Moat
What, If Anything, Protects Nu Holdings
1. Moat in One Page
Conclusion: narrow moat. Nu has a measurable and company-specific economic advantage, but it is concentrated in one geography (Brazil), one cost line (unit operating cost), and one cohort behavior (rising ARPAC on a flat cost-to-serve). It does not extend cleanly to Mexico, where Mercado Pago is the only peer with comparable returns, and it has not been tested through a deep Brazilian consumer-credit downturn at current scale. A moat exists; the case for wide moat is not proven.
The 2-3 strongest pieces of evidence: (1) ~14,000 customers per employee versus ~1,200 at Brazilian incumbents, driving a ~$0.80/month cost-to-serve that no listed peer matches; (2) a 30.3% FY2025 ROE earned alongside 27% revenue growth — a combination no other listed LatAm bank achieves, with Inter and PagBank stuck at 14% ROE despite being digital-native; (3) primary-banking-relationship penetration of roughly 58–65% of 1-year-tenured customers, anchoring a deposit franchise funded at 81% of CDI. The 1-2 biggest weaknesses: Inter funds itself at 65.3% of CDI — 160 bps cheaper than Nu — proving the deposit-cost advantage is a function of scale and brand, not digital-bank physics; and the cost-of-risk line at ~19% of average loans has never been tested through a real Brazilian recession at the current $27.7B loan book.
Moat Rating: Narrow moat · Weakest Link: Mexico replication risk vs Mercado Pago
Evidence Strength (0-100)
Durability (0-100)
A moat is a durable economic advantage — durable means it survives competition, cycles, and management changes. Saying "Nu has a great business" is not the same as saying "Nu has a moat." The moat question is specifically: what stops a well-capitalized competitor from copying this in five years? For Nu, the answer is "Brazil scale," not "everything they do."
2. Sources of Advantage
Five candidate moat sources are testable against the upstream evidence. Two pass with high confidence, two pass with qualifications, one fails. The point of this table is to be specific about how each one is supposed to work and what would falsify it.
Plain-English definitions used above: Switching cost = the friction (time, risk, lost data, lost convenience) a customer faces when changing providers. Open Finance reduces this structurally; brand and habit can still raise it psychologically. Network effect = each new user makes the product more valuable to existing users. Nu has a data network effect inside its own underwriting models, but it does not have a two-sided network like a marketplace. Scale economies = fixed costs spread over more revenue, lowering per-unit cost. This is Nu's strongest moat lever.
3. Evidence the Moat Works
Six observable data points, each linked to a moat claim and to a way it could be wrong. The point is to test the moat, not to celebrate it.
The ARPAC chart is the single cleanest visualization of the moat working. Monthly revenue per active customer has roughly tripled since 2019 on a cost-to-serve that has stayed essentially flat at $0.80. The $27/month mature-cohort number is what every other cohort could converge to as it ages — that gap is the moat's compounding mechanism, not a forecast.
The peer ROE chart tells the moat story most precisely: Nu earns a 30% ROE despite a worse cost of funding than Inter — meaning the moat is not cost-of-funding per se, it is scale and cross-sell stacked on top of a still-good (not best-in-class) cost-of-funding base. If Nu had Inter's funding cost, ROE would be materially higher; if Inter had Nu's scale, Inter's ROE would presumably also be much higher. The variable that separates them is scale + cohort maturity, which is exactly the moat.
4. Where the Moat Is Weak or Unproven
The honest skeptical reading of the moat case rests on five concerns, each of which would compress earnings power or refute one of the source claims.
The single most fragile assumption in the wide-moat case is that the Brazil ARPAC curve replicates in Mexico against Mercado Pago. The bull case requires Mexico customers to age from a low-ARPAC starting point to mature-cohort economics on a Brazil-equivalent curve. Mercado Pago is not Bradesco — it is a 36% ROE LatAm fintech with a captive e-commerce flywheel and the cheapest customer-acquisition channel in the region. If the moat must defend against MELI rather than ITUB, the math is materially harder.
5. Moat vs. Competitors
The Competition tab established the peer set. The point of this table is to translate that into moat-relative strength: where each competitor is stronger or weaker than Nu as a moat — not just on financials.
