I am studying NU, and I am still unsure wether it's undervalued or not.
I understand that the market is pricing in South America risks, but this growth and those margins are very rare at those multiples.
My take is that if you value it as a fintech, it's undervalued. If you value it as a bank (on P/B, ROTE, Deposit mix, and NLP trend), it's fairly valued. But this labelling is a bit silly because banks do not have those efficiency ratios and those margins. If you compare it with SOFI (which I consider overvalued) it's cheaper. Another possible peer is DAVE, but they are on different trajectories.
Finally I think I might buy, because it looks like a very high quality business, with good focus and clear growth potential. I like the CEO, and I think that there is sufficient margin of safety.
What is your opinion?
For those unfamiliar with the stock, I will drop an AI-generated overview of the company (guilty!)
\---
# The Business in One Paragraph
Founded in 2013, Nu is the largest digital bank outside Asia. 135M customers across Brazil, Mexico, and Colombia. It started with a no-fee credit card and now does everything: checking accounts, loans, insurance, crypto, investments, a travel portal, even a mobile phone service (NuCel). It's a super-app model — Lynch would call this a **fast grower** that's still early in cross-selling its existing base. Revenue went from $1.5B in 2021 to $17.5B TTM. It turned profitable in 2022 and has been compounding EPS at 45%+.
\---
# Six-Pillar Table
|Pillar|Score|Positives|Negatives|
|:-|:-|:-|:-|
|**Moat**|8/10|Structural cost advantage — 20-30x more efficient than Brazilian incumbents (Itaú, Bradesco). No branch network, legacy IT, or官僚 bureaucracy. Data moat: 135M customers generate granular credit data that feeds AI underwriting models. Brand is strong in Brazil — top-of-mind for millennials/gen Z. AI transformation is real: engineering throughput up 50% YoY, testing cycles 90% faster, real-time NPV pricing on loans.|No hard switching costs — a customer can download another app. The moat is operational efficiency, not a license or network effect with lock-in. Incumbents are fighting back with their own digital products (Itaú's Íon).|
|**Growth**|9/10|Revenue grew 43% YoY (TTM). Three engines: (1) customer acquisition — still adding millions per quarter in Brazil, Mexico, Colombia; (2) ARPAC expansion — $16/quarter and rising as customers use more products; (3) geographic expansion — Mexico just hit IFRS profitability ahead of plan, Colombia growing fast. Credit portfolio growing 40% YoY. TAM is enormous: Brazil's banking profit pool is \~$40B and Nu has maybe 10% share.|Growth is partly credit-driven (98% of new exposure is unsecured). If Brazil enters recession, loan growth stalls and provisions spike. Mexico and Colombia are still tiny relative to Brazil. The U.S. expansion is a call option with <100bps OpEx headwind, but unproven.|
|**Management**|9/10|David Vélez (CEO) and the team are exceptional communicators. On the last call they decomposed the $800M provision increase and showed 86% was growth + seasonality, not credit deterioration. They admitted when a cost beat was timing-related (marketing, real estate) rather than structural. They capped U.S. investment risk explicitly. This is the kind of candor Buffett looks for — they prioritize long-term value over making each quarter look perfect.|Vélez is the founder with super-voting shares. That's fine while he's good, but succession risk exists.|
|**Financial Safety**|8/10|Net cash position: $16.1B cash vs $5.2B debt. Debt-to-equity of 0.25 — very low leverage for a bank. Short-duration assets (mostly credit card receivables) mean they can reprice quickly if rates move. High provision coverage — they're building reserves ahead of NPL formation, not behind it. Deposit base is stable and low-cost.|Current ratio of 0.83 — this is normal for a bank (liabilities are mostly deposits, not debt), but it means they rely on deposit stickiness. Rapid credit growth (40% YoY) means the loan book is young and hasn't been tested through a full cycle. If Brazil unemployment spikes, the unsecured portfolio gets hit.|
|**Efficiency**|9/10|Operating margin of 23%, profit margin of 18%. Efficiency ratio (costs/revenue) was 17.6% in Q1 2026 — absurdly low for a bank. Even normalizing to \~20% for full year, that's still best-in-class globally. AI is driving this lower: customer service automated, credit decisions in milliseconds, no branches.|The Q1 beat was partly timing (marketing spend shifted). Full-year efficiency will be \~20%, not 17.6%. Still excellent, but the trajectory of improvement may slow as they invest in U.S. and AI.|
|**Valuation**|5/10|Trailing P/E of 21x, forward 18x, PEG of 0.48. For a company growing earnings 45% that's cheap on a PEG basis — Lynch would say a PEG under 1.0 for a fast grower is interesting. EV/EBITDA of 12.2x is below the peer median of 18x. Price/Sales of 3.8x is high vs peers (2.1x median) but that's because Nu has much higher margins and growth.|The cheapness is the Brazil discount. If you strip out the Brazil risk premium, what multiple would this trade at? A US fintech growing 30%+ with these margins would trade at 30-40x. The 21x is the market saying "we don't trust the macro." If Brazil has a currency crisis or political blow-up, the multiple could compress further. Also, the low PEG assumes 45% growth persists — growth will decelerate as the base gets bigger.|
\---
# What the Earnings Call Told You That the Numbers Don't
**The provision spike in Q1 2026 scared the market.** Provisions jumped $800M. The stock sold off. But management broke it down:
* **86% of the increase was growth + seasonality** (more loans → more reserves required by accounting rules; Q1 is seasonally heavy for Brazilian defaults because of holiday spending).
