Nu Holdings: A 33% ROE Flattered by a 14.2% Tax Rate, 90+ Day NPLs Up to 6.9%, and 5.7 Times Tangible Book Leaves No Margin of Safety

Tax is a particularly important line. Q2’s IFRS effective tax rate was only 14.2%, versus 25.8% for full-year 2025 and about 34.9% on the managerial EBT presentation. The interim tax note starts from a 42.5% theoretical Brazilian rate and records large reductions associated with different tax rates, interest on capital and other items that principally include foreign tax credits, tax-exempt sovereign-bond interest, incentives and interest on recoverable taxes. Q2 also contained $846 million of current tax expense offset by a $671 million deferred-tax benefit. As a through-cycle valuation assumption, a 14% effective rate looks poor. Brazilian legislation is also scheduled to increase CSLL rates for relevant payment institutions and SCFIs from 2028. My base valuation normalizes the consolidated effective tax rate near 27–28%, rather than extrapolating Q2.

Normalizing tax alone changes the interpretation of the headline 33% ROE. Applying a 27.5% tax rate instead of Q2’s 14.2% to Q2 pretax profit would reduce annualized net income by roughly $0.66 billion. Against June equity of about $13.25 billion, the mechanical ROE reduction is about five percentage points. The implication is that sustainable ROE is closer to the high-20s unless tax benefits recur on the same scale; it does not mean Nu’s 33% was incorrectly reported.

Credit is the other axis of the valuation. The operating 90+ day NPL ratio reached 6.9% from 6.5% in Q1 and 6.5% a year earlier. The 15–90 day ratio improved to 4.8% from 5.0%, but management attributed much of that improvement to seasonality and explicitly said intentional expansion into higher-risk, higher-return customers partly offset it. The 90+ deterioration was attributed to first-quarter delinquency migration. At the same time, risk-adjusted NIM rose to 12.4%, which shows that current pricing is more than covering currently recognized credit expense. The debate is whether current loss recognition is truly through-cycle.

Balance-sheet provisioning is substantial. Credit-card receivables carried $4.52 billion of allowance against $25.96 billion gross exposure; loans carried $2.12 billion against $13.44 billion. Cards had about 9.1% of gross balances in IFRS Stage 3 with 85.9% Stage-3 coverage; loans had about 8.3% in Stage 3 with 66.2% Stage-3 coverage. Across cards and loans, total allowance was about 16.9% of gross exposures. A rough comparison of the entire $6.64 billion allowance with the operational 90+ NPL balance implied by a 6.9% ratio gives coverage above 240%, although that comparison deliberately mixes an all-stage IFRS allowance with a delinquency-only NPL definition. Stage-3-specific coverage is the cleaner accounting measure, at roughly 80% across the two portfolios.

Caution comes from write-offs. Card and loan write-offs totaled about $2.33 billion in the first half of 2026, versus roughly $1.51 billion in H1 2025, an increase of about 54%. Using period-end gross credit as a crude denominator, the first-half number annualizes to a double-digit percentage of the portfolio. Growth and seasoning make that an imperfect ratio, but it is enough to reject any thesis that Nu has solved unsecured credit risk. Its achievement is different: so far, it has charged enough yield to earn very high returns despite bearing that risk.

FX has flattered the latest USD growth rates. On the managerial presentation, Q2 revenue rose 55.8% year on year in reported USD, while Nu’s stated FX-neutral growth rate was 39%. Net income rose 66.6% in reported USD while the company quoted 49% FX-neutral growth. Those are like-for-like comparisons within each metric and show roughly a high-teens percentage-point translation benefit. IFRS revenue, separately, rose 50.3% in reported USD. I never mix the 39% FX-neutral managerial growth rate with the 50.3% IFRS reported rate. The Brazilian real was around R$5.11 per dollar on September 22, a reminder that future USD earnings can move materially without equivalent changes in local operating performance.

Brazil remains the proven economic core. Mexico is already large enough to matter operationally, with 15.8 million customers at June and about 16 million by July. It began operating as a multiple-banking institution on August 6, 2026. Management says Mexican customers monetize much earlier than Brazilian customers did at the same point in their development, citing ARPAC of $12.3 versus $5.6. Yet Mexico’s loan-to-deposit ratio was only 35% at June, so today it is still carrying a sizeable pool of deposits that has not been converted into the high-return credit economics that make Brazil valuable.

Colombia has surpassed five million customers but remains earlier. The United States is earlier still. On September 10 Nu launched a 3.50% APY account and a no-annual-fee, 1.5%-cash-back credit card through FDIC-insured Lead Bank while its own OCC national-bank application remains conditionally approved. Nu Global launched simultaneously, offering yields on USDC and EURC balances and selected digital-asset trading. Those products make strategic sense as tests of Nu’s brand and software outside Latin America, but they enter markets with radically lower switching barriers and much more expensive customer acquisition. I assign positive option value to them and almost no current-earnings value.

The rate cycle cuts both ways. Brazil’s Copom made its fifth consecutive reduction in September, bringing Selic to 13.75%, while signaling caution over further easing; market expectations still allow another cut by year-end. A simple $45.3 billion deposit-base sensitivity illustrates why the net effect is subtler than “lower rates hurt Nu.” A 100-basis-point fall in asset yields would reduce gross annual float income by as much as roughly $453 million if that entire balance repriced one-for-one. At Nu’s disclosed consolidated deposit cost of 88% of interbank rates, the same move could lower annual funding expense by up to roughly $399 million on the same simplifying base. That leaves a direct net rate hit closer to $54 million before tax in that stylized calculation. If easier rates then reduce credit-loss formation by only 25–50 basis points on $39.4 billion of credit, that adds roughly $100–200 million of annual pretax benefit. Timing and asset-liability betas will decide the actual outcome.

At $14.16, using the June 30 economic share count of 4.831 billion shares, the equity is worth $68.40 billion. June tangible book value, after deducting about $409 million of goodwill and $747 million of intangibles from attributable equity, was approximately $12.09 billion, or $2.50 per economic share. On that basis Nu trades at roughly 5.7 times tangible book. Trailing IFRS earnings through June are around $3.6 billion by my calculation, putting the stock around 19 times trailing earnings.

Put side by side, the two multiples define the present disagreement. A 19-times P/E for a company still producing high-teens to 30%-plus operational growth can look inexpensive. A 5.7-times tangible-book multiple for an emerging-market unsecured lender is expensive unless ROE stays very high for a long time. Both statements are true at the same time. The P/E bull case assumes earnings growth converts into rapidly growing book value and that excess returns endure. The P/TBV bear case emphasizes that the market already capitalizes several years of unusually high returns.

Qualitative portrait: high-quality compounding growth. Valuation is the qualification. Brazil has already proven that Nubank’s branchless distribution, brand and underwriting can produce bank-scale profits with a structurally low cost base. Mexico has begun the transition from customer acquisition to banking economics. The U.S. and Nu Global are options rather than proven franchises. The stock presently prices substantial continuity in Brazilian ROE and successful maturation of at least Mexico. Credit seasoning, tax normalization and international spending decide whether that expectation is conservative or aggressive.