Confidence on peer data: Medium-high for Nu, ITUB, BBD, MELI, INTR (all 20-F filers); medium for PAGS; low for C6 (private). Customer-per-FTE ratios for peers are derived from disclosed customer counts and approximate headcount and should be treated as order-of-magnitude, not point estimates.
The moat-relative read of the peer bubble: Nu and MELI are alone in the upper-right quadrant where the market prices a genuine moat. Inter has one moat ingredient (funding) and is being priced at 1.4x book — proof that any single ingredient is not enough. ITUB has the legacy moat (corporate + affluent + insurance) and trades at 2.2x book — the market sees those as durable but not growing. Nu is the only listed LatAm bank for which a 5x book multiple makes sense, but the comparison that should keep an analyst honest is to MELI (15x book / 36% ROE) — that is what a wide moat would look like, and Nu is not currently priced there.
6. Durability Under Stress
A moat must survive things it has not yet faced. Six stress cases, with the realistic Nu response and the signal an analyst would watch.
Stress Cases — Probability, Severity, and Moat Resilience (0-100)
The heatmap is a reading aid, not a model output: numbers are author judgment, anchored on upstream evidence. The two stresses where moat resilience scores lowest are the Mexico/Mercado Pago contest and a deep Brazilian unsecured-credit downturn — both because Nu has not been tested through them. The two where resilience scores highest are incumbents-catch-up (slow erosion, decade-scale) and a single regulatory cap (Nu has navigated the 2023 cap and is reweighting toward installment + secured).
7. Where Nu Holdings Fits
The moat is not uniform across Nu's portfolio. Three observations matter:
1. Brazil consumer-credit and deposits — moat is strongest. This is the franchise that drives 85-90% of revenue and the great majority of profit. Customers-per-FTE, primary-banking penetration, brand favorability, deposit cost relative to CDI, and BCB-tracked share gains are all evidence of a real, measurable advantage here specifically. Within Brazil, the unsecured credit-card book (24.3% share of purchase volume) is where the moat shows up most clearly because that product mix capitalizes the lowest cost-to-serve and the highest revolving-yield combination.
2. Mexico — moat is partial and contested. Nu is the #3 credit-card issuer (6.6M cards, ~17-18% of total Mexican credit-card market as of June 2025) but Mercado Pago is the #1 fintech by MAU and the only competitor that out-grows and out-monetizes Nu in this market. The moat ingredients that travel best to Mexico are brand and unit cost; the ones that travel weakly are the BCB-driven regulatory tailwind (Mexican Sofipo regime is different) and the data-underwriting flywheel (Mexican credit history is shorter and Mercado Pago's e-commerce TPV gives it more underwriting signal in some segments).
3. SME, US, affluent BR — no moat yet. SME (Working Capital launched 2024, Charging Assistant 2025) is early and competes against PAGS and Stone which have multi-year head starts on merchant relationships. US (OCC conditional Jan 2026) is pre-revenue and Nu will be a sub-scale challenger in a market dominated by Chime, SoFi, and the largest US banks. Affluent Brazil banking is Itaú Uniclass and Bradesco Prime territory and Nu's Ultravioleta tier is small. These are call-options on the future moat, not part of the current moat.
The moat conclusion is largely a Brazil conclusion. Brazil consumer banking is roughly 80% of where the moat is doing work today; Mexico is contested optionality; SME, US, and affluent BR are call options that should not be priced in until the moat is observable in segment economics. When management reframes the equity story as "AI-first" and "US national bank," the analyst's job is to keep separating proven moat from prospective moat.
8. What to Watch
Eight observable signals will tell you whether the moat is widening, holding, or narrowing — in roughly the order they tend to move.
The first moat signal to watch is Nu's cost of funding as a percentage of CDI. It moves before earnings, before NPL, before customer count; it sits at the structural intersection of brand, primary banking, and deposit franchise; and Inter's 65.3% disclosure proves the ceiling is set by competition, not by digital-bank physics. A reading that stays in the 75-85% band keeps the moat intact. A reading that breaks 88% in a quarter where Selic is flat is the metric to act on — that is the deposit franchise weakening, and the deposit franchise is the spine of every other moat ingredient.