Nu Holdings — Blitz Index 65 (B)

ROE 33% and 39–56% FXN growth across 8 quarters — Mexico repeats the Brazil S-curve while the credit cycle watches.

Last re-scored: October 17, 2026.

Score breakdown

Think we scored it wrong? Score it yourself →

Radar axes (0–100): Moat 63.3 · Speed 84 · Creative 93.3.

Watch item

Gates: G1 ✓ · G2 ✓. Exactly on the 65 line — first score. A 73-point business minus 8 points of governance and credit-cycle demerits.

Strengths

• 8 straight quarters of 39–56% FXN growth with quarterly net income topping $1B (Q2 2026)
• Mexico 15.8M customers at breakeven — the Brazil S-curve repeating (ARPAC $12.3, 2.2x Brazil at the same stage)
• ROE 33% + risk-adjusted NIM 12.4%, both all-time highs — best-in-class banking profitability
• Funding cost at 87% of the interbank rate + $45.3B deposits — deposits are the blitz fuel

Concerns

• 90-day+ delinquency 6.9% (+35bp YoY) with Brazilian household debt at record highs — the credit cycle is the weather
• 85% interest-income dependence — rate and macro shocks flow straight into NII
• Vélez ~74.3% voting control + controlled-company status — founder rule without checks
• Brazil adult penetration ~60% — domestic customer-count growth is capped, monetization must carry it

Research notes

Nu Holdings Ltd. (NYSE: NU), the company behind the Nubank brand, is a digital bank built across Latin America with no branches at all, and its first BlitzLab scoring lands exactly on the line that matters most in this system: 65 out of 100, grade B, the threshold score itself. The arithmetic is 21 for Speed plus 38 for Moat plus 14 for Creative Destruction, minus 8 in Demerits, for a total of 65. Before demerits the business scores 73; the deductions are what pull it back to the boundary, a shape the score file itself compares to NET (62/B), another company whose demerits do the dragging. There is no FLAG: demerits of 8 sit well below the 15-point FLAG trigger. Both gates pass, so the scoring proceeds in full — G1, sustained growth or a second growth curve, as a confirmed pass, and G2, a new-market ten-year ceiling at least ten times current revenue, as an estimated pass. This is a research assessment under BlitzLab v1.0 (frozen), not investment advice, and the business score does not look at price.

Two reading rules from the score file govern everything that follows, and they are worth stating plainly because Nu is not an ordinary company. First, Nu is a bank-structure company, so deposits and provisions do not mean what debt and costs mean for a non-financial firm; the score file applies its own canonical rule here, treating deposits as customer funds rather than corporate borrowing. Second, Nu reports under IFRS as a Cayman Islands exempted company and foreign private issuer filing 20-F and 6-K forms, so the canonical figures in this draft are audited IFRS numbers from the 20-F, while the growth narrative is carried on the company's managerial, foreign-exchange-neutral (FXN) basis. Where the score file marks a figure as estimated or press-reported, this draft keeps that label. Confidence in the scoring is rated high, because it rests on SEC EDGAR 20-F and 6-K texts reviewed directly, with the company's ticker identifier confirmed from SEC company tickers data, while share price, market capitalisation and analyst material are press-reported. This is also a first scoring, with no blind phase behind it.

1. Current Position: a threshold score, earned from opposite directions

The headline tension in this score is that its two halves point in opposite directions. On the business side, Nu looks like a textbook blitz: creative destruction scores 14 of 15, speed scores 21 of 25, and the company has already become, by customer count, the third-largest financial institution in Brazil according to the Brazilian central bank, with about 118 million customers there — roughly 60 percent of the adult population — plus 15.8 million in Mexico and more than 5 million in Colombia, for a total the score file puts at 138.9 million customers. On the deduction side, two structural facts each cost real points: founder David Vélez controls about 74.3 percent of the voting power, and the credit cycle is visibly turning in the wrong direction on delinquency even as profitability reaches record levels. The net result is a company that clears the bar by nothing at all — 65 against a threshold of 65 — and that exactness is itself informative. It says the business, judged on its own terms, is comfortably above the line, and that the reservations are concentrated, nameable, and measurable rather than diffuse.

The growth gate, G1, passes on confirmed evidence. Revenue has grown 39 to 56 percent on an FXN basis for eight consecutive quarters on the managerial measure, from $2.9 billion to $5.88 billion per quarter, while quarterly net income has risen from $553.4 million in the third quarter of 2024 to $1,061 million in the second quarter of 2026, the first quarter above $1 billion, with return on equity at 33 percent. The second growth curve required by G1 is not a plan but a demonstrated repetition: Mexico has reached 15.8 million customers and breakeven, with average revenue per active customer (ARPAC) of $12.3, which the score file notes is 2.2 times the $5.6 that Brazil showed at the same stage of development. That comparison is the heart of the G1 case — the Brazilian formula, transplanted, is reproducing itself faster on monetisation than it did at home.