Company story, financial vertical and capital-market narrative

Nu began in São Paulo in 2013 around a narrow wedge into one of the most concentrated and disliked corners of Brazilian consumer finance: the credit card. Founder David Vélez combined financial-services and venture experience; Cristina Junqueira brought knowledge of incumbent Brazilian banking and product economics; Edward Wible brought the engineering architecture. The initial product was intentionally narrow. A no-annual-fee card controlled from a mobile application attacked fees, customer-service friction and branch dependence rather than trying to build a full bank on day one. The 2025 Form 20-F still traces today’s group back to that model.

The early competitive insight was as important as the software. Brazil already had enormous banks with cheap deposit franchises, dense branch networks and broad product menus. Nubank could not outspend them on physical distribution or offer the full balance sheet of Itaú, Bradesco, Banco do Brasil or Santander at inception. It chose a product where approval, servicing and engagement could be digitized. That allowed the company to learn underwriting while using a highly visible card as its customer-acquisition product. The purple card became marketing in physical form.

Product validation defined the first stage, roughly 2013–2017. The important asset created in those years was proof rather than card revenue: proof that a financial institution could acquire Brazilian mass-market customers without a branch, maintain a direct digital relationship and collect enough behavioral data to improve credit decisions. Venture backers including Sequoia and Kaszek financed that experiment before the economics could support the scale of lending required.

From the late 2010s into 2020, the second stage turned the card into an account relationship. Deposits changed the model. A card fintech that depends on wholesale funding and interchange has a weaker economic architecture than a deposit-funded bank. As Nu broadened from cards into accounts, payments, personal loans and investments, customer engagement became a source of funding and cross-sell rather than merely a source of interchange. That evolution explains why today’s float economics account for roughly a third of managerial gross profit.

Geographic expansion created the third stage. Mexico and Colombia offered some of the same ingredients Nu had exploited in Brazil: large consumer markets, concentrated incumbent banking systems and growing smartphone adoption. The strategy, however, carried an important asymmetry. Brazil supplied brand, capital and learning; each new country required a fresh regulatory stack, fresh underwriting models and local deposit economics. Customer counts can travel quickly. Credit-loss curves cannot simply be copied.

Nu went public in December 2021, near the end of the global zero-rate growth-stock boom. It priced its NYSE offering at $9 a share, after building a shareholder base that included major global venture and institutional investors. The capital market initially treated the company as a technology-led financial platform rather than as a conventional bank. That framing was rational while customer growth dominated reported profits; it became less useful as the balance sheet expanded. The company’s current 20-F confirms NYSE Class A trading, Cayman incorporation and the foreign-private-issuer reporting structure.

Then came the fourth stage, the 2022–2023 post-IPO reset. Higher global interest rates crushed valuation multiples for long-duration growth equities at the same time investors questioned whether Nubank’s rapidly expanding consumer credit would ever produce enough profit to justify the IPO story. The equity’s subsequent recovery came with a change in evidence: Nu moved from promising future operating leverage to actually reporting large profits. The market began valuing user growth through ARPAC, cost of funding, NIM, losses and ROE rather than through customer additions alone.

That shift is visible in the financial record. IFRS revenue reached $8.03 billion in 2023, $11.52 billion in 2024 and $15.77 billion in 2025, a two-year compound rate of about 40%. FY2025 net income reached $2.872 billion, up 45.6% from $1.972 billion in 2024, while expected-credit-loss expense rose to $4.205 billion from $3.169 billion. Nu grew profit rapidly even while absorbing a large and rising absolute credit bill.

† H1 2026 ECL figure is based on the interim statement and underlying quarterly expense; half-year comparisons should be read with portfolio growth in mind.

Now underway, the fifth stage is more ambitious and more difficult: transforming a highly successful Brazilian digital lender into a multinational regulated bank group. Mexico became a multiple-banking institution on August 6, 2026. In Brazil, Nu signed an agreement in July to acquire Banco Porto Real de Investimentos, seeking a full banking license as Brazil standardizes which regulated institutions can use banking terminology. The U.S. experiment began through Lead Bank on September 10 while Nu continues pursuing its own OCC charter. Nu Global moves the brand into stablecoin-based cross-border balances and crypto.

This stage has a different failure mode from the first four. Brazil has already proved product-market fit. The risk now is capital allocation. A dollar spent improving Brazilian secured lending or primary-bank penetration has measurable unit economics. A dollar spent acquiring a U.S. depositor at 3.50% APY has far less historical evidence behind it. The company can remain an excellent Brazilian bank while destroying incremental value by overinvesting in foreign options.

Balance-sheet growth remains sound enough to fund experimentation. At June 30, 2026 Nu had $82.75 billion of assets, $45.33 billion of deposits, roughly $39.4 billion of gross cards and loans, $4.68 billion of borrowings and $13.25 billion of equity. Deposits exceed the gross credit portfolio, reducing dependence on wholesale financing. Brazil supplied $36.4 billion of deposits, Mexico $5.7 billion and Colombia $3.3 billion.

A conventional industrial-company cash-flow analysis is misleading here. For a bank, deposits are simultaneously customer products and funding, while loan originations are operating assets rather than “capex.” Cash flow from operations can swing dramatically with deposit gathering and credit growth without saying much about distributable owner earnings. For that reason I do not use a five-year operating-cash-flow/net-income ratio as a valuation anchor. The economically relevant cash passthrough is earnings after expected credit losses, tax and the equity capital that must be retained to support balance-sheet growth. The financial statements also do not provide a reliable maintenance-versus-growth capex split comparable with an industrial company. Much of Nu’s growth investment is expensed through technology, personnel, marketing and credit provisions rather than recorded as physical capex.

That methodological choice makes “owner earnings” slightly more conservative than headline net income. Nu is profitable, but shareholders cannot treat all profit as immediately distributable while the loan book grows 30%-plus and international regulated entities require capital. So in the valuation I normalize credit losses and tax first, then value retained earnings through the residual-income model.

The Q2 buyback is an interesting capital-allocation test. The board authorized $1 billion and Nu spent $500.4 million by June 30 to acquire 40.66 million Class A shares at an average price of about $12.30. That was below the current $14.16 price but several times tangible book value. The repurchased shares remain in treasury and had not been canceled, leaving them available for future use. The program reduced the treasury-adjusted economic share count by about 0.8% relative to what it otherwise would have been, but stock-based compensation continues and a treasury share can later be reissued. I give the program credit for offsetting dilution and shrinking current outstanding equity, while permanent cancellation would provide the cleaner shareholder-return signal.

† Class B is unlisted but converts one-for-one into Class A and carries identical economic rights; omitting it understates market capitalization.

