XPCompany report
XP Inc. XP
The bet you're really making is that Brazilians keep pulling their savings out of the big banks and moving that money onto XP's investing platform, and keep paying XP a slice to look after it. You're betting they keep coming even as Nubank and Itaú chase the same money, and that XP earns a little more per client each year. Right now it is going well: the biggest quarterly profit yet, $0.53 a share, higher in each of the last four quarters and up from $0.45 late last year. You pay about 10 times earnings, near the cheapest the stock has been since it listed in 2019.
Key data
XP · price with moving averages
Source: market data.
The business
XP is where a growing share of Brazil's savers keep their money instead of the big banks. It began as a brokerage routing orders through a network of independent financial advisors, the largest such network in the country, and has since bolted on funds, retirement products, cards, insurance and a bank account, so a client who arrives to buy one stock ends up holding half of their financial life on the platform. The money is made two ways: fees and spreads on the assets clients hold and move, and net interest on the cash and credit sitting on XP's own balance sheet. The advisor at the client's kitchen table is the moat, because switching means firing a person you trust, not closing an app, which is why client money, once on, tends to stay and compound. Nubank now chases the same saver from below on price, and Itaú from above on scale.
The numbers
Earnings per share is the cleanest series XP prints, and it has climbed each of the last four quarters even as the beat against estimates has been uneven.
| Quarter | Revenue | Net income | Diluted EPS |
|---|---|---|---|
| Q2 2025 | R$4.38B | R$1.27B | R$2.46 |
| Q3 2025 | R$4.59B | R$1.29B | R$2.47 |
| Q4 2025 | R$5.06B | R$1.32B | R$2.49 |
| Q1 2026 | R$4.62B | R$1.29B | R$2.42 |
| Q2 2026 | R$4.99B | R$1.42B | R$2.74 |
Two beats, two narrow misses, but every quarter's actual topped the one before, and Q2 was the sharpest step up of the run, a 13% jump off Q1.
Zoom out and the profit line has compounded through a brutal rate cycle and a weak Brazilian tape.
| Fiscal year | Revenue | Net income | Diluted EPS |
|---|---|---|---|
| 2021 | R$13.0B | R$3.73B | R$6.26 |
| 2022 | R$14.2B | R$3.71B | R$6.25 |
| 2023 | R$14.8B | R$3.84B | R$7.16 |
| 2024 | R$19.9B | R$5.18B | R$8.23 |
| 2025 | R$18.2B | R$5.07B | R$9.72 |
| 2026, 1H to Jun | R$9.61B | R$2.71B | R$5.16 |
Net income compounded about 21% a year from 2020 to 2024, and operating cash flow swung from a big outflow in 2021 to R$11B by 2024. Interim figures for 2026 weren't in this run's data. The cash line is noisy for a broker, because client balances slosh through it, so read the profit line as the truth and the cash as confirmation that the profit is real. Consensus has earnings compounding toward roughly $13.50 a share by 2028, which even on conservative currency leaves the forward multiple in the high-single digits.
Here is the whole argument in one table: the earnings kept rising while the market kept paying less for them.
| Year-end | P/E | P/B |
|---|---|---|
| 2021 | 24.9x | 6.2x |
| 2022 | 12.6x | 2.6x |
| 2023 | 17.7x | 3.5x |
| 2024 | 8.9x | 2.0x |
| 2025 | 9.1x | 2.0x |
| Now | 10.0x | 2.1x |
The stock de-rated from 25 times to 9 while profit rose every year. What this memo believes that the tape does not is that the de-rating over-corrected for a growth scare that the numbers are not confirming. The single print that settles it is net new money: if it stays positive and the fee take holds for two straight quarters, 10 times a 21%-compounder is a mistake, and if net new money rolls over, 10 times is simply the honest price of a business whose land-grab is finished.
Management
The record is a founder-led firm that has grown book value and profit through the worst rate backdrop in a decade, and turned operating cash sharply positive since 2023. Reported return on equity runs around 22%, high for a financial and consistent with a capital-light distribution model rather than a balance-sheet lender. There was no open-market insider buying or selling in the trailing twelve months, so there is neither a vote of confidence nor a warning to read into the founder's own account this window, and 10b5-1 plan status is not disclosed. Pay design and buyback pace were not in this run's data, which is the one real hole in judging how shareholder-friendly the capital allocation actually is.
How it fails or surprises you
The rate cycle turns against the mix. Brazil's Selic sits high, and a fast cut helps XP's equity flows over years but squeezes the net interest it earns on client cash in the near term. Watch net interest income and the blended take rate together over the next two quarters, because a falling rate that arrives before flows recover is the pinch that shows up first.
The take rate bleeds to competition. Nubank and BTG are pricing to win the same client, and XP's edge is an advisor network that costs money to keep loyal. If blended retail take rate compresses for two straight quarters while advisor payouts hold, the moat is being rented, not owned, and the next print to reveal it is fee revenue per real of custody.
The de-rating fully reverses (right tail). The market is paying 10 times for a business it once paid 25 times for, and pricing in almost no re-rating. If earnings keep stepping up as Q2 did and net new money stays positive, a move back toward even the mid-teens multiple it held in 2023 doubles the stock without a single new client, and the first sign is two more quarters like the last one.
Closing thoughts
This is a case where one specific print settles the debate rather than a fog that never clears. The bear read and the bull read agree on the facts, cheap on history, compounding through a hard cycle, and disagree only on whether the growth is maturing or merely paused. Net new money and the fee take, read together for two quarters, convert the story one way or the other, and an ambiguous print, flat flows with a soft take rate, is itself the answer that the ex-growth price was fair. On judgment, not arithmetic, the left tail looks partly spent, because 10 times already prices a Brazil discount and a growth scare, while the right tail needs the market to re-believe a story the numbers are quietly supporting.
The bet is still that Brazilians keep pulling their savings out of the big banks and moving that money onto XP's platform, and keep paying XP a slice to look after it. It breaks if net new money stalls and the take rate slips in the same two quarters, and those are the one pair of numbers that tell you first. If earnings keep climbing while custody grows and the stock still sits at 10 times a year from now, the market will have told you the growth is over, and the numbers will have told you it was not.
Methodology
The year-to-date row is the sum of the 2 reported quarters of the current fiscal year, diluted EPS included; the five-quarter and five-year tables are the vendor income statements.
Sector frame: Brazilian capital-markets platform, valued on earnings and book against domestic financial peers (Nubank, BTG, Itaú), with the rate cycle as the dominant external variable. Data gaps: quarterly revenue and net income, net new money, client assets under custody, take rate, executive compensation and buyback pace were not in this run's data; the fiscal-year "Revenues" XBRL tag understates XP's total net revenue and was therefore left out in favor of the net income series. Bundle: net income and operating cash flow from as-filed annual XBRL (2020-2024, BRL); quarterly EPS actuals and estimates (USD), valuation history, and the live quote (USD) pulled this run. Sources: as-filed 6-K (filed 2026-08-17), annual XBRL series, vendor year-end P/E and P/B history, consensus estimates, and the real-time quote. Fact check: 1 currency error corrected (forward consensus EPS is USD not BRL); quarterly estimate differences are derived calculations; vendor estimates rounded to nearest cent. All bundle financials reconciled to as-filed XBRL. No Critical qualitative claims required web verification. Final analysis verified as of Sep 6, 2026.
Bid Cap
Daily ideas, a 390-name database, and a model long/short book from an investor who mostly covers financials. $70 a month or $700 a year.
Subscribe on Substack


