NYSE: NU

Nu Holdings Ltd.

Three credit cycles, one dollar income statement

Price $13.56

Data as of 22 September 2026

Rebuilt from primary sources on each publish

Nu earns in reais, pesos and Colombian pesos and reports in dollars, so the equity is a levered bet on three consumer credit cycles at once. The franchise side keeps compounding: 139 million customers at June 2026, $1.93bn of earnings already banked in the first half, and a base case that compounds earnings to $1.98 a share by FY2029. The credit bill is the offset. Losses took 30.5% of revenue in H1 2026 against 26.7% in FY2025 and Brazilian household delinquency sits at 5.81%, though the Selic has now been cut five times to 13.75%. The whole base case is one line: cost of risk on the book falling from 20% this year to 16.5% by FY2029. Get that and the three valuation routes still land between $12.91 and $26.61 a share. The spread between them, not the delinquency line, is what makes this a judgement call rather than a calculation.

$13.56

Traded price

NYSE: NU · priced 2026-09-10

$17.89

Three-method value

+32% versus market

27%

Base IRR to 2029

exit at 15x FY2029E EPS of $1.98

30%

Return on equity

FY2026E; 30% in FY2025

30.5%

Credit losses / revenue

H1 2026, from 26.7% in FY2025

5.81%

Brazil delinquency

+1.98pp over two years

The environment

Two charts that between them explain most of what happens to this stock. Delinquency sets the loss rate; the policy rate sets both the cost of funding and the customer's ability to pay.

Consumer credit delinquency

System-wide, Nu's three markets, last 24 months

Each regulator publishes on its own bucket, so read the trend, not the gap between lines. Brazil: households 90+ days (BCB SGS 21084). Colombia: credit cards 30+ days (Superfinanciera). Mexico: awaiting data feed.

Central bank policy rate

Sets Nu's funding cost and its customers' capacity to repay

End-of-month headline policy rate. Source: Bank for International Settlements (WS_CBPOL).

Translation and the credit bill

A dollar reporter earning in three softer currencies carries a translation effect that has nothing to do with operations. Beside it, the chart that matters most: whether the macro has reached the income statement.

Currency strength versus the US dollar

Indexed to 100 two years ago. Above 100 flatters reported dollar results

Month-end spot rates, inverted and rebased so a rising line means a stronger local currency. Source: Yahoo Finance.

Credit losses as a share of revenue

Where the credit cycle actually shows up. Grey bars are modelled

Expected credit loss over total revenue. This is the IFRS 9 P&L charge, not cash written off: most of it is the provision Nu books against loans that are still performing, so it moves when the book grows or the outlook darkens, ahead of any actual default — and it reverses if those loans cure. That is why it spikes with first-quarter delinquency, and why it holds near thirty percent of revenue in the forecast even as the loss rate falls: the credit book grows faster than revenue. The driver that falls is cost of risk on the book. Source: Nu Holdings IFRS filings through Q2 2026.

The customer base

Seven quarters of reported history, nothing modelled. The base still adds about four million a quarter, and the revenue each of those customers produces has risen in every single one of them.

Customers and net additions

Millions. Bars are the base, the line is each quarter's intake

The base grew 22% over these seven quarters while the quarterly intake eased from 4.4m to 3.7m. That is the arithmetic of a maturing base, not a broken one: the same 4m a quarter is a smaller percentage each time. Source: Nu earnings releases.

Revenue per active customer

Monthly ARPAC, US$, as reported each quarter

Up 54% in seven quarters, and rising in every one of them. Nu's definition: average monthly revenue divided by average monthly active customers. The model carries it to $20.54 by FY2029 — a slower climb than the one these bars have already delivered.

Unit economics

The compounding argument in two charts: what it costs to serve a customer against what that customer pays, and how much balance sheet each one carries.

Cost to serve and the leverage on it

US$ per active customer per month, against revenue over that cost

Cost to serve rose 25% over the period, on return-to-office costs and international expansion, and Nu has said it expects to hold around a dollar. It still earns 17 times that cost per customer, against 14 times at the start. Definition includes transactional expenses as well as support and operations.

Balance sheet per active customer

US$. Solid to FY2025, dashed is modelled

The gap between the two lines is the funding cushion expressed per customer, and the model has it closing: lending per customer roughly doubles by FY2029 while deposits per customer grow far more slowly. Brazilian incumbents carry several times this much credit per client, which is the room the thesis is paying for.