No post-June company disclosure found before the research base date quantified additional buyback shares. Absent that, I use the June 30 treasury-adjusted count rather than inventing an updated figure. Form 144 sales by insiders do not alter the company share count.

Those August Form 144 filings matter less than their number suggests. Cristina Junqueira or vehicles associated with her disclosed a 50,000-share sale on August 14 under a Rule 10b5-1 plan adopted March 27; an associated ZJ WY LLC filing covered 300,000 shares on August 17; another affiliate sale of 100,000 shares is recorded in the subsequent three-month-sales disclosure; the trust sold 25,000 shares on August 24; and Junqueira filed to sell another 65,000 on August 25. Much of the stock originated in long-held 2016 interests or vested compensation.

Officer Henrique Camossa Saldanha Fragelli separately filed on August 14 to sell 221,707 shares with an indicated market value of about $3.5 million, drawn from vested restricted stock. The combined known Junqueira-related activity and Fragelli proposal comes to roughly 762,000 shares, about 0.016% of Nu’s economic share count. At least part is explicitly pursuant to a pre-existing 10b5-1 arrangement. I read this as routine diversification and compensation monetization rather than evidence of insiders exiting the thesis. Not every filing identifies a 10b5-1 plan, so “routine” should not be stretched into “all automatic.”

The governance issue with real valuation consequence lies elsewhere. David Vélez beneficially owned 88.3% of the Class B shares as of March 1, 2026; with 20 votes attached to each Class B share versus one for Class A, that translated to roughly 74% of combined voting power. Class B converts one-for-one into Class A economically. Minority shareholders own the same per-share economics yet have little practical ability to replace the founder, force a strategic change or block transactions he supports. Cayman domicile and foreign-private-issuer status add another layer of distance from U.S.-domestic governance norms.

Formal board independence reduces agency risk but cannot remove this shareholder-level control structure. I treat it as a cost-of-equity input. My base model effectively carries about a 50-basis-point governance/control premium inside the discount rate relative to what I would demand from a similarly profitable bank with one-share-one-vote ownership.

The market has already shown it cares about earnings quality. When Nu reported strong Q4 2025 customer, revenue and profit growth in February 2026, the shares fell about 5.5% after hours as investors focused on costs and concerns around tax-related profit support. That reaction was useful: Nu is no longer rewarded simply for beating on customers or revenue. The quality and repeatability of ROE are now the dominant variables.

The lasting lesson from Nu’s history is that management has repeatedly converted distribution scale into a new financial asset: first card engagement, then deposits, then credit, then primary-bank relationships. The next test is whether that conversion process travels across borders without destroying the cost advantage that created it.

Business model, moat, industry and regulation

Nu’s reported IFRS income statement naturally resembles a bank more than a software company. Q2 IFRS revenue consisted principally of interest income and gains on financial instruments plus $753 million of fee and commission income. Interest and financial expenses were $1.557 billion, expected-credit-loss expense was $1.482 billion and transaction costs were about $128 million before operating expenses. The managerial presentation rearranges those same economics into $3.605 billion of credit income, $1.454 billion of float income and $816 million of fee income.

Broken down this way, the numbers explain why Nu can produce both spectacular growth and volatile earnings sensitivity. Credit earns a large spread but consumes loss provisions and equity. Float requires much less customer-level underwriting but depends on rates and deposit pricing. Fees are less balance-sheet intensive, though card rewards, network and transaction costs absorb part of their apparent revenue.

Costs come in three layers. Funding cost scales with deposits and market rates. Credit loss scales with loans, customer mix and the economic cycle. Transaction costs scale with payments volume. Above those sit the technology, customer support, marketing, compliance and corporate functions that create operating leverage as users grow. Q2 customer-support expense was about $226 million, G&A $600 million and marketing $103 million on the IFRS statement.

Nu’s 19.5% efficiency ratio is one of the best pieces of evidence for genuine operating leverage, although it is a managerial rather than IFRS metric. It rose from 17.6% in Q1, with management attributing the change partly to real-estate and marketing expenditure that shifted from the first quarter into the second and partly to continued international investment. A single quarter does not prove structural cost deterioration. A sustained ratio above roughly 23–25% while customer growth slows would be much more concerning.

The real moat is a combination of distribution, low servicing cost, deposit funding and accumulated underwriting data, strongest in Brazil. The brand gives Nu low-friction customer acquisition and unusually high activity. Branchless servicing turns scale into low unit costs. Deposits turn engagement into funding. Credit history then gives Nu a larger proprietary dataset for offer selection and pricing.

Equal weight belongs on the weaker moat claims. Financial accounts have low technical switching costs and consumers can multi-home. Payments infrastructure increasingly commoditizes basic transfer functionality. Deposits leave when competitors offer enough incremental yield. A credit-scoring edge also decays if management deliberately moves into borrower segments for which its historical dataset is less representative. Nu’s moat resembles a powerful operating flywheel within a regulated balance sheet rather than a software monopoly.

Customer stickiness is best evidenced by behavior rather than contracts. Of almost 118 million Brazilian customers at Q2, more than 86% were active monthly. Group activity was 83.5%. With 138.9 million total customers, that implies well above 110 million monthly-active relationships. Brazil’s penetration is now so high that the next stage of domestic value creation must come increasingly from primary-bank status, higher balances, payroll, secured lending, investments, affluent customers and SMEs rather than from simply adding first-time Nubank accounts.

The “largest digital bank in Latin America” label needs a definition. Nu itself uses the claim, and 139 million customer relationships makes it plausibly the largest standalone digital-banking platform in the region by reported customer count. I found no regulator-maintained Latin American category called “digital banks” that would certify the ranking across banks, wallets and payments platforms. Mercado Pago, for example, competes intensely for payments and credit but is embedded in a broader commerce ecosystem, making user counts definitionally different. I accept “largest” on the narrow basis of Nu’s reported banking-platform customer scale, not as a claim that it is Latin America’s largest financial institution by assets, deposits, loans or profit.

Brazil now presents a maturity paradox. Nu’s consumer reach is enormous, but its share of household financial wallets remains much smaller than its share of people. That gap leaves ample room for ARPAC growth if the company can become the primary financial institution for existing customers. Launching products aimed at more affluent Brazilian customers and expanding into secured lending matter more economically than headline account growth because they increase balances and reduce reliance on unsecured revolving economics. Management specifically highlighted primary banking and its Croma proposition in Q2.

Mexico is at an earlier point, so it offers a larger penetration runway. Nu had 15.8 million customers at June and described them as about 16.5% of Mexico’s adult population; the company crossed 16 million in July. Its Mexican website confirms that Nu began operating as a bank on August 6, 2026 and is regulated by CNBV, Condusef and Banxico.

On cohort monetization, the comparison is encouraging but not conclusive. Management says Mexican ARPAC at a comparable stage is $12.3 versus $5.6 for Brazil. That can indicate better product sequencing, learning transferred from Brazil, and stronger deposit monetization. It can also reflect a different rate environment and product mix. Mexico’s 35% loan-to-deposit ratio shows that current ARPAC is not yet proof of mature credit economics. The largest source of future profit is deploying those deposits into loans without damaging loss rates.

