Nu Holdings
This is the archived full research. Read the story
A bank that lives on a phone: 138.9 million people in Brazil, Mexico and Colombia use it for a no-fee credit card, loans and a deposit account, with no branches.
The scores
Growth4/4 →
Revenue grew 39% currency-neutral over years, organically, with customers, products and countries all broadening. Nothing was bought and nothing subsidised.
Business quality3/4 →
The cost position is measured against a rival's own accounts, but Brazilian account switching is cheap by design and Mercado Pago leads Mexico.
Financial position4/4 ↑
Deposits exceed the loans, all growth is funded from retained profit and it repurchased shares; the deferred-tax quarter of equity is the only caveat.
Execution record3/4 →
It beat its own guidance on what it earns after losses, and broke even in Mexico faster than Brazil. Barely five years listed.
Per share capture3/4 ↑
Diluted share count is flat and the first buyback covers stock compensation twice over, but one half-year is not yet a policy and grants continue.
Potential3/4 ↑
A large prize on two observable steps - revenue for each Brazilian customer, and Mexico now licensed - rather than anything a bank can contract.
Risk profile3/4 →
Several disclosed concentrations - one country at 90.9% of revenue, one founder at 74.3% of votes, one capped product - but no binary event.
Valuation3/4 ↑
About 4.6 times tangible book value, 93% of an excess-return value near $13.18 at an 8% return: fair, inside the 20% margin.
Evidence quality3/4 →
Everything central is filed and two rivals' own reports corroborate the moat and the threat, but the credit edge rests on an unverifiable management claim.
If it works 0.8× · if it fails you keep about 63%.
The Verdict
We rate Nu Neutral. It is a bank that lives on a phone, serving Brazil, Mexico and Colombia; serving a customer costs it very little, so it earns far higher returns than Brazil's old banks, and the business is strong. Its September quarter showed its lending margin holds without a government debt programme, but profit still leaned on a low tax rate and serious late payments rose again. The price sits a little under our value and assumes record returns become normal, leaving little cushion. As a bet on Mexico and newer countries it is priced for success: about three-fifths kept if they never earn.
The long caseMexicans are still moving from cash to digital accounts; if Nu repeats its Brazilian path there, Mexico could become a second Brazil.
How we weighed it
The price against our value
Cuts both ways
Price at 93% of our value
The price is a little under our value, with the bear case well below.
The business
In its favour
Strong
Serving a customer costs it very little, so it earns far higher returns than old banks.
Growth against what the price assumes
Cuts both ways
Growth is strong and the lending margin held without one-off help, but profit before tax grew far slower than profit after it.
What could go wrong
Against it
Brazilian defaults are rising, rates are falling, the tax rate will not last and one country is most of the business.
Against competitors
Cuts both ways
Far dearer than the Brazilian banks it names, paid for its returns.
How its group is behaving
Cuts both ways
Payments and lending names move together on rates and credit.
The latest news and report
Cuts both ways
The September report held the lending margin cleanly, but a low tax rate lifted profit, late defaults rose, and the call expects the tax rate back in the low twenties next year.
The evidence under this - 9 claims
The efficiency ratio was helped, the release says, by some expenses moving into the fourth quarter, so part of the improvement is timing.
- The size of the effect moved during the call. Cost of credit was said to be affected 'by just about 5%', the expected-credit-loss impact was 'less than $10 million', and the share of the surprise was 'about a third' - three different framings of one item, and the filing quantifies none of them.
- Management also said more than four fifths of the programme's benefit landed in Q2 2026 with a little more to come in Q3, so the comparison base for the next quarter contains a benefit that is ending.
- Running the other way: the same answer attributed the other two thirds to loan growth already on the books from Q4 2025 and Q1 2026 and to underlying credit performance, both of which are observable in the portfolio figures rather than asserted.
- 2026-11-12editedQ3 2026: the recurring margin after losses rose once the one-off items were set aside
- 2026-11-12editedre-sourced to the Q3 2026 release: late delinquency rose again
- The arithmetic assumes the multiple converges on the peer median, and there is no reason a business earning a far higher return should ever trade where its peers do. Read the other way, the calculation says how long the premium has to be earned rather than when it must disappear.
- It also assumes the return holds. The same quarter at the prior year's tax rate returns 29.6% rather than 35.1%, and a third of gross profit is interest on deposits Nu has not lent, into a policy rate already cut 125 basis points.
- Running the other way: on the sell side's own 2027 estimate of $1.13 a share the price is about 12.2 times, which is not a demanding number for a business growing profit before tax 40.6% - so the frame chosen changes the answer materially, and that is stated rather than hidden.
- The peer multiples are a vendor's, on four filers whose income statements that same vendor cannot reconcile. The book figures touch no income statement, which is why they are usable, but they are not the filings.
