Same ROE as the incumbent leader on half the leverage — that math is the entire bull case.

Two measurable edges do the work. A cloud-native cost base serves 14,314 customers per employee against ~1,234 at incumbents. A 100%-organic deposit franchise funds the book ~12 percentage points below where a wholesale issuer would pay — roughly $760M of annualized funding advantage on $42B of deposits at Selic 15%, before any loan is underwritten. Net result: Nu earns 2.7× Itaú's return per asset dollar on half the balance-sheet leverage.

Roughly a third of 30% ROE is rented from a 15% Selic — and Q1'26 was the first whisper of stress.

• NIM compressed, NPLs ticked up. Risk-adjusted NIM stepped from 10.5% to 9.5% in Q1'26 and 15–90-day NPLs jumped 89bps to 5.0%. Management called it Q1 seasonality plus "intentional risk expansion" via the NuFormer model; the tape priced it as the leading edge of a credit cycle and the stock printed -11.5% the next day.
• The funding moat is a percentage, not a dollar. The 88%-of-CDI edge produces ~$760M annualized at Selic 15%. The BCB path points to 9–11% over 2026–2028; the same percentage on a lower base rate halves the absolute funding edge — even before any drift to 92%+ as Open Finance enables deposit portability.
• A $32.7B unsecured book has never seen a recession at this scale. 92% of the credit portfolio is credit card + unsecured personal loans. A 100bps move in NPL 90+ on $32B equals roughly $320M of new ECL — about 11% of FY25 net income. Brazil's 2022–23 sector card NPL 90+ peaked above 14% on a smaller base.

Consensus FY26 EPS of $0.89 silently anchors on a tax rate that cannot persist.

• The EPS denominator is wrong. Q1'26 ETR printed 8.7% on a one-time interest-on-equity benefit; the Brazilian fintech statutory rate is stepping from ~40% to 45% in 2026. At a structural 22–27% ETR the same pre-tax base produces $0.70–0.75 of EPS — and the $19.50 sell-side target sits on a 26× forward multiple, not the 17× consensus advertises.
• The marginal seller is forced flow, not a bear. Short interest sits at 2.9% of float at 3.4 days to cover after a 35% drawdown. That fingerprint is LatAm/EM redemption rotation, not a directional credit-cycle short. A clean Q2 print would have to clear without help from short-cover demand.
• Mexico and the US charter are priced at zero. Mexico hit break-even Q1'26 at 15M customers; the OCC granted Nubank N.A. a conditional national bank charter on Jan 27. The stock fully unwound that rally inside eight weeks — leaving the second engine as unpaid optionality in the multiple.

Q2'26 in mid-August arbitrates the credit-cycle question without yet deciding the multiple.

• The trip-wire. NPL 15–90 falling back below 4.7% with risk-adjusted NIM holding at 9.5% invalidates the bear's credit-cycle read and re-validates through-cycle ROE durability. NPL 15–90 holding at 5.0% with NIM compressing toward 9.0% forces FY26 EPS cuts and a multiple reset.
• The October election adds tail risk. Brazil's general election (Oct 4) plus Haddad's proposed BCB mandate expansion plus the FGTS regulatory cliff make consumer-credit caps a top-tier 20-F risk for the first time. The tape now reads every Brasília policy headline as Nu-relevant.
• The disclosure machinery is thinning. The Q4'25 "Managerial P&L" framework runs on limited assurance only — not a full audit — and FY24/FY23 have not yet been re-presented. $243M of FY25 development spend was capitalized rather than expensed. None of this is fraud; it warrants a 100–200bps multiple haircut on a 7× tangible-book name.