Colombia remains an option with more than five million customers and $3.3 billion of deposits at June. Its early balance-sheet pattern resembles Mexico: customer and funding scale can arrive before a mature credit portfolio. That means headline customer growth can look better than near-term ROE.

Entering the U.S. is a fundamentally different problem. Lead Bank supplies the regulated banking infrastructure for the September launch, while Nu markets a 3.50% APY account, fee-free transfers and a credit card paying 1.5% cash back. That product set begins in a market where direct banks, money-market funds, brokerage cash products and rewards cards already compete aggressively on price. Nubank’s Brazilian brand carries far less organic acquisition value in the United States, so customer-acquisition cost, fraud, rewards expense and the cost of deposit incentives matter more than application design alone.

Nu Global increases both strategic optionality and the regulatory surface. It allows balances to be converted into USDC and EURC, with advertised yields of 3.50% and 2.20%, and provides selected digital-asset trading. Those balances should be analyzed separately from ordinary insured bank deposits. Stablecoin reserves, issuer concentration, custody, redemption mechanics and changing digital-asset regulation can create operational and reputational loss without a classic bank run occurring at Nu itself.

Neither of those U.S. products automatically inherits the economic moat. Brazil proves that Nu can build a low-cost digital retail bank in an under-served, concentrated market. It does not yet prove that Nu can out-acquire or out-monetize U.S. incumbents where digital onboarding is universal.

Rate cyclicality is central to all three Latin American businesses. Brazil’s Selic stands at 13.75% after the fifth consecutive rate cut, and the central bank signaled that further easing would be gradual and data-dependent. One private-bank forecast cited on September 22 expected 13.25% by year-end and 11% in 2027.

The first-order sensitivity can be separated rather than netted prematurely:

† These are research sensitivities, not company guidance. Actual asset and deposit betas, country mix and hedging will differ. Inputs use Q2 deposits, credit and Nu’s disclosed 88% cost-of-deposit ratio.

That table explains the unusual rate exposure. Higher Selic creates excellent gross carry on liquid balances but simultaneously raises deposit costs and eventually stresses household borrowers. Lower Selic reverses each channel. If credit losses fall enough, the gross float headline can decline while total economics improve. Mexico is somewhat more vulnerable to an early easing-cycle carry squeeze because its 35% loan-to-deposit ratio leaves more of the deposit base waiting to be converted into loans.

Brazilian tax policy adds a less cyclical headwind. Nu’s interim filing describes Complementary Law 224/2025 as setting CSLL rates for payment institutions at 12% in 2026–27 and 15% from 2028, while relevant SCFIs move from 17.5% to 20%. The consolidated effective rate cannot be derived by applying any one statutory rate to group pretax income because legal entities, tax-exempt instruments, credits and deferred taxes differ. Directionally, however, the 2028 schedule argues against using Q2’s 14.2% ETR in perpetuity.

Acquiring Banco Porto Real carries a second Brazilian regulatory effect. A full banking license resolves part of the naming/licensing issue created by Brazil’s standardized rules for regulated institutions, but a bank structure also subjects activities to a fuller prudential capital and liquidity framework. Nu has not published enough pro-forma RWA and regulatory-capital detail for me to quantify the acquisition as either capital-accretive or dilutive. I assign it operational value but no speculative capital-release value.

Mexico’s new banking status should ultimately make the local franchise more complete. Nu also said central-bank payment-rule changes become mandatory for institutions by year-end 2026; management framed its SPEI transfer experience as preparation for that transition. Regulatory compliance costs will rise with the franchise, but obtaining a full license also removes product constraints.

Governance regulation is the final structural issue. Nu’s founder-controlled dual class, Cayman parent and foreign-private-issuer status are legal, disclosed choices. They create a persistent minority-holder discount because economic ownership and voting influence are radically different. That discount stays justified even if management’s historical decisions continue to be good. Trust in a founder is not a substitute for a mechanism that allows shareholders to replace him.

Horizontal competition and current fundamentals

Nu competes in several ecosystems simultaneously, which is why no single peer multiple should set the valuation. The closest operating competitors are incumbent Latin American banks, while MercadoLibre is the most important digital ecosystem competitor. Kaspi is the closest conceptual analogue, and SoFi helps frame the U.S. opportunity. XP is a useful reference for Brazilian financial distribution and Selic exposure. Robinhood captures how markets price retail-finance optionality, though its trading and crypto economics are far removed from Nu’s deposit-funded lending.

Incumbent banks make the most important comparison because they reveal what Nu has become. Itaú, Bradesco and Santander Brasil earn money across corporate banking, consumer lending, insurance, payments, wealth and investments. Their physical and legacy infrastructure raises servicing costs but also creates product depth, long credit histories, diversified collateral and entrenched payroll/corporate relationships. Nu sacrificed that breadth in its first decade for lower distribution cost and faster iteration.

Current market capitalization makes the expectations visible. Using September 22 ADR/equity data, Itaú’s market capitalization was about $82.6 billion, Bradesco’s $38.1 billion and Santander Brasil’s $43.6 billion. Nu’s full two-class economic value is $68.4 billion. The market is already valuing Nu closer to Itaú than to the smaller Brazilian ADR comparables even though Nu’s absolute asset base is far smaller. Investors are paying for growth and ROE, not balance-sheet size.

† Nu value is recalculated using both economic share classes; vendor market-cap conventions can differ.

I do not publish an invented “like-for-like” Q2 2026 table of incumbent card charge-offs and unsecured loss ratios. The Brazilian banks do not disclose those categories on identical definitions, while Nu’s operational NPLs, IFRS Stage 3, ECL provision and write-offs themselves use different denominators. Forcing a false precision would be worse than acknowledging the comparability limit. Structure supplies the useful conclusion: incumbents own more secured and corporate diversification; Nu earns much higher yields from a younger, more unsecured book and must preserve a wider risk-adjusted spread to compensate.

MercadoLibre became something different from Nu. Mercado Pago began as transaction infrastructure for an e-commerce marketplace and evolved into a wallet, acquiring, deposits/investments and credit ecosystem. Commerce gives it proprietary merchant data and transaction flows that Nu does not possess. Nu has the cleaner banking franchise and larger dedicated financial relationship; MercadoLibre has the stronger connection between purchase intent, merchants, advertising and payments. Customers pick Mercado Pago because commerce and payments are already happening in the same ecosystem; they pick Nu because they want a primary financial product independent of a shopping marketplace.

Valuation follows from that difference. MercadoLibre traded near $1,827 on September 22 with a market capitalization around $92.6 billion and a trailing P/E close to 50 times, far above Nu’s roughly 19-times trailing IFRS earnings. The premium is partly the value of its commerce and advertising businesses, which do not consume bank equity in the same way. Comparing Nu to MELI and declaring Nu “cheap” would be a category error.

