrage — how fast does the IFRS ETR converge to the guided 15–20% (and the managerial 30–35%) as legislated CSLL increases bite? What is the steady-state normalized rate?
4. Risk-adjusted NIM trajectory — does it recover toward 10.5%+ in 2H-2026 as management guides, confirming the seasonal-not-structural read on credit?
5. Mexico unit economics at scale — now that Mexico is IFRS-profitable, what is its standalone ROE trajectory and cost of risk as the credit book seasons?
6. US go-to-market — management has deliberately withheld the strategy; what product/segment is the wedge, and what is the realistic path to (and probability of) product-market fit?
7. Residual early-backer stakes (Berkshire, Tencent) — now sub-5% and undisclosed in the 20-F; confirm via 13F/press whether fully exited.
8. CFO transition — does the Livingston-for-Lago handoff change disclosure practice or capital-allocation posture?

14. What Must Be True

Bull case — what must be true:

• The Q1-2026 credit/NIM softness is genuinely seasonal: risk-adjusted NIM re-expands toward 10.5%+ and 90+ NPLs stay range-bound (~6–7%) through 2H-2026.
• Brazil monetization continues (ARPAC keeps rising) and Mexico inflects toward a self-funding, profitable second engine.
• The cost-advantage moat holds — efficiency ratio (ex-investment) keeps improving and the AI transformation widens, not just defends, the gap.
• Normalized (30%-tax) ROE stays ≥~22%, justifying the book multiple.
• Falsification test: if, by Q4-2026, risk-adjusted NIM has not recovered toward 2H-2025 levels and 90+ NPLs have risen above ~7.5%, the credit thesis is broken and the bull case fails.

Bear case — what must be true:

• The 30% ROE is materially flattered by an unsustainable tax rate and a closing credit window — normalize both and ROE compresses toward the mid-to-high teens.
• The unsecured-book scaling into a mid/late cycle produces a cost-of-risk spike that the “short duration / NPV-positive cohorts” framing cannot contain.
• Competition (Mercado Pago, incumbent digital arms) compresses share and spreads, ending the disproportionate-share-gain dynamic.
• Falsification test: if, through 2026–27, NU sustains a ≥25% ROE on a normalized (30%-tax) basis while holding 90+ NPLs below ~7% and growing earnings 25%+, the bear’s “flattered ROE / closing window” thesis is falsified.

15. Source Appendix

Key primary sources relied upon:

• Nu Holdings FY2025 Form 20-F (filed 2026-04-08) — consolidated financial statements, MD&A, principal shareholders, governance, capital ratios, share-based payments, dividend policy.
• Nu Holdings FY2023 / FY2024 Form 20-F — multi-year financial history.
• Q1-2026 interim financial statements & press release (6-K, 2026-05-14) — Q1-2026 income statement, balance sheet, cash flow, tax note, asset-quality bridges.
• Q4-2025 / FY2025 results 6-Ks (2026-02-25); OCC US national-bank 6-K (2026-01-29); CFO-transition 6-K (2026-06-01); $1.0B buyback 6-K (2026-06-04).
• Q1-2026 earnings-call transcript (2026-05-14) — management framing of credit, efficiency, tax, US optionality.
• SEC Form 4 / Form 144 corpus (CIK 1691493) — insider-transaction read.
• Prior the author internal research (STNE, PAGS, SOFI, MELI, JPM, BAC) — industry structure and peer multiples .
• Public macro sources — Brazil Selic (14.50%, BCB), income-tax-exemption and Desenrola 2.0 program details (accessed 2026-06-13).
• Quantitative cross-checks: AZI fundamentals/valuation-index feed and yfinance (reconciled to filings).

This article is independent fundamental research for general information only and is not investment advice. The body contains no recommendation and no price target; the sole exception is the clearly-labeled author’s-opinion block. Do your own due diligence.

APPENDIX A — Standard Diligence Questionnaire — Nu Holdings Ltd. (NYSE: NU)

Supplemental to the research memo. Grounded in the research log; Fact/Interpretation/Assumption labels applied where material. Report date 2026-06-13.