The market gate, G2, passes as an estimate, and the score file is candid about the thinner margin here. The Latin American banking revenue pool is put at about $418 billion on a 2022 estimate from McKinsey, as cited in press reporting, against which Nu holds roughly a 4 percent share. If that pool grows 7 percent a year, it reaches roughly $770 billion by 2035; a 5 percent share would imply revenue of about $38 billion, which at a price-to-earnings multiple of 20 would imply a market capitalisation near $190 billion, above the bar of about $163 billion set by ten times fiscal 2025 managerial revenue of $16.3 billion. The United States conditional approval from the Office of the Comptroller of the Currency, granted on 29 January 2026 and confirmed in SEC filings, is an additional upside option rather than part of the core case. The constraint the score file names is real: with Brazilian adult penetration already near 60 percent, further domestic customer-count growth is bounded, so the G2 case depends on monetisation, on Mexico and Colombia, and on United States execution actually happening.

2. Moat: distribution and depth, not network effects

Moat scores 38 of 60, or 63.3 on the rescaled radar axis — the weakest of the three business axes in relative terms, and the section where the score file is most discriminating about what kind of defence Nu actually has. The four subscores are network effects 7 of 18, distribution scalability 13 of 18, market size 9 of 12, and gross margin 9 of 12. The pattern is a company whose defences are built from reach and from customer depth, not from the self-reinforcing network that the top of the network-effects anchor describes.

Network effects earn only 7, placed in the 5-to-9 anchor band for data and ecosystem effects, and the score file states the reason directly: direct network effects are weak, and value is created instead through cross-selling and through customers making Nu their main bank. The evidence for depth is a cohort figure: customers of eight years or more show ARPAC of $27.3 per month, against an overall $12.2 in the second quarter of 2025, a multiple of 2.2 times. Long-tenure customers are not merely staying; they are worth more than twice as much per month as the average customer. The premium Ultravioleta tier reinforces the same point, with purchase volume up 41 percent year on year. But the score file refuses to call this a social-graph lock-in, because it is not one: the switching cost is the friction of moving accounts, not membership in a network whose value rises with each additional member. Seven is the score for deepening relationships without a true network effect.

Distribution scalability, at 13 of 18, is the strongest moat subscore and the closest thing in this assessment to a structural advantage. Nu operates zero branches, so the app itself is the distribution network, and customer acquisition cost is described as converging on effectively zero, driven by referral and viral organic inflow on a press-reported basis — the score file notes net additions of about 4 million customers per quarter occurring without marketing. More than 100 new products were launched in fiscal 2025, and the NuCel fifth-generation mobile virtual network operator, launched in May 2026, extends the app further into customers' daily lives. Monthly serving cost per customer is about $1.0, and in Mexico it is already below $1. The discount from a higher score is a genuine friction the file does not wave away: know-your-customer checks and regulatory onboarding are not the same as frictionless digital virality, and every new market re-imposes that onboarding cost. Distribution here is radically scalable but not frictionless.

Market size earns 9 of 12, and the reasoning mirrors the G2 gate with a tighter margin. The same $418 billion pool, grown at an estimated 7 percent a year to about $770 billion by 2035, stands at about 2.6 times the $163 billion bar — a pass, but without a large cushion, and the score file contrasts this with a peer scoring 10 on a pool roughly three times its bar. The United States conditional approval is again the upside option. The full 12 is withheld on two stated conditions: quantification of the Mexican pool and final approval of the United States charter. Until those arrive, 9 records a large market that is proven in outline but not yet fully measured in its second and third geographies.

Gross margin earns 9 of 12 on an audited figure: IFRS gross margin of 42.0 percent for fiscal 2025, computed as $6,625.0 million divided by $15,774.7 million, confirmed in SEC filings. The surrounding profitability measures are at record levels — net interest margin of 22.9 percent in the second quarter of 2026, risk-adjusted net interest margin of 12.4 percent, up 290 basis points year on year and the highest on record, and return on equity of 33 percent, also a record. The revenue mix behind those margins is 85 percent interest income and 15 percent fees, and that mix is exactly why the score stops at 9: for a bank, a 42 percent gross margin is strong, but dependence on interest income ties the result directly to the interest-rate and economic cycle. Margin strength and cycle sensitivity are the same fact viewed from two sides.