Kaspi.kz is the closest business-model analogue because banking, payments and marketplace activity sit inside one emerging-market consumer application. Where they differ is geographic concentration. Kaspi has built exceptional penetration in Kazakhstan and expanded through acquisitions, while Nu operates across much larger Latin American populations. Kaspi shows what extremely dense financial and transaction engagement can produce; it also shows that a super-app’s multiple can carry a large country-risk discount. Its September 22 U.S.-traded share price was about $96.45.

SoFi is more relevant to the U.S. launch. It uses a bank charter, deposits and digital acquisition to fund lending while cross-selling brokerage, cards and technology services. Its U.S. customer has many more digital-bank alternatives than Nu’s first Brazilian customers had. SoFi traded at $17.16 with a roughly $23.2 billion market capitalization and a trailing P/E around 35 times on September 22. Nu can enter the United States with better scale and more profit than SoFi had during its early bank phase; it cannot import Brazil’s competitive structure.

XP has the opposite balance-sheet emphasis. Its franchise is investment distribution and advisory rather than mass-market unsecured credit. It shares Brazil’s Selic, BRL and regulatory exposure with Nu but has less direct consumer-credit loss formation. XP’s market capitalization was about $11.6 billion at a $21.21 U.S. price. The comparison helps most with how lower Brazilian rates alter the attractiveness of deposits, fixed-income products and investment flows, not with valuing Nu’s loan book.

Robinhood belongs in the capital-market comparison rather than the bank comparison. It monetizes trading activity, cash balances, subscriptions and crypto, and its earnings are more exposed to market participation than household NPLs. At about $124.25 and a $113.3 billion market capitalization, with a trailing P/E near 55 times, it illustrates the multiple public markets will award retail-finance platforms when investors believe optionality is expanding. As a valuation anchor for a lender, it offers little protection.

† Nu P/E is my calculation using full economic market capitalization and trailing IFRS net income through June; peer figures are market-data-provider trailing values.

Across that peer set, Nu occupies an unusual ecological niche. Against traditional banks it is the high-growth, low-cost challenger. Next to digital wallets it has a deeper banking balance sheet, and compared with fintech growth stocks it already has bank-scale profits. That combination explains why neither a traditional-bank discount multiple nor a 40–50-times fintech P/E is satisfactory.

The latest four-quarter direction is stronger than the simple Q2 snapshot. By Q4 2025 Nu had reached about 131 million customers and continued posting roughly 40%-plus revenue growth while net profit approached $900 million for the quarter. Q1 2026 net income then reached roughly $871 million, and Q2 moved above $1 billion. The sequence shows operating leverage and ARPAC growth continuing after the customer base passed 100 million.

Q2 itself delivered reported IFRS revenue of $5.513 billion, up 50.3% year over year in USD. Currency translation boosts that number. Managerial revenue rose 39% FX-neutral, and net income rose 49% FX-neutral. Monthly ARPAC reached $17.1, up 22% FX-neutral, while customer growth slowed to 13% year on year. The mix is healthy: monetization is replacing raw customer additions as the main growth engine.

Net interest income hit $3.7 billion and NIM 22.9%, up 180 basis points sequentially. Cost of credit fell 9% from Q1 and risk-adjusted NIM reached 12.4%. Those metrics explain the quarter’s profitability better than customer growth: Nu earned more on each active relationship while the near-term credit cost eased.

The credit mix is evolving. Of the $39.4 billion portfolio, about $26.0 billion was credit cards, $10.3 billion unsecured lending and $3.1 billion secured lending. Secured lending is still only about 8% of the total. Brazil’s next stage of growth should improve risk diversification if payroll, secured and higher-income products grow faster than cards. If the company instead gets most incremental growth from deeper penetration of higher-risk unsecured segments, the current 12.4% risk-adjusted NIM needs to remain very wide.

The credit verdict today is “well reserved, highly profitable, but unseasoned at the newest risk frontier.” IFRS allowances are large and Stage-3 coverage is not thin. Yet write-offs are rising faster than the book, 90+ delinquencies have moved up, and management openly acknowledges higher-risk expansion. The current evidence supports neither a crisis thesis nor a “credit is solved” thesis.

Stage 2 deserves more attention than the headline 90+ NPL ratio. About 12% of cards and 14.5% of loans have already experienced a significant increase in credit risk under IFRS 9 and therefore carry lifetime rather than merely 12-month expected loss. If these pools stabilize, current provisions can prove conservative. If Stage 2 migrates into Stage 3 while management adds riskier new cohorts, provisions could remain elevated even after Selic falls.

Public disclosure is insufficient to run the ideal vintage test. Nu discusses underwriting performance and has highlighted particularly favorable delinquency for its PBR segment, but the interim statements do not provide the full origination-vintage curves, renegotiated-balance waterfall and cured/defaulted cohort tables needed to judge every recent risk cohort independently. For a lender growing this quickly, that is a meaningful information gap.

U.S. card lenders Capital One and Synchrony are useful conceptual references because they show the economic reality of high-yield consumer credit: high gross yields do not equal excess return until net charge-offs, reserve building, rewards and funding are deducted. The comparison should stop there. U.S. CECL, product mix, borrower income, legal collections and funding structures differ from Nu’s IFRS 9 Latin American portfolio. I do not import their loss rates into Nu’s model.

Today the market is mainly trading the durability of Brazil’s high ROE plus the possibility that Mexico repeats Brazil faster. U.S. launch headlines add option value and investor attention, but current earnings still come overwhelmingly from the established Latin American operation. The first Investor Day on December 8 is likely to sharpen that debate because management has promised discussion of long-term strategy, growth opportunities and value drivers across markets.

The positive expectation is that customer growth can slow without earnings growth slowing because ARPAC, primary-bank penetration and secured credit take over. The negative expectation is that 2026 margins are being flattered by three conditions at once: favorable FX translation, a very low IFRS effective tax rate and a credit book whose newest higher-risk cohorts have not fully seasoned.

Those are testable views rather than competing narratives that can persist indefinitely. The next several quarters will show tax normalization, Stage-2 migration, write-offs and Mexico loan deployment.

Valuation, risks, catalysts and tracking

The current $14.16 share price implies a full economic equity value of $68.4 billion. At June tangible equity of approximately $12.09 billion, the stock trades around 5.66 times tangible book and 5.16 times reported book. Using trailing IFRS earnings through June gives a P/E around 19 times.

Because ROE is so high, those two multiples tell different stories. A bank earning 30% on tangible equity can justify a large P/TBV premium, but only while investors believe much of that ROE spread over the cost of equity will survive. A fintech growing earnings 25–30% can justify a 19-times P/E, but only while enough earnings are retained at high incremental returns rather than consumed by credit deterioration or foreign expansion.