The company
Nu sells a credit card with no annual fee, unsecured personal loans and a deposit account, all through a phone app. Its loans total $39.4B, two thirds of them card balances, and its $45.3B of deposits more than fund them.
The alternative is a branch bank. Brazil's big banks each run up to 3,955 branches and service points and employ up to 85,206 people.
The evidence under this - 4 claims
- 2026-11-13editedexpired evidence re-sourced to the Q3 2026 release: 142.6 million customers
- 2026-11-12editedre-sourced to the Q3 2026 release
- 2026-11-12editedre-sourced to the Q3 2026 release
How it makes money
Nu earns interest on card balances and loans, fees on payments, and interest on the deposits it has not lent out. In the June quarter that last source was 34% of gross profit, against credit's 41% and fees' 25%. So a third of the profit rides on Brazil's interest rate, which has been falling since March.
The evidence under this - 2 claims
Where the money comes from, stream by stream
| Stream | Share of sales (Q2 2026) | How it earns it |
|---|---|---|
| Credit card interest | 32.8% | Interest charged on card balances that customers revolve or pay in instalments, mostly in Brazil. |
| Loan interest | 31.5% | Interest on unsecured personal loans and the newer secured and payroll-deductible loans. |
| Interest on investments and other receivables | 22.1% | What Nu earns investing customer deposits it has not lent out, the float, in securities and interbank assets. |
| Card interchange and prepaid income | 9.7% | A slice of every card purchase, paid by the merchant's side of the card network. |
| Late fees, insurance and other fees | 3.9% | Charges for paying late, insurance commissions and other fees. |
Who runs it
David Vélez founded Nu in 2013 and still runs it; Guilherme Lago has run finance since 2021, and co-founder Cristina Junqueira leads the move into America. Insiders own about a fifth of the shares but, through super-voting shares, most of the votes, nearly all Vélez's. Pay is mostly shares, but Nu does not say what targets earn them or what each executive is paid.
The evidence under this - 3 claims
Cayman law lets Nu withhold individual pay, so what the chief executive is paid on cannot be read at all; the board has delegated its authority to make share awards to the chief executive, and setting officers' pay needs his written approval under the shareholder's agreement.
Why it wins, and who could beat it
Why customers come back
Nu's advantage is cost. It runs on about 10,400 staff where Itaú Unibanco, the incumbent it measures itself against, employs 92,470. Nu spends 19.5 cents to earn each revenue dollar; Itaú spends 38.8.
That cost lets Nu serve small balances incumbents cannot price at all. And customers who bank mainly with Nu, over 60% of its Brazilian mass market, default at roughly half the average, management says.
The evidence under this - 5 claims
- The number appears in no filing. It is management's answer to an analyst who had noticed headcount falling, and the fall is explained by a return-to-office policy rather than by automation - a friendlier reading than the question implied.
- The 20-F's own year-end figure of 10,027 is four months older and lower, so the direction between the two is up, not flat. Nothing reconciles them.
- Running the other way: the ratio the claim is used for - against Itaú's 92,470 - is so large that the difference between 10,027 and 10,400 cannot change the finding.
- Nothing outside Nu measures this. Neither the delinquency split by primacy nor the primacy share itself appears in the financial statements, the 20-F or any regulator series read here - so the single most load-bearing claim in the credit case is one only the company can see.
- The company's incentive is direct: this is the claim that makes a growing unsecured book into a defensive one, and it was made in the quarter the market was selling the shares on credit worry.
- The direction of causation is not established. Customers who make a bank their main account are on average better credit risks to begin with, so the primacy may be selecting the outcome rather than producing it - management asserts the second and the data shown cannot distinguish them.
- Santander Brasil's own 20-F states the identical strategy in almost the same words - primary relationship built on transactionality, investments and credit - so the mechanism, if it works, is not proprietary.
The threat
Mercado Pago, MercadoLibre's finance arm, is the rival to watch. It claims the most monthly users of any fintech in Mexico and second place in Brazil, 78 million in all, and is building a card business of its own, though its bad-debt provisions rose with it. Inter & Co runs the same branchless model at a smaller scale.
The evidence under this - 3 claims
The rivals: MELI, MELI, INTR, BBD, BSBR, ITUB, and 1 more that is private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| MELI | Mercado Pago, the financial arm of MercadoLibrePriced against it: One segment only | Brazil | On sale | One segment only |
| Mercado Pago in Mexico, the financial arm of MercadoLibre | Nu outside Brazil: Mexico, Colombia and the United States | On sale | ||
| INTR | Inter & Co (Banco Inter)Priced against it: Yes | Brazil | On sale | Yes |
| Private | Banco C6, part-owned by JPMorgan ChasePriced against it: Private | Brazil | On sale | Private |
| BBD | Banco BradescoPriced against it: Yes | - | On sale | Yes |
| BSBR | Banco Santander (Brasil)Priced against it: Yes | - | On sale | Yes |
| ITUB | Itaú Unibanco HoldingPriced against it: Yes | - | On sale | Yes |
How big it could get
How big this could get
Card balances across Nu's three countries come to about $159B, inside roughly $1,005B of consumer loans. Nu holds $26.0B of them. Its 138.9 million customers compare with about 311 million adults in the three countries.