3. Speed: fast growth, funded by cheap deposits, run at a fraction of incumbent cost

Speed scores 21 of 25, or 84.0 rescaled — the second-strongest axis — with subscores of growth 9 of 10, capital velocity 4 of 5, hiring and productivity 4 of 5, and share velocity 4 of 5. What distinguishes this speed profile from a typical venture-funded blitz is that the fuel is not raised capital but the banking model itself: cheap deposits, gathered through the app, funding a loan book at a cost advantage over incumbents.

Growth earns 9 of 10 on the combination of scale and durability that few companies sustain together. Audited IFRS revenue for fiscal 2025 is $15,774.7 million, up 37 percent, while the managerial measure is $16.3 billion, up 45 percent on an FXN basis — the two figures differ because one is the audited IFRS canonical number and the other is the company's managerial, currency-neutral narrative, and the score file keeps both visible rather than choosing silently between them. The eight-quarter run of 39 to 56 percent FXN growth, from $2.90 billion in the third quarter of 2024 to $5.88 billion in the second quarter of 2026, is the evidence, with the quarterly figure up 39 percent FXN in the latest quarter. Net income over the same span rose from $553.4 million to $1,061 million, up 49 percent FXN year on year, with return on equity at 33 percent. The score file's comparison is explicit: sustaining growth in the 40 percent range on a $16 billion base, together with profitability, sits one step above a peer scoring 8 whose growth had decelerated. The tenth point is withheld on a stated condition — evidence of acceleration after the Mexican bank launch in July 2026 — so the 9 is a score with a named path upward rather than a ceiling.

Capital velocity earns 4 of 5, and the mechanism is the deposit franchise. Funding cost equals 87 percent of the interbank rate, confirmed for the fourth quarter of 2025, which means Nu funds itself more cheaply than the interbank market itself — the low-cost deposit base is the advantage. Deposits reached $45.3 billion in the second quarter of 2026, up 18 percent FXN. The Common Equity Tier 1 ratio of 13.0 percent sits above the regulatory minimum, and available funding of $38.8 billion is about twice the credit portfolio, so the score file finds no bottleneck in capital access. A $1.0 billion share repurchase authorisation approved on 4 June 2026, funded from retained earnings, is further evidence of capital flexibility. For a blitz, this is an unusual and favourable shape: the faster the customer base grows, the more low-cost funding arrives, rather than the faster growth consuming a finite war chest.

Hiring and productivity earn 4 of 5 on an efficiency story that is unusually well documented. Full-time headcount is 10,027 as of 31 December 2025, confirmed in SEC filings, and the efficiency ratio — the share of revenue consumed by operating costs — is 19.5 percent in the second quarter of 2026, against 40 to 50 percent for incumbent banks, on a trajectory the score file traces from 29.9 percent through 27.7 percent to 19.5 percent. Monthly serving cost per active customer is about $1.0, which is 5.8 percent of ARPAC of $17.1. Artificial intelligence is doing real work inside that ratio rather than decorating it: NuFormer, a 330-million-parameter model used in underwriting, more than 60 percent of consultations handled by AI agents, and an AI Private Banker product with more than 15 million monthly active users. The point is not headcount growth but headcount leverage — a customer base approaching 139 million served by about ten thousand employees.

Share velocity earns 4 of 5 on numbers verified by the regulator rather than claimed by the company. Brazil's roughly 118 million customers make Nu the third-largest financial institution in the country by customer count on central bank data, confirmed, and its 6.8 million small-business customers rank first. Mexico's 15.8 million customers rank third, and Colombia adds more than 5 million. Activity rates show these are not dormant accounts: 83.5 percent overall, and above 86 percent in Brazil. The score file's judgement is that quantified share, demonstrated by central bank statistics, supports a strong score — while the remaining point presumably awaits the same depth of proof in the newer markets that Brazil already provides.

4. Creative Destruction: the highest axis, and the clearest story in the file

Creative destruction scores 14 of 15, or 93.3 rescaled — the strongest axis by a wide margin — with replacing the existing order at 5 of 5, irreversible behaviour change at 4 of 5, and cost-structure change at 5 of 5. If any single section justifies calling this a textbook case, it is this one, because all three subscores rest on measured facts rather than on promises.