Historical valuation percentiles are less useful than they look. Nu spent its first post-IPO years either unprofitable or only beginning its profit ramp, so a P/E history mixes fundamentally different earnings regimes. The 2022 trough reflected both collapsing growth multiples and doubts over profitability; the 2023–2026 rerating reflects the evidence that Nu can generate very large profits. The most useful historical conclusion is that the valuation regime itself changed.

P/TBV offers a cleaner expectation test. Under the classic steady-state clean-surplus relationship,

justified P/TBV = (ROE − g) / (cost of equity − g).

At the current 5.66-times tangible-book multiple and a 12.5% USD cost of equity, a perpetual 6% growth rate would require an implausibly high roughly 43% sustainable ROE. A perpetual 8% growth assumption lowers the implied ROE into the mid-30s. A 9% perpetual growth assumption gets close to 29%. The market therefore cannot be rationalized as a mature bank today; it requires a prolonged high-growth transition before a lower terminal multiple.

My USD cost of equity is deliberately higher than that used for a mature U.S. digital bank. The base case uses 12.5%, the conservative case 13.0% and the optimistic case 11.5%. These are valuation assumptions rather than company guidance. They incorporate emerging-market sovereign and currency exposure plus a control premium for the dual-class structure. Brazil’s current 13.75% local policy rate is not itself the USD cost of equity, but it illustrates the macro-risk regime in which Nu earns its returns.

My residual-income model starts from $12.09 billion of tangible equity and asks how much value the future ROE spread creates above that cost of equity. The conservative path assumes approximately 28% near-term sustainable ROE, 16% tangible-book growth during the high-growth period, 13% cost of equity, and a terminal ROE around 22% with 5.5% long-term growth. That produces roughly $10.2 per share. The base path assumes roughly 30% high-growth ROE, 18% tangible-book growth, a 12.5% cost of equity and eventual 24% terminal ROE with 6% growth; that produces about $14.5. An optimistic path of roughly 32% ROE, 20% book growth, 11.5% cost of equity and 26% terminal ROE produces about $24 per share.

I intentionally cross-check those residual-income values against P/E rather than accepting them mechanically. For the conservative case I use normalized forward EPS around $0.75 and a 14–15-times multiple. The base uses roughly $0.95 and 16–17 times. The optimistic case assumes about $1.10 and 19–20 times. The blended ranges below reflect both methods.

This is valuation-scenario analysis within a research framework, not investment advice.

The conservative scenario requires the tax rate to normalize, credit losses to stay structurally higher than bulls expect and the market to demand a more conventional emerging-market bank multiple; it does not require a recession. The base requires Brazil to sustain high-20s or low-30s ROE and Mexico to start converting deposits into profitable credit without a large deterioration in group losses. The optimistic outcome requires both those things plus meaningful value creation outside Brazil.

A country-option cross-check reaches a similar answer, although Nu does not disclose enough country-level profit to make this the primary methodology.

This is deliberately rough. Brazil is worth most of the current equity. Mexico has meaningful option value because customer scale and a banking license already exist. Colombia is smaller. The U.S. and Nu Global deserve little present value until customer-acquisition and credit economics become observable.

The most important sustainable-ROE sensitivities are now quantifiable.

A one-percentage-point increase in annual credit-loss formation on the $39.4 billion portfolio costs about $394 million pretax. At a normalized 28% tax rate that is roughly $284 million after tax, equivalent to a little over two percentage points of June equity ROE.

Tax is similar in magnitude. Raising Q2’s 14.2% IFRS effective rate to 27.5% on its existing pretax earnings knocks roughly five percentage points off annualized ROE. Raising a normalized 25.8% full-year-style rate to 30% would have a much smaller effect, roughly 1½ to 2 ROE points depending on pretax earnings.

If assets and deposits reprice together, the direct Selic effect is smaller than many headline analyses imply. On the simplified Q2 balance sheet a 100-basis-point cut costs about $453 million of gross float yield while saving up to $399 million of funding expense. The credit-loss response can overwhelm that $54 million direct gap. This is why the direction of risk-adjusted NIM matters more than float income alone.

Margin-of-safety verdict: none. The current $14.16 price is about 26–42% above the $10.0–11.2 conservative intrinsic range. Nu can still deliver a good return from this price, but that return depends on the base or optimistic scenario occurring. The conservative scenario provides no downside cushion.

Sustainable ROE is the most fragile base-case assumption. Cutting the base 29–30% assumption to 70% of its level means roughly 20–21% ROE. At the same 12.5% cost of equity, the residual-income premium collapses; my blended value falls to approximately $9–11 per share, roughly 25–35% below the current quote. The reason is mathematical: at 5.7 times tangible book, the equity has far more sensitivity to the duration of excess ROE than to a small difference in near-term EPS.

The flat-earnings test is harsher. If IFRS earnings remain unchanged for three years, Nu pays no meaningful dividend and the exit P/E remains exactly where it is today, the investor’s annualized return is approximately 0%. A flat-earnings return of about zero is below a positive 10-year government-bond yield. There is no margin of safety at this buy price under that test.

Permanent-loss risk concentrates in five variables.

Credit is medium-to-high probability and high impact. The observable indicators are 90+ NPLs, Stage-2 migration, write-offs and ECL as a percentage of average credit. If 90+ NPLs move above roughly 8% while annualized write-offs remain in the low-to-mid teens, the market will stop capitalizing 30%-plus ROE and begin treating current margins as compensation for delayed losses.

Tax normalization is high probability and medium-to-high impact. Q2’s 14.2% rate is already visibly below 2025’s 25.8%, while statutory changes point upward from 2028. A consolidated effective rate persistently above 30% would lower sustainable ROE and P/TBV simultaneously.

International overinvestment is medium probability and medium impact initially, potentially high over several years. The observable indicators are the efficiency ratio, Mexico loan-to-deposit conversion, U.S. customer-acquisition economics and capital allocated to new jurisdictions. A few hundred million dollars of annual investment is affordable today. A multi-billion-dollar build that does not produce deposits and credit at Brazil-like unit economics would consume much of the incremental value investors currently assign to options.

FX is medium probability and medium impact. Q2 reported growth enjoyed a large translation tailwind relative to FX-neutral growth. A renewed BRL depreciation can reduce reported USD earnings and book value even if local operating metrics continue improving. September 22’s BRL/USD rate around R$5.11 should not be mistaken for a stable conversion rate.

Founder control is low probability of acute failure but high potential impact. With roughly 74% voting control, a major capital-allocation mistake cannot be corrected quickly by ordinary Class A holders. The observable indicator is the size and return profile of acquisitions, foreign expansion and related governance decisions, not the share price.