The evidence under this - 4 claims
- 2026-11-12editedre-sourced to the Q3 2026 release
- 2026-11-12editedre-sourced to the Q3 2026 release
The results
What growth costs
Growth has made Nu leaner: running costs fell from about a fifth of each sales dollar to about a seventh in two years. Bad loans and funding still take about half of each sales dollar, so they do not get cheaper with size, and this year running costs crept back up.
The evidence under this - 2 claims
Nu's own efficiency ratio improved to 19.5% in Q2 2026 from 21.3% a year earlier (C07), on a managerial definition with a different denominator, so the company's measure and the filed cost lines point opposite ways over the same year.
The last report (12 Nov)
Quarter ended June 2026, filed figures.
In the quarter to September 2026 Nu earned a record $1.09B, up 39.5%. Profit before tax rose only 21.0%, to $1.35B, because the tax rate was 19.2% against 29.9% a year earlier.
Its lending margin after losses was 12.1%, and 12.7% without one-off items, so it held without the government debt programme. Serious late payments rose to 7.3% from 6.9%, and to 12.1% from 11.6% on the loans that earn interest.
On the call management said it expects the tax rate to settle in the low-to-mid 20s in 2027, and the lending margin to stay strong into December.
What it did to the thesisOne answer for the case and two against. The lending margin held without government help, but profit leaned on a low tax rate management expects to rise next year, and late payments kept rising, so the business looks as strong as we thought and no stronger.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.22 against $0.23 expected.
The evidence under this - 5 claims
The efficiency ratio was helped, the release says, by some expenses moving into the fourth quarter, so part of the improvement is timing.
- The size of the effect moved during the call. Cost of credit was said to be affected 'by just about 5%', the expected-credit-loss impact was 'less than $10 million', and the share of the surprise was 'about a third' - three different framings of one item, and the filing quantifies none of them.
- Management also said more than four fifths of the programme's benefit landed in Q2 2026 with a little more to come in Q3, so the comparison base for the next quarter contains a benefit that is ending.
- Running the other way: the same answer attributed the other two thirds to loan growth already on the books from Q4 2025 and Q1 2026 and to underlying credit performance, both of which are observable in the portfolio figures rather than asserted.
- 2026-11-12editedQ3 2026: the recurring margin after losses rose once the one-off items were set aside
- 2026-11-12editedre-sourced to the Q3 2026 release: late delinquency rose again
- On the Q3 2026 call the CFO said 90-day-plus delinquency measured on interest-earning balances alone, the basis the mix explanation rests on, rose to 12.1% from 11.6%; the early ratio on that basis improved, but the late one rose even after the mix is set aside.
- An analyst pressed this twice on the call and was not satisfied, saying 'a little bit' and asking to follow up later - the exchange is on the record and the question was about exactly this distinction.
- The company's own bridge concedes that intentional risk expansion added 24 basis points to early delinquency and $170M to the allowance in the quarter, so the higher-risk cohorts are real and their cost is already visible.
- Brazil's system-wide household default rate rose from 3.75% to 4.88% over thirteen months on central-bank data, so a mix explanation has to survive a background in which everybody's book is deteriorating.
- Running the other way: allowances were built at 113% of new 15-plus delinquency formation and coverage of 90-plus balances stood at 244%, both in line with Nu's own history, so the balance sheet is not being thinned to hold the reported number up.
- 2026-11-12editedQ3 2026: the late measure rose again while the early one improved
- 2026-11-13editedthe Q3 2026 call gave late delinquency on interest-earning balances, 12.1% from 11.6%, which the mix explanation does not cover
- The tax rate has moved from 27.6% to 14.2% to 19.2% in three quarters against management's own earlier silence on it, so a range for a whole year is a forecast without a filed reconciliation behind it.
- The CFO added that Q1 carries a higher seasonal cost of risk and that faster loan growth brings more provisions up front, so the Q4 margin outlook does not extend into 2027; no guidance for next year was given.
What management did with the money
It began buying back shares: $500.4M in the first half, against $216.8M of stock pay, leaving the share count up 0.1% on a year earlier. It also opened as a bank in Mexico in August, and entered America in September.
Insiders in 2026Q2: 2 purchases and 7 sales, from their Form 4 filings.
The evidence under this - 5 claims
- 2026-11-13editedexpired evidence re-sourced to the September 2026 interim statements: no shares bought in Q3
The price, and where we disagree
In short. Fair. The price is about 93% of what we think the business is worth, at an 8% yearly return.