The franchise

Revenue compounded at 49% a year from FY2022 to FY2025 and the first half of 2026 ran ahead of it. The question is no longer whether it grows, it is what the growth costs.

Revenue and gross profit

US$ billions. Solid bars actual, grey modelled

Source: Nu Holdings IFRS filings and the Aadi institutional model, base case. Forecast years carry the model's operating-ratio assumptions.

Margin structure

Operating leverage is real; the gross line is where credit costs bite

Each line as a percentage of total revenue. H1 2026 is six months of trading, directly comparable on ratios.

Profitability

Quarterly earnings against the equity that produced them. The first quarter of each year is seasonally the weakest, which is worth holding in mind before reading any single print as a trend.

Net margin and return on equity

Per cent. Net income over managerial revenue, against reported ROE

Both lines dip in the first quarter of each year: Brazilian delinquencies peak early and the credit provision follows them. Read Q1 against Q1, not against the quarter before it. Margin is on the managerial revenue base, which is how Nu reports ARPAC too.

Returns and capital

A thirty-percent return, on an equity base holding near a sixth of assets

Return on equity uses average equity; H1 2026 net income is annualised. Both lines are percentages, one scale.

Funding

Deposits funding a smaller credit book is the structural advantage. Through H1 2026 the book grew faster than the deposits behind it, so the gap between the two lines is closing and the cushion is being spent.

Credit book versus deposit funding

US$ billions. Deposits above the book is the cheap-funding advantage

Credit book is card receivables plus loans to customers, gross. The gap narrowed sharply through H1 2026 as lending outgrew deposits. Source: Nu Holdings balance sheet.

Deposits and deposit growth

US$ billions, with the sequential growth rate

Deposits grew 57% across these seven quarters, but the rate is easing and Q1'26 was almost flat on a seasonal outflow that Q2 recovered. Growth here is in dollars, so it carries translation; Nu quotes +18% year on year FX-neutral for Q2'26 against +24% in reported dollars. Cost of deposits held at 88% of the interbank rate.

Valuation

Three routes, a factor of two apart. Residual income lands under the traded price, the exit multiple lands at twice it, and which one you believe is the entire investment decision.

Excess-return value per share

Residual income. Shares trade at $13.56; warm cells sit below that

The model charges 13.0% for equity — CAPM plus a blended country risk premium. Read down that column: on a residual-income basis the shares are close to fairly priced, and the upside in this page is an argument about the multiple, not about the discount rate.

Value per share by method

The routes disagree by a factor of two. Reference line is the $13.56 traded price

Residual income charges for the equity that growth consumes and lands below the price. The exit-multiple route capitalises FY2030 earnings and lands at twice it. The disagreement is the finding, not a rounding error: this is a multiple story, not a discounted-value story.

What would change the view

The five things worth watching, in the order they would move the estimate.

Cost of risk falls from the H1 2026 run-rate

The base needs 20% of the average book this year and 16.5% by FY2029. Q1 2026 annualised at 23%, Q2 at roughly 18% as early delinquencies normalised. Every point it fails to fall is about $0.11 of FY2029 EPS against a base of $1.98. This is the whole argument.

Brazilian delinquency stops rising

Household 90-day arrears moved +1.98 points over two years. Nu's book is more concentrated in exactly the borrower this series describes.

The easing cycle keeps running

Copom has cut five times from 15.0%, most recently to 13.75% on 17 September 2026, a net +3.00 points over two years. Cuts help twice: cheaper deposits and healthier borrowers. The base case has the annual average at 12.0% in FY2027 and 9.3% by FY2030, so a pause costs it.

Deposits keep up with the credit book

They did not in H1 2026. Lending grew far faster than funding and the balance-sheet ratio went from 66% of deposits at FY2025 to 72% at June. The base case takes it to 91% by FY2029, which is still conservative for a bank, but the cheap-funding cushion is being spent.

Mexico and Colombia keep improving

Colombian card delinquency has fallen steadily and Banxico has been cutting. Nu took a full banking licence in Mexico in August 2026 and passed five million customers in Colombia. If the non-Brazil business scales into that, the diversification stops being a story and starts being earnings.