* **Only 14% was from intentional risk expansion** — they deliberately extended credit to slightly riskier segments because their models said they could price for it.
* **90+ day NPLs were stable.** Coverage ratio was high.
This is the key debate: is management right that the credit cycle is benign, or are they taking on risk that will blow up later? The stock is pricing in the latter scenario. If they're right, the stock is very cheap.
**Mexico is the proof-of-concept.** Mexico went from a $30M quarterly loss to IFRS profitability ahead of plan. ARPAC nearly doubled. Efficiency ratio dropped 78 percentage points in 4 years. This shows the model works outside Brazil and reduces the "one-country risk" argument.
**The U.S. is a controlled experiment.** Management capped the OpEx headwind at <100bps on the efficiency ratio for 2026 and 2027. They said additional investment depends on product-market fit. This is a small, disciplined bet — not a capital incineration project.
\---
# The Bear Case You Need to Address
1. **Brazil macro is real.** The real has been volatile. Interest rates are high (Selic at 14.25%). If Brazil enters a recession, Nu's unsecured loan book takes losses. The stock will re-rate lower regardless of fundamentals because it's a Brazil proxy.
2. **Growth will decelerate.** 43% revenue growth on a $17.5B base is harder to sustain. As they penetrate the Brazilian market, marginal customers are lower quality. ARPAC expansion can only go so far before hitting income constraints.
3. **It's not really a bank — it's a consumer lender.** Nu makes most of its money on credit card interest and unsecured loans. In a downturn, those are the first things to default. Itaú has a diversified corporate and investment bank to cushion consumer losses. Nu doesn't.
4. **The cheapness is a trap if earnings fall.** 21x trailing looks cheap for 45% growth, but if earnings drop 20% in a recession, 21x becomes 26x on lower earnings. The multiple compression + earnings decline is the permanent-loss scenario.
\---
# What to Watch
* **Risk-adjusted NIM in Q2/Q3 2026.** Management promised it would revert to \~10.5% (H2 2025 levels). If it doesn't, the credit story is wrong.
* **Mexico's contribution.** If Mexico keeps scaling profitably, it reduces the single-country risk.
* **ARPAC trajectory.** If ARPAC stalls, the cross-sell thesis is capped.
* **Insider selling.** No data available right now, but if Vélez starts selling meaningful amounts, that's a red flag.
\---
# Bottom Line
Nu is a high-quality business trading at a discount because of where it's domiciled. The six-pillar framework says: **great moat, great growth, great management, great efficiency, safe balance sheet, uncertain valuation.** The valuation pillar is the only one that's not a clear positive — and that's entirely driven by macro risk, not business risk.
If you believe Brazil avoids a crisis, this is a Graham-style net-net on growth: you're paying 18x forward earnings for a business compounding at 30-40% with a net cash balance sheet and a founder-led management team that communicates with unusual candor. That's a rare combination.
If you think Brazil is headed for a lost decade (currency collapse, political instability, high rates forever), then the cheap multiple is a value trap and you'll lose money even if the business executes perfectly.
The stock doesn't look cheap because it's misunderstood. It looks cheap because the market is pricing in a non-trivial probability of something going wrong in Brazil. The question is whether you think that probability is overestimated.