Nu Holdings (NU) Blitzscaling Score: 65/100 (B) — Moat, Speed & Risk

The business scores 73 — a textbook creative destruction of Latin American banking, growing 40% a year at a 33% return on equity. Then governance and the credit cycle take 8 points back, and the total lands exactly on the line.

11 min read · Blitz Index v1.0 · 2026Q4 · T = October 17, 2026 · First scoring · Not investment advice

The business in one paragraph

Nu Holdings is Nubank, the digital bank of Latin America, headquartered in São Paulo and listed in New York. It has 138.9 million customers — about 118 million in Brazil, where it is the third-largest financial institution by customer count according to the central bank, with roughly 60% of the adult population as customers; 15.8 million in Mexico, also third; and more than 5 million in Colombia. It has no branches at all. In the second quarter of 2026, revenue reached $5.88 billion on a managerial basis, up 39% year over year in FX-neutral terms, and net income passed $1 billion in a quarter for the first time — $1,061 million, up 49% FX-neutral year over year, at a 33% return on equity. That was the eighth consecutive quarter of 39–56% FX-neutral revenue growth. Audited IFRS revenue for fiscal 2025 was $15.77 billion, up 37%, with net income of $2.87 billion. This is a first scoring: there is no blind history and no estimated series behind it, only the measured quarter.

Two numbers frame everything else. The efficiency ratio — operating cost as a share of net revenue — is 19.5%, against 40–50% at incumbent banks. And the cost of serving an active customer is about $1.00 a month, against monthly average revenue per active customer of $17.10. A bank that runs at half the cost of its competitors, adds about 4 million customers a quarter almost entirely through referrals, and earns a 33% return on equity is not a story about a good app. It is a story about a different cost structure meeting a protected, high-fee incumbent order — which is precisely what the Creative-destruction axis is built to measure.

Why it scored 65: three drivers

1. A textbook creative destruction — C scores 14 of 15. This is the highest creative-destruction score in the 2026Q4 batch, and the evidence is unusually concrete. The order being displaced is the incumbent Brazilian bank: high fees, branch-centred, with Itaú, Bradesco, Banco do Brasil, Caixa and Santander Brasil as the named incumbents. Nu is already third by customer count, and the financial-inclusion evidence is quantified rather than asserted: 52% of its customers received their first-ever credit limit from Nu, and 35% of its Mexican customers opened their first-ever bank account with it. Behaviour change (C2) earns 4 — mobile-first banking is irreversible in Brazil, where the activity rate is 86% and deposits are flowing in as salary and payment money, though Mexico and Colombia are still early in the transition. Cost-structure change (C3) earns the full 5: an efficiency ratio of 19.5% against 40–50% at incumbents, a per-customer monthly cost of about $1.00 — roughly a fifth to a tenth of the incumbents' — and AI automation deepening the advantage, with more than 60% of customer chats handled by AI agents and the NuFormer underwriting model running on 330 million parameters.

2. The second growth curve is being reproduced in the financials. The growth gate (G1) passes on the Mexico evidence: 15.8 million customers, break-even reached, and average revenue per active customer of $12.30 — 2.2 times what Brazil showed at the same stage of development ($5.60). The Brazil formula is being replayed, not merely announced. Speed (A) is the strongest axis at 21 of 25: growth (A1) earns 9 on eight consecutive quarters of 39–56% FX-neutral growth at a $16 billion revenue base, with profitability alongside it — one step above a company like Robinhood, whose growth had already decelerated. Capital velocity (A2) earns 4, and here the banking structure matters: the "fuel" is deposits, not equity raises. Funding costs just 87% of the interbank rate, deposits total $45.3 billion (up 18% FX-neutral), the CET1 capital ratio is 13.0%, available funding of $38.8 billion is about twice the credit portfolio, and a $1.0 billion buyback was approved in June 2026. Hiring and productivity (A3) earns 4 on 10,027 employees and an efficiency ratio that has improved from 29.9% through 27.7% to 19.5%. Share velocity (A4) earns 4 on central-bank-verified share: third in Brazil, third in Mexico, and the number-one SME bank in Brazil with 6.8 million business customers, with an activity rate of 83.5% — accounts that are used, not merely opened.

3. The business is 73; governance and the credit cycle take 8 back. Add the three positive axes and Nu scores 73 — comfortably above the line, the profile of a strong blitzscaler. The deductions are where this article's real subject lies. Governance (D3) takes −4: Class B shares carry 20 votes each against 1 for Class A, and founder David Vélez holds 88.3% of the Class B shares, giving him about 74.3% of total voting power. Nu is a controlled company under NYSE rules, exempt from independent-director requirements, and Vélez can nominate a board majority once voting power exceeds 40%. Revenue quality (D2) takes −3: the 90-day-plus delinquency rate is 6.9% and rising, up 35 basis points year over year; Brazilian household debt is at a record high; interest income is 85% of revenue, tying earnings directly to the rate and economic cycle; and the loan-loss provision balance rose to $6.6 billion from $6.1 billion. Regulation (D4) takes −1: Brazil's central bank has fully applied Basel III since January 2025, and Nu prepaid $186.4 million in March 2026 as an emergency capital contribution to the deposit guarantee fund after bank failures in the sector. The total, 73 − 8, is exactly 65.