The shares fell 14.0% in the year to 18 September while revenue per share rose 53.1%, so buyers now pay 43.8% less for each dollar of sales. The worry is Brazilian consumer credit: household defaults are at the highest in the series, and analysts have cut ratings.
The September quarter showed the lending margin holds without the government programme. Next is an investor day on 8 December, where management may set targets, and then the December quarter: whether profit before tax keeps growing near 20% once the tax rate stops falling, and whether late payments stop rising.
The evidence under this - 5 claims
The efficiency ratio was helped, the release says, by some expenses moving into the fourth quarter, so part of the improvement is timing.
How we value it
- Method
- The owners' book plus the returns it earns above the rate
- Required return
- 8%
- Long-run growth
- 4%
- Moat
- An average moat
| Bear | Base | Bull | |
|---|---|---|---|
| Book growth, years 1 to 3 | 15%, 12.2%, 2% | 30%, 24.3%, 4% | 37.5%, 30.4%, 5% |
| Normal return on equity | 24.1% | 30.2% | 33.2% |
| Return on new capital after year ten | 8% | 12% | 15% |
| Value a share | $7.50 | $13.18 | $17.88 |
- Price
- $12.32
- Priced on
- 11 Nov
- Price as a share of the base value
- 93%
- Yearly return the price offers on the base case
- 8.4%
Where each figure comes from
| Figure | Source | Where exactly |
|---|---|---|
| The owners' book | Filed report | Interim report (6-K) for the period to 30 Sep 2026, filed 12 Nov 2026. |
| Shares | Filed report | Interim report (6-K) for the period to 30 Sep 2026, filed 12 Nov 2026; the diluted share count. |
| Book growth, years 1 to 3 | Our judgement | Our reading of the evidence; each case below names it. |
| Normal return on equity | Analysts' forecasts | FMP's consensus for the year to 31 Dec 2028, the average of 5 analysts' estimates: the return on equity of their last forecast year. FMP publishes how many analysts, not who they are. |
| Return on new capital after year ten | Our method's rule | Set by our reading of the moat: the 8% rate with none, 12% with an average one, 15% with a strong one; the bear case earns the rate. |
| The bear and bull cases | Our method's rule | Our method's spread: the bear case takes half the growth and a fifth off the margin, the bull a quarter more growth and a tenth more margin. |
| Price | Market price | FMP's closing price on 11 Nov 2026. |
What each case rests on
Bear caseEquity grows at half that path, 15%, 12% and 2%, the return on equity settles a fifth lower at 24.1%, and new equity earns only the 8% rate: the downturn test.
Base caseEquity grows 30% and 24%, the analysts' revenue path (C71) held to 30% a year, then 4% in the third year, where their 2029 average rests on two forecasts and falls below 2028, then fades toward 4%; the return on equity moves to 30.2%, the analysts' last forecast year on the filed basis.
Bull caseEquity grows a quarter faster, 38%, 30% and 5%, the return on equity reaches 33.2%, and new equity earns 15%.
What moves the base value
| Lower | Higher | |
|---|---|---|
| Growth in years 1 to 3, a quarter lower or higher | $12.04 | $14.42 |
| The normal return on equity, 15% lower or higher | $12.00 | $14.36 |
| Return on new capital after year ten, at the rate or one step up | $10.17 | $15.44 |
| The rate, a point higher or lower | $11.16 | $16.39 |
| Long-run growth a point lower | $12.39 | - |
Value a share as growth and margin change
| Normal return on equity ↓ · growth → | 50% | 75% | 100% | 125% | 150% |
|---|---|---|---|---|---|
| 24.1% | $9.76 | $10.64 | $11.60 | $12.64 | $13.77 |
| 27.1% | $10.37 | $11.34 | $12.39 | $13.53 | $14.76 |
| 30.2% | $10.99 | $12.04 | $13.18 | $14.42 | $15.75 |
| 33.2% | $11.60 | $12.73 | $13.97 | $15.30 | $16.75 |
| 36.2% | $12.21 | $13.43 | $14.76 | $16.19 | $17.74 |
Growth is shown as a share of the base case's. Green: the price is at or under 80% of that value. Red: at or over 120%.
How we read the multiple
Buyers pay about $5.50 for each $1 of Nu's net worth, what it owns minus what it owes. The four Brazilian banks it names as rivals sell for $1.10 to $2.50. To catch up with them, Nu must:
- Grow that net worth about 3.5 times.
- That is roughly four years of earning 35% a year on it.
- Two supports of that return, a tax rate that fell sharply this year and record lending profits, are not yet shown to last.
Against its own history. Cheaper than its own past on sales and on what it owns: the market has cooled on the same business. That past is short and mostly loss-making, so anchors little.