Replacing the existing order earns the full 5 because the target is named, large, and already partly displaced. The incumbents are Brazil's established banks — Itaú, Bradesco, Banco do Brasil, Caixa and Santander Brasil — operating on a high-fee, branch-centred model, and Nu already ranks third by customer count against them on central bank data. The financial-inclusion evidence is quantitative: 52 percent of Nu's customers received their first-ever credit limit from Nu, and 35 percent of Mexican customers opened their first-ever bank account with Nu in the second quarter of 2026. Free transfers through Pix and a zero-branch model do not merely compete with the incumbents' fee pool; they remove the tollbooths the pool was built on. A full score here means the replacement is not hypothetical — it is already the third-largest institution in the market doing the replacing.

Irreversible behaviour change earns 4 of 5, with the discount marking geography rather than doubt. In Brazil, the shift to mobile-first banking has reached what the score file calls the irreversible stage: an activity rate above 86 percent, deposits up 41 percent FXN in the second quarter of 2025, and salary and payment inflows showing customers making Nu their main account. Behaviour has changed where the product has been present longest. The withheld point reflects Mexico and Colombia, which are still in the early stage of transition — breakeven and bank launch, respectively, rather than entrenched habit. The score therefore describes a behaviour change that is complete in one country and in progress in two others, which is also an accurate description of the company's stage as a whole.

Cost-structure change earns the full 5 on the starkest comparison in the file. An efficiency ratio of 19.5 percent against 40 to 50 percent for incumbents means Nu operates at less than half the cost burden of the order it is replacing, and the per-customer monthly cost of about $1.0 is estimated at one-fifth to one-tenth of incumbent levels. AI agents and NuFormer in underwriting and customer service are expanding automation further, so the cost advantage is on a trajectory that deepens rather than a snapshot that might erode. This is the unit-cost collapse that the creative-destruction framework looks for, demonstrated in audited and reported figures rather than asserted. Taken together, the three subscores describe a disruption that has already happened in Brazil — the old order replaced, behaviour changed, costs collapsed — and is now being repeated elsewhere.

5. Deductions & Risks: two structural deductions and one cyclical warning

Demerits total −8, composed of debt 0 of −10, revenue quality −3 of −10, governance −4 of −5, and regulation and social cost −1 of −5. The composition matters more than the total, because the zeros and the deductions tell different stories: the balance sheet carries no debt trigger at all, while governance and credit quality each cost points for reasons that are structural and therefore persistent.

Debt scores 0, and the score file insists on the bank-structure reading that makes that zero correct. Deposits of $41,925.1 million, as cited from the 20-F, are customer funds, not corporate borrowing, under the score file's canonical rule — treating them as debt would be a category error for a bank. Real corporate leverage is borrowings and financing of $4,398.2 million against equity of $11,321.6 million, a ratio of about 0.4 times, confirmed in SEC filings. Credit-card network payables of $13,633.9 million are settlement in nature, not borrowing. The score file is explicit that the risk in this company is not leverage but the credit cycle, which is scored under revenue quality instead. A reader who applied a non-financial debt ratio to Nu would manufacture a crisis the filings do not show; the file's treatment avoids that error, and this draft preserves it.

Revenue quality, at −3, is where the credit cycle enters the score. The 90-day-plus delinquency rate is 6.9 percent in the second quarter of 2026, up 35 basis points year on year, reported in the 6-K. The score file notes the mitigating context — the rise reflects seasonal migration from first-quarter early delinquency — but records the direction as worsening, and that direction is what costs points. Two further factors compound it: Brazilian household debt is at a record high on a press-reported basis, which raises sensitivity to any credit-cycle turn, and interest income dependence of 85 percent in fiscal 2025 ties revenue directly to rates and the economy. The loss-provision balance has risen from $6.1 billion in the first quarter to $6.6 billion. Against those pressures, the file records two genuine offsets: the 15-to-90-day delinquency rate improved to 4.8 percent from 5.0 percent in the first quarter, and risk-adjusted net interest margin reached its record 12.4 percent. The −3 is therefore a measured warning, not an alarm: credit is deteriorating at the margin while the pricing of that credit is, for now, more than compensating.

Governance, at −4, is the largest single deduction and the most structural. Class B shares carry 20 votes each against 1 vote for each Class A share, and founder David Vélez holds 88.3 percent of Class B, giving him about 74.3 percent of total voting power as of 31 December 2025, confirmed in the 20-F. The score file checks this against its own anchors — a peer at about 76 percent voting control also scores −4, and another at 58.8 percent scores −3 — so the deduction is consistent with the model's treatment of comparable structures. Two further structural facts accompany it: Nu is a controlled company under New York Stock Exchange standards, which waives independent-director requirements, and a shareholders' agreement gives holders of more than 40 percent of voting power the right to nominate a majority of the board. A sale of 33 million shares by Vélez in August 2025, reported in the press as for personal wealth management, is explicitly noted as not a change of control. The deduction is for entrenchment of founder control, not for any alleged misconduct, and its structural nature means it will not fade with a good quarter.

