Nu Holdings clears $5B revenue and $871M profit — and its credit book is seasoning faster than the bulls admit

---

There is a particular kind of earnings report that is impossible to argue with on the headline and impossible to ignore underneath it. Nu Holdings delivered exactly that on May 14, 2026. For the first quarter of 2026 the company reported revenue above $5 billion for the first time in any opening quarter, net income of roughly $871 million — a record first quarter and up about 41% year over year on an FX-neutral basis — and a customer base that crossed 135 million people across Brazil, Mexico and Colombia. The efficiency ratio hit a record low near 17.6%. Mexico, the expensive new frontier that bears have flagged for two years, reached break-even ahead of the company's own internal plan. Every line a growth investor wants to see was there.

And yet the most important numbers in the release were the ones management spent the most time explaining. The credit-loss allowance build. The early-stage delinquencies. The risk-adjusted margin. Companies do not pre-empt questions about things that are going well. The forensic thesis here is not that Nu is a bad business — it is plainly an extraordinary one. The thesis is that the market is pricing a secular compounder while the income statement is quietly behaving like a fast-growing consumer lender entering the seasoning phase of an aggressively expanded, increasingly unsecured loan book — in a country whose policy rate is still close to 14%. Those are two very different things to own at the same multiple.

The allowance build is the tell

Start where management started: the credit-loss allowance. In Q1 2026 Nu's allowance build rose roughly 33% quarter over quarter on an FX-neutral basis, to nearly $1.8 billion in provisions for the quarter. That is not a rounding adjustment. CFO Guilherme Lago broke the sequential increase into three components, and the breakdown is more revealing than the total. Of the roughly $800 million quarter-over-quarter increase in expected-credit-loss allowance, about $267 million came from seasonality, about $423 million from portfolio growth, and the remainder from product mix shifting toward higher-loss unsecured products.

Read that allocation slowly. More than half the build — the $423 million growth component — is the direct cost of originating loans faster. That is the mechanical reality of provisioning under expected-loss accounting: when you book a new unsecured loan, you immediately reserve against its lifetime expected loss before it has earned you a single real of net interest. A lender growing its credit portfolio 40% year over year, with unsecured lending up 53%, is front-loading an enormous provisioning bill against revenue it has not yet collected. In good times this looks like conservatism. In a downturn it looks like the moment the music stopped. Either way, the provision is a present-tense cash-equivalent drag against a future-tense revenue promise — and the gap between the two is exactly the risk the multiple is not pricing.

The seasonality explanation is true and also incomplete. Brazilian consumer delinquencies do peak in the first quarter — the post-holiday, post-Carnival hangover is a real and recurring pattern. But "seasonality" is a comforting word for a phenomenon that only fully resolves if the underlying credit quality of each successive vintage holds. When a lender is simultaneously seasonal, growing fast, and migrating its mix toward riskier products, the three forces become difficult to disentangle in real time. That is precisely the environment in which a credit problem can hide inside a growth story for several quarters before the cohort data forces it into the open.

Risk-adjusted margin: the number that actually pays the bills

Nu reported a net interest margin near 21% and proudly noted it expanded as the credit portfolio outgrew its funding base. The bulls quote that figure. The number that actually determines profitability, though, is the risk-adjusted net interest margin — NIM after credit losses — and that one declined about 100 basis points sequentially, to roughly 9.5%.

That is the entire forensic case in two numbers. Gross margin up; margin-after-losses down. The spread Nu earns on lending is widening on paper and narrowing in economic substance, because the losses embedded in the portfolio are growing faster than the gross yield. A 100-basis-point sequential compression in risk-adjusted NIM is the financial-statement signature of a lender whose growth is outrunning the quality of its underwriting — or, more charitably, of one deliberately accepting lower-quality credit in exchange for higher gross rates. Management would frame it as the latter. The accounting cannot tell you which it is. Only the next several vintages can. And by the time the vintage data is unambiguous, the loans have already been made.

This is the denominator illusion in its purest form. Headline NIM uses a gross-yield numerator that flatters the lender; the honest denominator is loss-adjusted. When a company foregrounds the flattering metric and footnotes the honest one, the disclosure itself is a signal worth weighting.

The delinquency leap the seasonality frame partly obscures

The early-stage delinquency bucket — loans 15 to 90 days past due — jumped to 5.0% from 4.1% at year-end 2025. A 90-basis-point move in a single quarter in the leading indicator of credit deterioration is significant, and the 15–90 bucket matters more than the 90+ bucket precisely because it is forward-looking: today's early-stage delinquencies are tomorrow's charge-offs. Management noted the 90+ NPL ratio actually ticked down slightly, to about 6.5% from 6.6%, and attributed the early-stage rise mostly to seasonality with a secondary contribution from intentional expansion into higher-risk segments.