General

What thoughtful questions have other investors asked about this company? The recurring institutional debates (confirmed on the Q1-2026 call) cluster on three themes management itself flagged: (1) asset quality — is rising provisioning a sign of credit deterioration or just seasonality/growth/mix? (the dominant concern); (2) US internationalization — bull (breaking into the largest retail-banking market) vs. bear (US neobanking is a graveyard); (3) the role of AI — real productivity/underwriting edge vs. hype. Additional investor questions: the sustainability of the low (8.7%) effective tax rate; whether the new “managerial P&L” obscures comparability; Brazil’s household debt-service ratio; the SME opportunity (5M customers built at ~zero CAC); and the high-income segment push.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: a mixed picture. Operating leverage is at a structural high (record-low efficiency ratio), but reported net income is flattered by a cyclically/structurally low 8.7% tax rate and depressed by elevated, seasonally-high provisioning. Net-net, the quality of the Q1-2026 print is lower than the headline; normalized earnings (30% tax, mid-cycle credit) are below reported.

Driven by external environment or internal actions? Both. Internal: customer growth, ARPAC monetization, cost discipline (durable). External: Selic (14.50%, NIM-geared), the Brazilian consumer-credit cycle, and FX (cyclical/exogenous).

How stable are revenues? Highly recurring and granular (135M customers, 83% activity, deposit-funded). The variability is in NII (rate-sensitive) and the cost of risk (credit cycle), not in customer churn.

Outlook for products/services; how big is the market? Large and growing. Brazil served-profit-pool >$100B/yr gross profit (NU ~7% share); Mexico >$40B (NU <1% share). Assumption (mgmt): multi-year double-digit profit-pool growth plus share gains. Domestic and international (Mexico, Colombia, US optionality).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More — the digital-challenger cohort (Mercado Pago, Inter, C6, PicPay, incumbent digital arms) is crowded and well-funded; the cost advantage that was once unique is now shared among digital operators. But it remains less competitive than the deteriorating payments/acquiring pool (PIX-compressed).

How profitable is the business (ROIC/ROE)? ROE ~30% (FY2025), ROA ~4.6% — exceptional for a bank (most large banks ~1.0–1.5% ROA, mid-teens ROE). Interpretation: normalize the tax rate to 30% and ROE is ~22–25% — still excellent.

How profitable is the industry — competitors, barriers to entry? Brazilian banking has historically earned wide spreads (oligopoly of 5 incumbents). Barriers: regulatory (banking licenses), scale/cost, deposit-funding, brand/trust, and — increasingly — proprietary data. Barriers to digital entry have fallen (hence the crowded cohort), but barriers to profitable scale remain high.

Can the business be easily understood? Yes, at the unit level: customers × ARPAC × cost-to-serve − cost of risk − funding cost. The complications are the bank-specific accounting (IFRS-9 provisioning, deferred tax) and the new managerial framework.

Can it be undermined by foreign low-cost labor? Not directly — it is a domestically-regulated, deposit-taking bank. (Its own cost advantage partly derives from low-cost LatAm engineering talent.)

Do brands matter? Yes — Nubank has one of the strongest, highest-NPS consumer-finance brands in LatAm (“fans, not customers”), a genuine source of low-CAC growth and captivity.

Nature of competition; switching costs? Competition is on cost, product breadth, brand, and underwriting. Switching costs are real but not absolute — primary-account principality (direct deposit, bill-pay, card-on-file, credit history) creates stickiness; Open Finance is lowering friction.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The brand, the proprietary data/AI stack, and the customer base are not capitalized — genuine unrecognized intangible value. Interpretation: arguably the largest “hidden asset.”

Off-balance-sheet liabilities? Off-balance-sheet credit-card limits ($70.7B total exposure vs. $37.2B on-balance-sheet portfolio) — every undrawn dollar carries upfront IFRS-9 provisioning, so this is largely captured in the P&L provision build (a feature of the Q1-2026 ECL story, not a hidden liability).

How conservative is the accounting? Mixed. Provisioning is conservative (coverage 16.2%, running at 153.8% of new NPL formation — over-reserving). The tax line is aggressive-looking (8.7% IFRS rate, large deferred-tax benefits building a $3.0B DTA). The new managerial P&L is a presentational watch-item. Net: balance-sheet/reserving conservative; earnings-presentation requires scrutiny.

How CapEx-hungry? Very light — ~$300M/yr of capitalized software on a ~$78B balance sheet; asset-light, branchless.