The evidence under this - 4 claims
- The arithmetic assumes the multiple converges on the peer median, and there is no reason a business earning a far higher return should ever trade where its peers do. Read the other way, the calculation says how long the premium has to be earned rather than when it must disappear.
- It also assumes the return holds. The same quarter at the prior year's tax rate returns 29.6% rather than 35.1%, and a third of gross profit is interest on deposits Nu has not lent, into a policy rate already cut 125 basis points.
- Running the other way: on the sell side's own 2027 estimate of $1.13 a share the price is about 12.2 times, which is not a demanding number for a business growing profit before tax 40.6% - so the frame chosen changes the answer materially, and that is stated rather than hidden.
- The peer multiples are a vendor's, on four filers whose income statements that same vendor cannot reconcile. The book figures touch no income statement, which is why they are usable, but they are not the filings.
- The size of the effect moved during the call. Cost of credit was said to be affected 'by just about 5%', the expected-credit-loss impact was 'less than $10 million', and the share of the surprise was 'about a third' - three different framings of one item, and the filing quantifies none of them.
- Management also said more than four fifths of the programme's benefit landed in Q2 2026 with a little more to come in Q3, so the comparison base for the next quarter contains a benefit that is ending.
- Running the other way: the same answer attributed the other two thirds to loan growth already on the books from Q4 2025 and Q1 2026 and to underlying credit performance, both of which are observable in the portfolio figures rather than asserted.
- 2026-11-12editedQ3 2026: the recurring margin after losses rose once the one-off items were set aside
Priced against its rivals
The competitors its own annual report names, each confirmed against the rival's filing. Measured on price for each $1 of net worth excluding goodwill, cheapest first:
| Ticker | Company | price for each $1 of net worth excluding goodwill |
|---|---|---|
| BBD | Banco Bradesco | 1.1× |
| INTR | Inter & Co | 1.4× |
| BSBR | Banco Santander (Brasil) | 1.7× |
| ITUB | Itaú Unibanco Holding | 2.5× |
| NU | Nu Holdings | 5.5× |
The middle of the other 4 is 1.5. Nu Holdings Ltd. is dearest of 5.
Rivals' figures come from a data provider, so read the spread of the group rather than any single pair.
Far dearer than every Brazilian bank it names, on net worth. The gap is Nu's own, paid for its returns, not a price the whole group carries.
Also checked and set aside: XP (a substitute, not the same product); PagSeguro Digital (a substitute, not the same product); StoneCo (a substitute, not the same product).
Private or foreign rivals that cannot be priced: Caixa Econômica Federal, Banco do Brasil S.A., Banco BTG Pactual S.A., Banco C6 S.A., PicPay Instituição de Pagamento S.A.
The bet: what it pays if it works
The bet is on Nu outside Brazil: Mexico, where it now holds a full banking licence, Colombia, and the United States, which it has only just entered. Mexico's sales grow fast but no profit is disclosed, so Brazil is counted separately, at the earnings multiple of Brazil's best old bank.
If Mexico keeps growing fast, the shares are still worth less than today, because the price already pays more for the countries outside Brazil. If they never earn, shareholders keep about three-fifths.
| Nu outside Brazil: Mexico, Colombia and the United States revenue in two to three years | Active customers at Mexico's revenue each needed | At the stated multiple |
|---|---|---|
| $2.50B | about 16.9M at 100% kept | 0.8× |
| $4.00B (the big win) | about 27.1M at 100% kept | 0.8× |
| $6.50B | about 44.0M at 100% kept | 1.0× |
To pay 2× (a Fair bet) it would need about $26.79B of Nu outside Brazil: Mexico, Colombia and the United States revenue (about 181.5M active customers at Mexico's revenue each); to pay 5× (a Cheap bet), about $85.40B of Nu outside Brazil: Mexico, Colombia and the United States revenue (about 578.6M active customers at Mexico's revenue each), at the same multiple.
Each row is Nu outside Brazil: Mexico, Colombia and the United States alone, plus the rest of the company (Brazil) counted at about $41.76B.
The steps it has to take
- Done Mexico reaches break-even
- Done Mexican banking licence granted and the bank launched
- Done United States opened through a partner bank
- In progress Mexico's deposits funding lending at scale due 2027-12
- Not started Profit by country disclosed in the annual report due 2027-04
- Not started United States credit models built and lending begun due 2028-03
The detail behind the label
The evidence under this - 6 claims
Mexico's deposits fell modestly again in the quarter, to $5.7B, which management calls a deliberate cut in funding cost: the revenue is growing faster than the funding base it will need once lending scales.
- The filed interim statements give revenue by country but no profit by country, so break-even cannot be checked; the 20-F is the next filing that could show it.
- MercadoLibre says its fintech arm leads Mexico on monthly active users (C12), so Mexico is not an open field the way Brazil was in 2016.