Hold both facts at once. The lagging indicator improved marginally; the leading indicator deteriorated sharply. In a seasoning credit book, that is exactly the order in which the numbers move before they roll. The 90+ figure reflects loans originated quarters ago, when the book was smaller and arguably better underwritten. The 15–90 figure reflects the newer, faster-grown, riskier-mix vintages. If those early-stage loans roll forward at historical rates, the 90+ line follows with a lag. The bulls are watching the rearview mirror; the windshield shows the road getting rougher.

"Intentional" risk-taking is still risk-taking

The most striking phrase in the call was the framing of higher delinquencies as the product of "intentional expansions into higher-risk segments." This is presented as strategic confidence — Nu choosing to lend to thinner-file, higher-yield customers because its data and underwriting are good enough to price that risk profitably. It may well be right. Nu's data advantage is real, and its history of expanding the credit box without blowing up is genuinely impressive.

But "intentional" is a description of the decision, not a guarantee of the outcome. Every consumer lender that has ever cratered did so by intentionally expanding into higher-risk segments it believed it could underwrite. The graveyard of fintech credit is filled with companies whose models worked beautifully until the macro turned and the marginal borrower — the one added at the edge of the credit box to keep growth alive — defaulted in clusters the model had not seen because the model had never lived through that environment. Nu's models have never operated through a full Brazilian credit downturn at this scale and this mix. The 53% growth in unsecured lending to roughly $10 billion is the company pressing the accelerator on exactly the segment that performs worst when unemployment rises and real incomes fall. Intentional, yes. De-risked, no.

Cyclical engine, secular multiple

Here is the valuation crux. At a share price around $12 in mid-May 2026, Nu traded at a forward P/E in the low-to-mid teens on 2026 estimates — roughly 14x against an EPS estimate near $0.89 — for a company growing earnings around 40–48% annually with a return on equity above 30%. On a pure growth-adjusted basis, bulls argue that is cheap, even a bargain, and several models peg fair value at twice the price.

The forensic objection is not to the growth rate. It is to the implicit assumption baked into a secular multiple: that the earnings stream is durable and the loss rate is structural rather than cyclical. A 30%-plus ROE earned by a fast-growing unsecured lender in a 14% policy-rate economy is not the same asset as a 30% ROE earned by a payments network or a software company. The first is leveraged to the credit cycle; the second is not. If you believe Nu's current loss rates are near a cyclical trough — and with the leading delinquency indicator already rising, that is a live possibility — then the through-cycle normalized ROE is lower than the trailing figure, and the "cheap" multiple is cheap on peak-ish earnings. Cyclicals always look cheapest right before the cycle turns, because the denominator is about to fall.

The market is paying for the customer-acquisition story — 135 million users, ARPAC rising toward an eventual $30–$40 from roughly $16 — and treating the lending engine that monetizes those users as a stable annuity. It is not a stable annuity. It is a credit book.

Brazil's rate environment cuts both ways — mostly the wrong way

Nu's defenders note that Brazil's central bank has begun easing, cutting the Selic to 14.25% by mid-2026 from the 15% level held since June 2025. Lower rates eventually reduce Nu's funding costs on its roughly $42 billion deposit base and can stimulate borrowing. True. But the sequencing matters, and it does not flatter the near term.

A still-elevated policy rate near 14% does two things to a consumer lender simultaneously, and both are visible in this quarter. It keeps Nu's funding expensive, compressing the spread on the liability side. And it keeps debt-service burdens punishing for the very borrowers Nu is adding fastest, raising the probability that the newer, riskier vintages default. The benefit of easing — cheaper funding — arrives with a lag and accrues gradually. The cost of the high-rate regime — stressed borrowers and rising delinquencies — is already in the Q1 numbers. The rate cycle, in other words, is currently working against Nu on the credit line faster than it is working for Nu on the funding line. A market betting on the easing tailwind is betting on a benefit that shows up after the risk it is supposed to offset.

Mexico break-even is real — and the smallest part of the bet

The Mexico break-even milestone is genuine and arrived ahead of plan, with the unit serving more than 15 million customers; Colombia is approaching 5 million. These are legitimate achievements and the strongest part of the bull case for durable, multi-country growth. But it would be a mistake to let the new-market good news anchor the whole thesis. Brazil — 115 million-plus customers and the overwhelming majority of revenue, deposits and credit — is where the earnings are made and where the credit risk concentrates. Mexico reaching break-even changes the long-term optionality; it does not change the fact that the Q1 margin compression and delinquency rise originated in the mature, scaled, Brazilian book. The new markets are the call option. The Brazilian credit cycle is the underlying. You do not get paid on the option if the underlying breaks.