The scorecard

Total: 21 + 38 + 14 − 8 = 65, Grade B. No FLAG — demerits are 8, below the 15-point FLAG line. The structure mirrors NET (62/B): a business score dragged to the line by its deductions.

The gates

G1 passes on the Mexico reproduction described above. G2 — the new market's ten-year ceiling must be at least ten times current revenue, a bar of roughly $163 billion — passes, but with less room than the growth numbers might suggest. The Latin American banking revenue pool is about $418 billion (2022 estimate, McKinsey), with Nu holding roughly 4% of it; growing 7% a year, the pool reaches about $770 billion by 2035, and a 5% share would imply revenue of about $38 billion. The conditional approval from the US OCC in January 2026 is an additional upside option for a US retail entry. But the constraint is visible in the same paragraph: with adult penetration in Brazil already near 60%, domestic customer-count growth is bounded — future growth must come from monetisation, from Mexico and Colombia, and from execution in the United States.

One reading note matters for a bank scored on a model built for ordinary companies. Deposits of $41.9 billion are customer funds, not corporate debt, so the debt item (D1) is judged on genuine borrowings against equity — about 0.4 times — rather than on a meaningless total-liabilities ratio. Revenue is judged on audited IFRS figures as the canonical source, with managerial FX-neutral growth quoted alongside for the growth narrative. The score file proposes both rules be written into v1.1, because a bank cannot be scored with a non-financial company's ruler.

The other side: what to worry about

The Moat axis is the weakest positive axis at 38 of 60, and the reason is instructive. Distribution (B2) earns 13 — the app is the distribution network, with a 5G MVNO (NuCel, launched May 2026) and more than 100 new products in fiscal 2025 — but network effects (B1) earn only 7. There is no social-graph lock-in; the lock-in is the deepening of the primary-banking relationship, visible in the cohort data (customers of eight years or more generate $27.30 in monthly revenue per customer, 2.2 times the $12.20 average) and in the premium Ultravioleta tier, where purchase volume rose 41% year over year. That is real, but it is not a network effect in the sense the anchor means, and the score says so.

The deeper worry is the one the score file names as the single biggest blitzscaling risk: the credit cycle is the weather of this blitz. Interest income is 85% of revenue. The 90-day-plus delinquency rate is 6.9% and rising, Brazilian household debt is at a record high, and the provision balance is climbing. The offsets are genuine — the 15–90-day delinquency rate improved to 4.8% from 5.0%, and the risk-adjusted net interest margin of 12.4% is the highest on record, up 290 basis points year over year — which is why D2 is −3 rather than worse. But the failure mode here is not slow growth. It is a fast credit collapse: a downturn would travel from net interest income to provisions to return on equity in sequence. And the governance structure that makes decisions fast in a crisis — one founder holding 74.3% of the vote — is also a structure with no check on that speed.

Three S-curves, out of phase

The score file's blitzscaling-fit analysis puts Nu in a category of its own: a compound S-curve blitz. Brazil is in the harvest phase of the first act — 60% penetration, growth now driven by monetisation rather than headcount, with average revenue per customer up 22% FX-neutral even as customer growth slows. Mexico is the second act just opening — the banking licence conversion completed in July 2026, with a 15.4% capital ratio, and the S-curve reproducing faster than Brazil's did. The United States is a third-act option, opened by the OCC's conditional approval. Three curves, overlapping with a time lag, in a region-by-product matrix. That is why the growth numbers look so durable: the company is not riding one explosion but sequencing them.

It is also why the score is exactly 65 rather than 73. The model does not grade the curves; it grades the whole company, deductions included. A founder with three-quarters of the vote and a credit book exposed to a record-indebted household sector are part of the company, not footnotes to it.

What would change the score

Up: Mexico's banking launch (July 2026) accelerating beyond the Brazil-at-the-same-stage benchmark would push A1 toward the full 10, which the score file explicitly reserves for that evidence. A final US OCC charter approval, plus a quantified Mexican revenue pool, would lift B3 toward the full 12. And a stabilising delinquency rate would begin to unwind the −3 on D2.

Down: a rising 90-day delinquency rate in a downturn is the direct path to a deeper D2 deduction and a score below the line — the model has the credit-cycle anchor ready for it. Further concentration of voting power, or any governance event that tests the controlled-company structure, would deepen D3. And if Brazilian penetration stalls before monetisation and Mexico can carry the growth, the G2 case narrows. The single number to watch next quarter is the delinquency rate: it is both the speedometer and the brake of this blitz.