- The counterfactual assumes last year's rate was the right one, and it may not be. Two of the three reconciling items behind the fall - profits taxed at lower rates outside Brazil, and the Brazilian interest-on-capital deduction - are structural rather than one-off, and both should grow as Mexico and Colombia grow.
- The largest single item is not structural in the same way: 'other amounts', which the filing defines as foreign tax credits, non-taxable sovereign-bond interest, tax incentives and non-taxable interest on tax recoverable, contributed $183.5M against $65.0M a year earlier. Nothing in the filing says which part of that recurs.
- Running the other way: the statutory Brazilian combined rate Nu reconciles from rose from 40.0% to 42.5%, so the headline rate moved against the company while the effective rate halved. The gap is larger than a simple rate comparison shows.
- 2026-11-13editedre-read against the Q3 2026 call, which guides the 2027 tax rate to the low-to-mid 20s (C74), so last year's higher rate stays the fairer base
Where we disagree with the market
Brazil's consumer credit cycle will catch Nu. Analysts have cut ratings, and buyers pay less for each dollar of sales.
The credit worry is real but second. Two of the three legs of the record quarter were tax and interest rates, not credit.
The tax detail sits in the interim financial statements, a separate filing; neither the results release nor the call mentions it. And this market sold Nu on the same credit worry in September 2025 and bought it back: the objection recurs rather than resolves.
We are wrong if the lending gap holds near 12.4% while the tax rate returns to about 27%.
What analysts say, and the evidence
The evidence under this - 6 claims
What could go wrong, and how we would know
The risks
- Brazilian households falling behind. Household defaults rose from 3.75% to 4.88% in a year, and Nu's own 90-day defaults from 6.5% to 6.9%.
- Falling rates. A third of gross profit is interest on unlent deposits, and Brazil has cut its policy rate by 1.25 points since March.
- A tax rate unlikely to last. The quarter's 14.2% came from reconciling items, not a lower legal rate.
- One country. Brazil was 90.9% of first-half revenue, and its law caps what a card lender may charge at the original debt.
- One vote that counts. The founder holds 74.3% of the voting power.
The evidence under this - 8 claims
- 2026-11-12editedre-sourced to the Q3 2026 release: late delinquency rose again
What we still do not know
- Where the tax rate settles once this year's reconciling items pass.
- What the lending gap looks like after the government renegotiation programme ends.
- What Nu charges on revolving card debt. No filer publishes it, so the customer's side of the bargain is unmeasured.
The evidence under this - 5 claims
- The size of the effect moved during the call. Cost of credit was said to be affected 'by just about 5%', the expected-credit-loss impact was 'less than $10 million', and the share of the surprise was 'about a third' - three different framings of one item, and the filing quantifies none of them.
- Management also said more than four fifths of the programme's benefit landed in Q2 2026 with a little more to come in Q3, so the comparison base for the next quarter contains a benefit that is ending.
- Running the other way: the same answer attributed the other two thirds to loan growth already on the books from Q4 2025 and Q1 2026 and to underlying credit performance, both of which are observable in the portfolio figures rather than asserted.
- 2026-11-12editedQ3 2026: the recurring margin after losses rose once the one-off items were set aside
What would change our view
- Neutral→AttractiveThe price falls 20% under fair value
It needs:
- ○The share price falls to about $10.54, 80% of what we think the business is worth.about $10.45 today, 24% below the last close
- Neutral→UnattractiveThe price runs 20% above fair value
It needs:
- ○The share price rises to about $15.82, 20% above what we think the business is worth.about $15.80 today, 15% above the last close
- Neutral→AttractiveThe return survives a normal tax rate
It needs all of:
- ○The return on shareholders' money stays at 30% or more with the tax rate back above 25%.checked at the next results, due by 15 Mar
- ○What Nu earns on loans over what it pays for money, after losses, stays above 10% once the government programme ends.checked at the next results, due by 15 Mar
- Neutral→AttractiveMexico becomes a second Brazil, not a second market
It needs all of:
- ○Mexico's revenue per customer keeps growing faster than Brazil's did at the same stage.checked at the next results, due by 15 Mar
- ○The new Mexican banking licence turns into lending funded by deposits, at scale.watched as it happens
We would also be wrong if
- Customers who bank mainly with Nu stop defaulting less than the average.
- The cost of serving each customer rises faster than the revenue each brings in.
- The tax rate returns above 25% and the return on shareholders' money falls below 25%.
What is coming
- 8 December 2026 First Investor Day, New York: Whether management sets medium-term targets on return, efficiency and Mexico, and says how capital will be returned or deployed after the first buyback.
Every dated event ahead
- 8 December 2026 First Investor Day, New York: Whether management sets medium-term targets on return, efficiency and Mexico, and says how capital will be returned or deployed after the first buyback.