Quality of earnings: managerial vs. the seasoning curve

Nu's $871 million figure is its managerial net income, the company's preferred presentation. It is a clean, defensible number and Nu's disclosure is better than most. But every reported profit in an expected-loss lending model is a function of the allowance management chooses to hold against future losses — and that allowance is, by construction, an estimate of how vintages not yet seasoned will perform. Nu reports gross coverage of new 90+ NPL formation above 150%, which is reassuring on its face. The forensic caveat is structural, not accusatory: in a book growing 40% a year with a mix shifting toward unsecured, the adequacy of today's allowance depends entirely on whether the loss curves of the newest, riskiest vintages resemble the historical curves the reserves are calibrated against. If the new cohorts season worse — as the 15–90 leap hints they might — then today's profit was overstated by an under-reserve that only future quarters will reveal. This is not an allegation of wrongdoing. It is the irreducible epistemics of lending into a fast-growing, mix-shifting book.

The product-mix migration is a one-way ratchet

There is a subtle structural point inside the "intentional risk-taking" framing that deserves its own examination, because it changes the character of the risk over time. Nu is not merely adding higher-risk customers at the margin; it is re-weighting its entire credit mix toward unsecured products — personal loans and credit cards — which carry both the highest gross yields and the highest loss rates in the portfolio. Unsecured lending grew 53% to roughly $10 billion while the overall credit book grew 40%. That differential is the mix shift made numerical: the riskiest segment is growing fastest, by design.

Mix shifts of this kind are easy to enter and hard to reverse. Once a lender's revenue and ARPAC trajectory depend on the gross yield of unsecured products, dialing back that segment to de-risk in a downturn means accepting slower growth and lower revenue per customer precisely when the market is most punishing of decelerating fintechs. The migration toward unsecured credit is, in effect, a one-way ratchet on the income statement: it lifts revenue today and embeds loss-rate sensitivity tomorrow, and unwinding it gracefully in a stressed environment is far harder than the smooth path up suggests. The same lever that powers the ARPAC story is the lever that makes the earnings cyclical. You cannot separate the two; they are the same decision viewed from opposite ends.

This matters for how to read every future quarter. As long as the mix keeps shifting toward unsecured, gross NIM will keep looking strong and risk-adjusted NIM will keep being the number that tells the truth. The widening gap between those two figures is not noise — it is the running tally of how much cyclical risk Nu is accumulating in exchange for its secular-looking growth.

What the bulls genuinely get right

Intellectual honesty requires conceding the bull case is strong — stronger, in fact, than most short theses ever face. Nu is one of the best-executed consumer-finance stories of the past decade, and the bears who dismissed it at $4 and $8 have been comprehensively wrong.

The customer engine is real and cheap: Nu acquires customers at a fraction of incumbent-bank cost and monetizes them with a record-low efficiency ratio near 17.6% that legacy banks cannot approach. The ARPAC runway is genuine — at roughly $16 monthly against incumbents near $45, there is years of monetization upside as customers adopt more products, entirely independent of adding new users. The deposit franchise, north of $42 billion and growing over 20%, is a low-cost, sticky funding base that most fintechs would kill for and that gives Nu a structural funding advantage over both neobank peers and many incumbents. The Mexico break-even proves the model travels beyond Brazil, validating the multi-country thesis that bears long doubted. And the 30%-plus ROE is real cash-on-equity, not an accounting artifact — Nu is genuinely one of the most profitable banks in the world per unit of equity. Management's disclosure on the allowance build was candid and granular in a way that many lenders would have buried. If the credit book seasons in line with history rather than worse, the stock at a mid-teens multiple on 40%-plus growth is, straightforwardly, underpriced. None of the forensic concerns here require Nu to be a bad company. They require only that the market has chosen to price the good outcome as the certain one.

The asymmetry, not the direction

The honest forensic conclusion is not "Nu will collapse." It is that the risk is asymmetric and the price does not reflect it. If the credit book seasons normally, the bull case prints and the stock works. If the newest unsecured vintages season worse than history — entirely plausible given the deliberate move down the credit spectrum into a high-rate economy — then risk-adjusted NIM keeps compressing, the allowance keeps building, reported ROE normalizes lower, and a stock priced as a secular compounder gets re-rated as the cyclical lender it partly is. You are not being paid much for taking the second outcome, because the multiple already assumes the first.

The kicker

A growth multiple says the future is certain; a rising allowance says the company isn't sure. When the leading delinquency indicator jumps 90 basis points, risk-adjusted margin gives back 100, and the loss reserve climbs a third in a single quarter, the market's job is to ask whether it is paying for a compounder or for the best-looking quarter of a credit cycle that has only just begun to turn — and right now, at a secular multiple on a book that has never seen a full downturn at this scale and this mix, it is being asked to take that question on faith.

The headline is a record and the footnote is a warning, and in consumer lending the footnote always gets the last word.