Capital Allocation & Management

How much FCF does it generate; how is it used? Bank caveat: “FCF” is not meaningful (CFO is deposit/loan-flow-distorted; Q1-2026 CFO was −$1.2B despite +$871M NI). The economic equivalent is retained earnings reinvested at ~30% ROE plus a growing excess-capital buffer (~$5.3B). Capital is retained-and-reinvested, with a new $1.0B buyback as the first return of capital.

Significant acquisitions recently? Only small, equity-funded tuck-ins (Easynvest/NuInvest, Spin Pay, Cognitect, Hyperplane, Olivia) plus a Dec-2024 minority stake in Tyme Group. No large/debt-funded M&A.

Buying back shares? Yes — first-ever $1.0B Class A repurchase authorized 2026-06-04. Issuing shares to insiders? SBC $359M (FY2025), declining; dilution ~1%/yr. Founder “Group CEO Award” (up to 2% of FD shares, hurdle-gated) is the one large grant.

Compensation policy / management motivations? Aggregate director+key-mgmt comp $91.3M (FY2025), equity-weighted. Founder Vélez’s economic interest is massively aligned with shareholders on the upside; governance is misaligned via 74% voting control (controlled-company exemptions).

Valuation & Market Data

ADR, MLP, or K-1 issuer? NU lists Class A ordinary shares directly on NYSE (not an ADR) and BDRs (“NU Units”) on B3. Cayman-incorporated foreign private issuer — files 20-F/6-K, not a K-1/MLP. Tax note for US holders: a foreign-corporation share; verify PFIC status with a tax advisor (a financial-services operating company is generally not a PFIC, but confirm).

Dividend policy? No dividend, none planned (appropriate at ~30% reinvestment ROE).

How profitable; is net income diverging from cash from operations? Highly profitable (30% ROE). Yes, NI diverges from CFO — but for benign, bank-specific reasons (loan/deposit growth dominates CFO). The more relevant divergence: cash taxes paid ($1.64B FY2025) far exceed booked tax expense ($1.0B) — the opposite of the usual red flag, indicating the low booked rate is a deferred-tax/timing effect, with real cash taxes higher.

Risks & Downside

What would cause the stock to decline? A hard credit turn (90+ NPL >7.5%); tax-rate normalization compressing reported EPS; competitive share/spread loss; Brazilian macro/FX shock; faster-than-expected growth deceleration; multiple compression from 4.7x book.

Risk of catastrophic loss / total loss? Catastrophic loss: low — over-capitalized (~2.5x minimum), over-reserved, deposit-funded (low run risk), short-duration assets. Total loss: very low — a profitable, liquid, well-capitalized bank; the realistic downside is a cyclical-earnings + de-rating drawdown, not a zero. Principal tails: severe Brazil macro shock during unsecured-book scaling; (lower) governance/key-person event.

Recent News & Events

Has the business environment changed recently? Yes — materially: OCC US national-bank charter (Jan-2026), Mexico full-banking-license conversion advancing, CFO transition (Livingston ex-Visa replacing Lago, July-2026), first-ever $1.0B buyback (Jun-2026), and the Q1-2026 “disappointing” quarter that triggered sell-side PT cuts (BofA $17→$16, UBS lower). Macro: Brazil Selic easing began (14.50%), the R$5,000/month income-tax exemption (a consumer tailwind), and Desenrola 2.0 debt-relief (effective May-2026).

Significant acquisitions / accounting-policy changes / new markets? Tyme minority stake (Dec-2024); the new non-IFRS “managerial P&L framework” (Q4-2025); US market entry; Mexico license activation.

APPENDIX B — Source Appendix — Nu Holdings Ltd. (NYSE: NU)

Report date 2026-06-13. Primary sources first. Primary sources first. Citations distinguish primary filings and transcripts, third-party market data, and public industry context.

1. SEC Filings — Primary (Nu Holdings, CIK 0001691493)

2. Earnings-Call Transcripts

3. Third-Party / Market Data (reconciled to filings)

4. Public Macro & Industry Context

5. Analytical Frameworks

• Greenwald & Kahn, Competition Demystified — moat typing (low-cost-producer / economies-of-scale + customer captivity), barriers-to-entry, ROIC/share-stability tests.
• Marathon / Chancellor, Capital Returns — supply-side capital-cycle analysis (consumer-credit mid/late boom; acquiring-pool breakdown); asset-growth/mean-reversion warning on the +40%-growing unsecured book.