- 15 March 2027 Q4 2026 results: Whether profit before tax keeps pace once the tax rate stops falling, whether the recurring margin after losses holds near 12.7%, and whether late defaults stop rising.
- April 2027 FY2026 annual report on Form 20-F: The full-year effective tax rate, a refreshed ownership table and pay total, and whether the buyback continued across the year.
Since we wrote this
- 13 NovEarningsCall read: tax rate to rise next year, view holdsManagement expects its tax rate in the low-to-mid 20s in 2027, so this year's low rate will not last, and late defaults rose even on the loans that earn interest. Nothing moved our value; the view stays Neutral.
- 12 NovEarningsSeptember quarter: lending margin held, tax rate still lowThe margin after losses held at 12.1%, 12.7% before one-offs, without the government programme, but profit before tax rose 21% against 39.5% after tax, and late defaults rose. The view stays Neutral.
- 9 NovThesis reviewStill Neutral: priced at 98% of a value near $12.26Re-valued on its equity and the returns above 8% on it, the method drawn on every lender and insurer: the price sits inside the 20% margin of safety.
- 11 OctThesis reviewNu outside Brazil declared an early business, with a betThe owner's rule of 2026-10-11: every thesis names its lines, and an early line owes a bet. Mexico's profit is undisclosed, Colombia unreported and the United States just opened.
- 5 OctThesis reviewThe long case now sits under the verdictAfter the five-name pilot, the owner asked for the long case on every thesis.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the 20-F and filings.
- 1 OctNewsNu denies pursuing Monzo and says its capital rules are unchanged. View heldNu said it is not pursuing a transaction with Monzo, after a press report of talks at about 10 billion pounds.
- 18 SepThesis writtenNeutral, medium convictionWritten a month after the first $1B quarter, as the shares gave back the gain on worries about Brazilian credit.
Also checked 3 times with nothing new to act on.
Sources 27
Sources - 27
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-04-08The product description and the competitor list; the revolving-credit cap under Law 14,690/2023 and CMN Resolution 5,112/2023; the incumbent branch and staff ranges; the central-bank loan-balance figures for the three markets; the dual-class voting structure; the 85%-lower cost-to-serve estimate.backsC01C09C10C14C16C43C58C60C61C62C63C64C65C66
- S02primarycompany filing2026-08-13Customers, activity rate, revenue and cost to serve for each active customer, the credit portfolio split, deposits and cost of deposits, the managerial profit and loss statement and its reconciliation to IFRS, the efficiency ratio, return on equity, risk-adjusted margin, both delinquency ratios, allowance and coverage.backsC02C07C11C18C20C21C26C30C31C33C34C55C56C65
- S03primarycompany filing2026-08-13The filed income statement and balance sheet; the income tax note and its reconciliation, which is where the effective-rate finding lives; the deferred tax note; the earnings-per-share and share-count note; the equity note with the first treasury shares; the geographic revenue split; the deposit guarantee fund advance; the Mexican banking authorisation in subsequent events. Same origin as the release - same management, same day - but a different document, and almost nothing used from it appears in the release.backsC20C21C22C24C25C34C38C40C41C42C48C49C52C59C60C69
- S04primarytranscript2026-08-13Nu Holdings Q2 2026 earnings call transcriptManagement's prepared remarks, including the NuFormer description, the support-automation figure and the Mexican penetration comparison; and sixteen turns from nine analysts, of which the credit-mix exchange and the Desenrola sizing are the two that carry evidence the filings do not. The questions are a separate origin from the answers.
So a lot of -- everything -- every single underwriting decision already assumes a deterioration by default and has a pretty significant cushion in terms of what do we need to see for that decision to continue to be NPV positive.
backsC08C11C18C30C32C47C55C56C70 - S05primarycompany filing2026-09-10The United States launch date, the partner-bank structure with Lead Bank, the product terms, and the status of the national bank charter application.backsC46
- S06primarycompany filing2026-04-30The incumbent's own staff count, return on average equity and efficiency ratio - the entire external evidence base for the cost-position moat, from the party with the opposite incentive. Also the retail product list and the retail share of its credit portfolio, which fixed the cohort perimeter.backsC05C06
- S07primarycompany filing2026-02-25The substitute's own claim of fintech user leadership in Mexico and second place in Brazil, its fintech monthly-active-user series, and its statement that expanding the credit card portfolio cost group operating margin.backsC12C13
- S08primarygovernment data2026-09-18Banco Central do Brasil, SGS series 21082 - household non-performing loans, non-earmarked credit Banco Central do BrasilThe monthly system-wide household default series from June 2025 to July 2026, which is the external measure of the objection the market is pricing.backsC28
- S09primarygovernment data2026-09-18Banco Central do Brasil, SGS series 432 - Copom Selic target rate Banco Central do BrasilThe full path of policy-rate changes from 15.00% in June 2025 to 13.75% on 17 September 2026.backsC27
- S10primarygovernment data2026-09-18World Bank World Development Indicators - population ages 15-64 and 65 and above, Brazil, Mexico, Colombia The World BankThe adult population denominators for the prize base and for Nu's own penetration ratio - Brazil 170.3 million, Mexico 98.8 million, Colombia 42.2 million in 2024.