Regulation and social cost, at −1, is mild, but the single item behind it is concrete. Nu operates under multiple regulators: Brazil's central bank and National Monetary Council apply Basel III in full from January 2025, Mexico's conversion from a SOFIPO to a bank is complete with a capital ratio of 15.4 percent, Colombia's regulator reports 16.9 percent, and the United States conditional approval would bring a federal regulatory framework on final approval. The scored cost is a $186.4 million prepayment in March 2026 to the Brazilian deposit guarantee fund's emergency capital top-up, confirmed in SEC filings — a real cost imposed by bank failures elsewhere in the system, landing on a healthy bank's accounts. No securities litigation or fines have been confirmed, which is why the deduction stays at −1 rather than growing. It records a system in which even a well-run bank pays for its neighbours' failures.

6. What to Watch: the blitzscaling lens, and the triggers that would move this score

Read through the blitzscaling lens the score file itself applies, Nu is a destruction-plus-distribution blitzer. Its main levers are creative destruction — the replacement of the incumbent model, scored 5 of 5 — and distribution, the zero-branch, near-zero-acquisition-cost network scored 13 of 18. Its secondary lever is capital, but capital in an unusual form: low-cost deposits are the blitz fuel, so growth generates its own funding rather than consuming it. Its weak lever is network effects, where lock-in comes from main-account depth rather than from a social graph. The file's summary phrase for the resulting shape is a compound S-curve blitz: not a single market exploding once, but a matrix of regions and products exploding in sequence, with time lags between them. Brazil is in the harvest phase of its first blitz act, with penetration near 60 percent and growth now driven by monetisation; Mexico has just opened its second act with the bank launch; the United States is a third-act option behind the conditional approval. The file also names the characteristic failure mode for this shape, and it is worth repeating exactly because it reframes the risk section: the scenario is not slow growth but fast credit deterioration. With delinquency rising, household debt at a record, and 85 percent dependence on interest income, an economic downturn or sharp rate move would transmit through net interest income into provisions and then into return on equity in sequence, while founder control makes the response fast but unchecked.

The triggers that would move this score are mostly stated as conditions inside the subscores themselves, and they are best watched as a checklist rather than as a forecast. On the upside, growth could reach 10 of 10 on evidence of acceleration following the July 2026 Mexican bank launch, and market size could reach 12 of 12 on quantification of the Mexican pool together with final approval of the United States charter. On the risk side, the direction of the 90-day-plus delinquency rate from 6.9 percent, the provision balance from $6.6 billion, and the 15-to-90-day rate from its improved 4.8 percent will show whether the credit warning in this scoring deepens or stabilises; Brazilian household debt levels, press-reported at a record, set the background for that reading. The governance deduction will persist unless the voting structure itself changes, which no filing in the score file suggests, and further deposit-guarantee or resolution costs of the kind seen in the $186.4 million prepayment would test the regulation score. The next earnings release — the third-quarter 2026 results, which fall after T and are therefore excluded from this scoring — is the next scheduled moment when these triggers get fresh evidence, and the score file's own discipline applies: anything after 17 October 2026 belongs to the next scoring, not to this one. For now, the judgement stands as the file states it: a business scoring 73, a threshold score of exactly 65 after deductions, and a company whose second growth curve is already reproducing the first — with the credit cycle as the weather it cannot control.

Head-to-head

Nu Holdings vs SoFi Technologies — side-by-side scorecard comparison.

More Platforms companies

Other Platforms names on the Blitz Index, highest score first: Alphabet (83/A), Meta (78/B), Airbnb (76/B).

Next earnings

Nov 12, 2026 · After close · Confirmed

Coverage

Nu Holdings Ltd. · Q3 2026 due Nov 12, 2026 (expected) · Score 65 · Grade B

Embed this score

Copy the HTML below to show this score on your site. The badge links back to this page and updates automatically when the score changes.

<a href="https://blitzlab.app/companies/nu/"><img src="https://blitzlab.app/badge/nu.svg" alt="Nu Holdings Blitz Index 65/B — BlitzLab"></a> Copy HTML

Preview:

BlitzLab scores businesses against a scholar’s standard — not investment advice. No price targets, no buy/sell recommendations. Figures are from SEC filings (XBRL) as labeled in the source scorecard.