Positive catalysts over the next year are concrete. A peak in 90+ NPLs while risk-adjusted NIM remains above 11–12% would validate the underwriting expansion. Mexico loan growth that lifts the 35% loan-to-deposit ratio without sharply increasing early arrears would establish the second country as a real profit engine. A normalized tax rate in the mid-20s rather than around 30% would raise sustainable ROE. Approval of the Brazilian bank acquisition would remove a regulatory uncertainty. Completing the remaining roughly $500 million buyback below intrinsic value would increase per-share earnings. The December 8 Investor Day can add value if management quantifies international investment and return hurdles rather than simply expanding the TAM narrative.

Negative catalysts are equally measurable. Stage-2 balances migrating into Stage 3, another sequential rise in 90+ NPLs above 7%, an effective tax rate normalizing above 30%, a sustained efficiency ratio above 23%, or Mexico remaining heavily overfunded while its loan book fails to scale would attack the high-ROE narrative. U.S. growth purchased through high APYs, cashback and marketing without disclosed contribution economics would raise the same concern on a longer clock.

† Q2 expense annualized against period-end gross credit, used only as a monitoring approximation.

Nu had not announced the Q3 2026 earnings date through the company materials reviewed; third-party earnings calendars estimated around November 12, 2026, so that date should be treated as provisional until Nu confirms it. The first Investor Day is confirmed for December 8, 2026 in New York.

Cross-synthesis, investment conclusion, uncertainties and sources

Vertically, Nu has proved one capability beyond reasonable dispute: it can take an initially narrow financial product, acquire tens of millions of customers through digital distribution, turn those customers into active account holders, turn their balances into cheap funding, and then monetize the relationship through credit and services while maintaining an exceptionally low operating-cost ratio. The move from a card startup to a $45 billion deposit franchise earning more than $1 billion in a quarter required product execution, funding development, underwriting and cost discipline over more than a decade; it was not the result of a single favorable rate cycle.

Some of the current profitability is cyclical. A high Selic supports the yield on liquid assets. BRL strength boosted reported USD growth in Q2. A 14.2% IFRS effective tax rate is exceptionally favorable. Credit cohorts added during a risk expansion have not all reached terminal loss behavior. Those conditions can change without invalidating the business.

Underneath them lies the durable capability: cost-to-serve. A branchless institution with 83.5% group activity and 86%+ activity in Brazil can spread product, technology and compliance expenditure across a customer base approaching 140 million. That is why the operating model can tolerate expensive credit losses and still earn high ROE.

Horizontally, Nu’s biggest advantage over incumbents is distribution economics. Itaú has greater product breadth, collateralized lending capability and institutional depth. MercadoLibre has richer commerce transaction data. SoFi faces the U.S. market directly. Kaspi has an even tighter super-app integration in its core market. Nu’s distinctive asset is the combination of mass consumer scale, deposits and a branchless cost base across the two largest Spanish/Portuguese-speaking consumer markets it is targeting.

Its weakness is credit concentration. Nearly two-thirds of the current card-plus-loan book is credit cards, with most of the remainder unsecured lending. Secured lending is still small. That makes Nu capable of exceptionally high spreads but leaves earnings sensitive to household stress and underwriting drift.

Investors are paying for the proven advantage before they have proof that it transfers internationally. Brazil alone supports a large part of today’s equity value. Mexico is far enough along to deserve real option value because it already has 16 million customers, $5.7 billion of deposits and a full banking operation. Colombia deserves smaller option value. The U.S. and Nu Global deserve very little until unit economics become visible.

The most likely market misjudgment is subtler than “Nu is overvalued” or “Nu is cheap.” Investors looking at the roughly 19-times P/E can underestimate how much long-duration ROE is embedded in a 5.7-times tangible-book multiple. Those looking only at P/TBV can underestimate how quickly retained earnings can expand tangible book when ROE remains around 30%. The stock is cheap if high-20s ROE lasts a decade and expensive if it lasts only another two or three years.

That is why tax and credit matter more than the U.S. launch to the 12-month valuation. A five-point sustainable-ROE swing from tax normalization is worth billions of dollars in residual income. A one-percentage-point swing in credit-loss formation moves annual after-tax profit by roughly $0.28 billion. The first few hundred thousand U.S. accounts cannot compete with those sensitivities yet.

Over one year, I would focus on four numbers: 90+ NPLs, risk-adjusted NIM, effective tax rate and Mexico’s loan-to-deposit ratio. If NPLs peak around current levels, risk-adjusted NIM stays in double digits, taxes normalize below 30% and Mexico starts deploying deposits into good credit, the base case strengthens materially.

Over three years, the important question becomes country replication. Mexico needs to generate a recognizable return on its own equity and operating cost. Brazil needs to deepen wallet share enough to offset slowing customer additions. Secured lending needs to become a meaningful counterweight to unsecured exposure. International spending needs to remain small relative to Brazilian excess returns.

Over five years, Nu’s fate turns on whether its moat is a Brazil-specific response to a historically poor incumbent customer experience or a transportable operating system for retail banking. Successful replication in Mexico would be strong evidence for the latter. U.S. success would be far more powerful evidence, but also much harder.

Bull reasons:

• Q2 reported ROE reached 33% while risk-adjusted NIM rose to 12.4%, showing that current credit pricing is generating large excess returns even after recognized credit cost.
• Customer growth has slowed to 13% while ARPAC grew 22% FX-neutral, which is the desired transition from account acquisition toward wallet monetization.
• Deposits of $45.3 billion exceed the $39.4 billion gross credit portfolio, giving Nu a strong retail-funded balance sheet rather than dependence on expensive wholesale funding.
• Mexico already has roughly 16 million customers and a banking license, while management’s same-stage ARPAC comparison indicates monetization is developing faster than Brazil did.
• At about 19 times trailing IFRS earnings, Nu trades at a much lower earnings multiple than high-growth public fintech platforms such as MercadoLibre, SoFi and Robinhood despite much higher current ROE.

Bear reasons:

• The 90+ NPL ratio has risen to 6.9% and H1 card-plus-loan write-offs increased about 54% year on year while management is deliberately moving into higher-risk segments.
• Q2’s 14.2% IFRS tax rate is well below the 25.8% FY2025 rate and benefited from a $671 million deferred-tax offset, meaning reported 33% ROE overstates a reasonable tax-normalized run rate.
• At roughly 5.7 times tangible book, the stock requires high excess ROE to persist for many years; a drop toward 20% sustainable ROE produces material valuation compression even if the company remains profitable.
• Mexico is only 35% loan-to-deposit and the U.S. launched into an intensely competitive market, so much of the international customer scale has not yet proved Brazil-like return economics.
• Founder voting control of roughly 74% leaves ordinary Class A shareholders unable to correct an international capital-allocation mistake through normal shareholder influence.

Pre-mortem script one: during 2027 Brazil’s household-credit cycle deteriorates after the deliberate 2025–26 move into higher-risk borrowers. Nu’s 90+ NPL rises from 6.9% to above 9%, annualized write-offs move into the mid-teens, and IFRS ECL formation rises by roughly 300 basis points of loans. Risk-adjusted NIM falls below 8%. At the same time the effective tax rate settles around 30%. Sustainable ROE drops below 20%. Investors reprice Nu from roughly 19 times earnings to 10–12 times normalized EPS around $0.60–0.70, putting the stock around $6–8, a decline of roughly 45–60% from the current price.