"Population ages 15-64, total" - "Total population between the ages 15 to 64. Population is based on the de facto definition of population, which counts all residents regardless of legal status or citizenship." Brazil 2024: 146884644; Mexico 2024: 87993476; Colombia 2024: 36993721. Ages 65 and above, 2024: Brazil 23424113; Mexico 10794162; Colombia 5171459.
backsC15 - S11primarygovernment data2026-09-18World Bank Global Findex - account ownership at a financial institution or with a mobile-money-service provider The World BankThe dated series behind the driver rate: Mexico 27.4% in 2011 to 53.0% in 2024, Brazil 55.9% to 86.4%, Colombia 30.4% to 57.1%.
"Account ownership at a financial institution or with a mobile-money-service provider (% of population ages 15+)" - Mexico 2011: 27.4, 2024: 53.0320833316729. Brazil 2011: 55.9, 2024: 86.4. Colombia 2011: 30.4, 2024: 57.1.
backsC19 - S12primaryvendor data2026-09-18Research pack for NU, frozen at content/NU/research-pack.jsonPrice and the 252-day return, the price decomposition into business and rating, the analyst rating series and the sell-side estimates, FINRA short interest at the 31 August settlement, the insider base rate, and the filed-competitor cohort's book and tangible-book multiples.backsC34C35C36C38C53
- S13primarycompany filing2026-09-18Nu Holdings Form 4 filings, year to 18 September 2026, read by scripts/insider_fetch.pyEvery open-market transaction by code with the Rule 10b5-1 checkbox: seven sales, no purchases, 86.8% of shares sold outside a plan, and the April option exercises that put the August sale in context.backsC44C45
- S14primarygovernment data2026-09-18Banco Central do Brasil, SGS series 20741 - average interest rate, credit card revolving credit, individuals Banco Central do BrasilThe system-wide price of the product Nu mostly sells: 137.30% a year in July 2026, in a 137% to 147% band through 2026.backsC54
- S15primarycompany filing2026-04-30Confirmation that the second listed Brazilian digital bank sells the same product to the same buyer, down to the same transactor-versus-revolver reporting construction.backsC57
- S16primarycompany filing2026-02-27A 69-million-customer incumbent stating the primary-relationship strategy in almost the same words Nu uses - evidence that the mechanism, if it works, is not proprietary.read, but no claim rests on it
- S17primarycompany filing2026-03-25An independent description of the same retail product bundle, and a second filer's account of the Desenrola law and the revolving-credit interest limits.read, but no claim rests on it
- S18primarycompany filing2026-09-02The date and place of Nu's first investor day, and that the founder and senior management will present long-term strategy and drivers of value creation.backsC68
- S19primarycompany filing2026-09-30That Nu is not pursuing a transaction with Monzo, that its capital allocation framework is unchanged, and the criteria that framework names.backsC67
- S20primarycompany filing2026-09-30A new officer, Chief Legal Officer, from 28 September 2026.backsC62
- S21primarycompany filing2026-11-09shareholders' equity, goodwill and intangibles, preferred stock, diluted earnings per share and the share countread, but no claim rests on it
- S22secondaryvendor data2026-11-09the analysts' average revenue and earnings per share for the next fiscal years, and the closebacksC71
- S23primarycompany filing2026-11-12Q3 2026 customers, revenue, profit before tax, tax, net income, margins after losses, delinquency, deposits, efficiency and the efficiency outlookbacksC01C03C04C17C29C30C32C33C51C73
- S24primarycompany filing2026-11-12equity, goodwill, intangible assets, total assets, revenue and the diluted share count at 30 September 2026backsC37C40C51C72C73
- S25primarycompany filing2026-11-12the recurring margin after losses, 12.7% against 11.9% in Q2, and delinquency by income bandbacksC32
- S26secondaryvendor data2026-11-12the analysts' average revenue and earnings per share for fiscal 2026 to 2029, and the $12.32 close of 11 NovemberbacksC72
- S27primarytranscript2026-11-12The 2027 effective tax rate expected in the low-to-mid 20% range, the one-time charge on payment-institution receivables (63 basis points of margin, about $53M of net income), the outlook for the margin after losses into Q4, 90-day-plus delinquency on interest-earning balances, and ten turns from eight analysts on credit, tax, payroll lending and the US.
This result also reflects a sequential increase of our effective tax rate to 19.2%. After a volatile 2026, we expect our 2027 ETR to settle in the low to mid-20% range.
backsC30C33C74