Nu Holdings Ltd. NU

Key data

The business

Nu is a bank with no branches, run through a phone app, built around a purple card that a Brazilian first gets for credit and then uses for everything. It sells credit cards, checking-style accounts, deposits, personal loans (secured and, increasingly, unsecured), investments and a shopping marketplace, to almost 118 million customers, with more than 86% active in a month for the first time. It makes money two ways: interest on the money it lends, and small fees each time a card is swiped. The lending is funded by $45.3 billion of customer deposits that cost Nu only 88% of the local interbank rate, cheaper than any bank with tellers can match. That funding gap is the moat: once Nu is your primary bank, the switching cost is your whole financial life, and each new customer costs almost nothing to serve. Mexico and Colombia are the same playbook, years behind.

The numbers

The earnings line has stepped up every quarter for a year and beaten the estimate each time.

Q2 delivered the biggest profit in the company's history, $0.22 against a $0.198 estimate, on gross revenue of nearly $5.9 billion, up 39% year over year. Net interest income reached $3.7 billion, and the margin on lending widened 180 basis points in a single quarter to 22.9%, driven by the deliberate shift into unsecured loans.

The Street models revenue compounding into the low thirties of billions and earnings roughly doubling across two years.

The lender's real gauge is credit, and here the two moving parts tell opposite stories.

Early delinquency improved 16 basis points, mostly seasonal, and the cost of credit fell 9% to $1.7 billion. But the serious bucket, loans 90-plus days late, rose another 35 basis points to 6.9%, exactly the drift flagged last quarter as the unsecured book seasoned, and it did not reverse: the deep-delinquency watch-item held its upward path rather than rolling over. That is the tell in the whole memo, early buckets healing while the back end keeps climbing, and it means today's lower cost of credit is a timing gift, not a structural win.

The compounding case rests on return: Nu earns about a 30% return on tangible equity, and with no dividend that book compounds on itself while ARPAC rises and Mexico's lending is barely started. The variant view is narrow. The market prices the growth but discounts the unsecured push, betting losses catch up; if 90-plus delinquency stabilizes near 7% while the risk-adjusted margin keeps expanding, 14 times forward is too cheap for a 30% compounder. The next two prints of 90-plus NPL and cost of credit settle it.

Management

Founders David Vélez and Cristina Junqueira still run and own a large slice of this. The record over the last year is one-directional selling: zero insider buys, and about $10.3 million sold across seven Form 4s, led by Fragelli's $3.5 million in August and Junqueira's combined $4.4 million in March. Plan status is not disclosed on the filings, so planned and discretionary cannot be separated, which is itself a small demerit at a company that could easily have shown 10b5-1 cover. The sales are modest against the float and against founders' remaining stakes, and capital is otherwise fully reinvested, no dividend, no buyback, which is the right call while equity earns 30%. Guidance has been met or beaten every quarter shown here.

How it fails or surprises you

The unsecured book turns. The $10.3 billion unsecured portfolio is the intentional risk expansion, and 90-plus NPL is already 6.9% and climbing 35 basis points a quarter. A Brazilian consumer slowdown pushes cost of credit back above $2 billion a quarter and erases the margin gain within two prints. This is the fact the bull read explains least well.

Mexico becomes a second Brazil (right tail). Mexico holds $5.7 billion of deposits against a loan-to-deposit ratio of just 35%, almost no lending yet. If Nu turns that funding into a Brazil-style credit book, a whole second earnings engine appears that today's price ignores. The first standalone Mexico profit disclosure is the tell.

Funding or currency squeezes the margin. Deposits cost 88% of the interbank rate today. A jump in Brazilian rates, or a real that moves against the dollar reporting, compresses the spread that the entire model prints on, independent of credit quality.

Closing thoughts

The next two quarters of 90-plus day NPL and cost of credit tell you whether the unsecured push is working or breaking. Underneath that sits Brazil's economy and currency, which no quarterly report predicts. The fatter tail is upside: a 30%-return franchise still adding customers and barely lending in two of three countries, priced at a market multiple. But the left tail is genuine and can be permanent, a domestic recession landing on a book that was deliberately steered toward higher-risk, later-loss loans while the deep-delinquency line was already rising. What is at risk if credit breaks is the entire margin story; what the upside is worth is a re-rating on top of doubling earnings.

The bet is still that Nubank keeps signing up people across Brazil, Mexico and Colombia, gets them to borrow more, and keeps getting paid back. What breaks it is the pair to watch: 90-plus day NPL and cost of credit. If delinquency pushes through roughly 8% while the risk-adjusted margin stalls over the next two quarters, the unsecured bet is breaking; if it holds near 7% and Mexico lending inflects, you are underpaying at 14 times forward.