Pre-mortem script two: Mexico accumulates customers and deposits through 2027–28 but its loan-to-deposit ratio remains below 45% because underwriting cannot scale at acceptable losses. Nu simultaneously spends heavily to acquire U.S. customers through deposit yield, cashback and advertising. The consolidated efficiency ratio stays above 25% for four quarters, group customer growth falls below the mid-teens and ROE settles near 22%. The company remains profitable, yet the market stops paying a growth-bank premium and compresses P/TBV toward roughly 3 times. Even with book-value growth, a price around $9–11 becomes plausible.

Of the two, the first script is the one most likely to produce a permanent 50% loss. The second is more likely to produce several years of opportunity cost and multiple compression than insolvency.

My research judgment changes positively if the evidence establishes three things together: 90+ NPLs peak below roughly 7.5%; tax-normalized ROE remains above 28%; and Mexico raises loan-to-deposit above 50% while keeping early delinquencies controlled. At that point the 5.7-times tangible-book multiple would rest on more than Brazil.

I would overturn the thesis negatively if Stage-2 and 90+ deterioration persist through two additional quarters while risk-adjusted NIM falls below 9.5%; if normalized ROE falls under 24% without a recession; or if international operating expenditure pushes the efficiency ratio above 25% without measurable country-level returns.

Nu is a very good operating company at a price that already recognizes much of what has been proven. Brazil merits a premium bank multiple because its combination of scale, cost efficiency, engagement and ROE is unusual. Mexico deserves option value. The U.S. and stablecoin initiatives warrant curiosity, not much current valuation. The share price offers a plausible base-case return but almost no conservative-case cushion.

At $14.16 I would own rather than chase the stock. The principal reason is valuation asymmetry: approximately 19-times trailing earnings looks manageable, while 5.7-times tangible book leaves significant downside if sustainable ROE falls from around 30% into the low-20s. A better entry price would let the investor receive Mexico and U.S. optionality without paying in advance for most of their success.

【Company-profile scores】

• Fundamental quality: high
• Growth: high
• Moat: medium
• Financial soundness: strong
• Management credibility: high
• Valuation attractiveness: medium
• Risk level: high
• Suitable investor type: long-term growth

【Investment rating】

• Rating: Hold
• One-line thesis: Brazil’s 33% reported ROE is exceptional, but 6.9% 90+ NPLs, tax normalization and 5.7x tangible book leave limited downside protection.
• Current-price classification: acceptable hold
• Whether to wait for a better price: yes. I would require approximately $8 or less for the template’s strict 20%-plus margin below conservative intrinsic value; a less demanding growth investor could reassess near $11–12 if credit and tax evidence improves.
• Target holding horizon: 3–5 years
• Expected annualized return, conservative scenario: roughly 0–4% over 3–5 years, with a negative 12-month mark-to-value result possible
• Expected annualized return, base scenario: roughly 13–16% if earnings compound and the multiple normalizes gradually
• Expected annualized return, optimistic scenario: roughly 22–27% while Mexico becomes a second high-return engine
• Max-loss risk: approximately 50–60%, to roughly $6–8, under the credit/tax/multiple-compression pre-mortem
• Reassessment trigger: 90+ NPL above 7.5% for two consecutive quarters
• Reassessment trigger: risk-adjusted NIM below 9.5%
• Reassessment trigger: tax-normalized ROE below 24%
• Reassessment trigger: Mexico loan-to-deposit above 50% with stable early delinquencies would be positive
• Reassessment trigger: efficiency ratio above 25% for four quarters without corresponding international revenue acceleration would be negative

【Ideal Buy Price】7.50–8.00 USD

Basis: the conservative blended intrinsic range is approximately $10.0–11.2; the upper buy signal is capped at least 20% below the conservative floor, following the assignment’s unusually strict margin-of-safety rule.

【Valuation Range】

• current: 14.16 (close as of 2026-09-22)
• bear (conservative · ideal buy zone): [7.50, 8.00]
• base (fair · acceptable hold zone): [13.00, 17.70]
• bull (optimistic · above the clearly-overvalued line): [26.60, 29.00]

The bear band is intentionally much lower than the conservative intrinsic estimate because the requested “ideal buy” rule requires an additional 20%-plus margin below conservative value. The base band is centered on the $14.5–16.2 base intrinsic range with approximately ±15% tolerance. The clearly-overvalued threshold begins roughly 10% above the $24.2 optimistic residual-income endpoint.

Research uncertainties are material in five areas. First, June 30 is the latest company-disclosed treasury-adjusted share count I could quantify; additional post-June buybacks may have reduced shares modestly by September 23. Second, Nu does not disclose country-level profit and capital in enough detail to make the Brazil/Mexico/Colombia/U.S. option values independently auditable. Third, the interim credit filing does not provide complete origination-vintage and renegotiated-loan tables, limiting a true cohort stress test. Fourth, current incumbent-bank disclosures do not put card and unsecured charge-offs on definitions identical to Nu’s IFRS staging and operational NPL measures; I declined to fabricate a like-for-like table. Fifth, U.S. and Nu Global launched only on September 10, leaving essentially no public unit-economics history.

Source hierarchy: the valuation and accounting work relies principally on Nu’s August 13, 2026 SEC-furnished Q2 earnings release, IAS 34 interim financial statements, credit and tax notes, and the separate managerial P&L reconciliation with KPMG’s limited-assurance report. Governance and historical information relies on the FY2025 Form 20-F filed April 8, 2026. Insider-sale analysis uses the underlying SEC Form 144 filings. Current strategic developments use Nu’s September 2 Investor Day notice and September 10 U.S./Nu Global disclosure. Current prices use September 22 market data; Brazilian rates and FX use September 22 reporting on the latest Copom minutes and market close.

Other tickers mentioned

• MELI.US: MercadoLibre and Mercado Pago are Nu’s most important listed digital-commerce/payments competitor in Brazil and Mexico.
• KSPI.US: Kaspi.kz is the closest listed emerging-market super-app analogue with banking, payments and marketplace economics.
• SOFI.US: SoFi provides the most relevant listed digital-bank comparison for Nu’s U.S. entry.
• XP.US: XP is a Brazilian financial-distribution peer useful for Selic, BRL and investment-platform exposure.
• HOOD.US: Robinhood illustrates U.S. retail-finance platform valuation and trading/crypto optionality, though it is not a bank analogue.
• ITUB.US: Itaú Unibanco is the most important incumbent Brazilian bank valuation and competitive benchmark.
• BBD.US: Bradesco is a major Brazilian universal-bank incumbent competing across consumer credit and deposits.
• BSBR.US: Santander Brasil is a large incumbent consumer lender and deposit competitor in Nu’s core market.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.