Dlocal Limited DLO

Key data

The business

dLocal is one API that lets a company in the US, Europe or China collect from and pay out to consumers across 40-plus emerging markets, using whatever local rail those consumers actually use: Pix in Brazil, UPI in India, Mada in Saudi Arabia, not a foreign Visa charge that gets declined. The pitch is conversion: the same card clears up to 20 percentage points more often when dLocal routes it locally, and the company now touches roughly 38% of India's credit-card processing and 14% of Brazil's debit volume. It keeps a sliver of each dollar of Total Payment Volume. The moat is the licensing and banking plumbing in dozens of hard countries plus that conversion lift, which together make it painful for a merchant already live to rip dLocal out. It is capital-light: property spend is under 1% of revenue.

The numbers

The engine is volume, and volume is compounding hard while the price of that volume erodes. The clearest way to see it is gross profit divided by TPV, the cents kept per dollar moved. Q3 2025 TPV was not disclosed in the available filings.

The erosion is the whole story. Q2 2026 extended it: TPV jumped 92% year over year to $17.7 billion, revenue hit $400 million, but the cents kept per dollar of volume fell to 0.72%, the lowest in this sequence. Management calls it deliberate: Brazil normalizing off a strong Q4, a mix shift toward big lower-priced merchants, and narrower currency spreads in Egypt. A securities case alleging dLocal hid this exact take-rate trend was dismissed on appeal in April, the court noting volume, revenue and gross profit all kept growing through it.

Revenue compounded 46% a year since 2021; EPS only 27%, and 2024 actually went backwards on the same margin squeeze plus a tax item before 2025 rebounded. That gap between the two growth rates is the tension. TTM revenue is now $1.36 billion (Q3 2025 through Q2 2026). Against consensus, the 2028 estimate is $2.51 billion of revenue and $1.33 of EPS, so the market already expects volume to keep winning; at 12 times that number, it is not paying much for it. What I believe the market underrates: at 32% return on invested capital and near-zero capital intensity, if the cut merely stops falling, gross-profit dollars compound with volume and this is far too cheap. The print that settles it is one quarter where gross profit grows in line with TPV.

Management

Pedro Arnt, MercadoLibre's former finance chief, runs it now; founder Sebastian Kanovich has stepped back and, over the last twelve months, sold about $16.2 million of stock across four sales, the largest $14.6 million on July 1, against a single $237,000 insider buy. Plan status is not disclosed in the filings I can see, so read the founder's selling as neither reassuring nor damning without the 10b5-1 footnote. Capital return is real and growing: $86 million of buybacks and $57 million of dividends flowed out in the recent period, funded from a balance sheet carrying net cash of $727 million (market cap less enterprise value). Returns on equity near 38% say the underlying franchise earns well; the open question is price per dollar of volume, not whether the model works.

How it fails or surprises you

The cut keeps outrunning the volume. Kept-per-dollar fell from 1.08% in Q2 2025 to 0.72% by Q2 2026. If that pace outruns 50%-plus volume growth, gross-profit dollars flatten and 22 times earnings becomes a value trap. First tell: a quarter where gross profit grows slower than TPV two periods running.

Operating leverage finally shows (right tail). Volume is compounding above 90% on a cost base that barely moves and ROIC of 32%. If the cut stabilizes anywhere near 0.70% while TPV keeps compounding, gross profit dollars inflect up sharply and the 12-times-2028 multiple is wrong by half. The market is paying for the fear, not the volume.

One country or one merchant breaks. Net revenue retention is a huge 152%, which means the top merchants also concentrate the risk, and the money rides through Brazilian, Nigerian, Argentine and Egyptian currencies. A capital-control freeze or a large merchant leaving would hit fast, exactly as margins whipsawed in 2024.

Closing thoughts

This settles fast. The question is whether the take-rate compression stops, and one quarter where gross profit grows with volume answers it. If it does, volume compounding at 50%-plus against a fixed cost base makes today's multiple look cheap, and that upside tail is the fatter one, because the volume growth is already in the filings and the compression is at least partly a choice to trade price for scale. The left tail is a genuine emerging-market accident, a currency freeze or a lost anchor merchant, and only the net-cash balance sheet, not any forecast, protects against that. What is at risk if the downside linchpin breaks is the entire earnings-growth thesis; what the upside is worth is a re-rating on gross-profit dollars that finally compound with volume.

The bet is still that dLocal moves enough new money for the world's big companies into and out of emerging markets to outrun the shrinking cut it earns on each dollar. It breaks when the cut falls faster than volume rises, and the one pair to watch is TPV growth against gross-profit growth: the quarter they converge is the signal the thesis is proven, the quarter gross-profit dollars stop growing is the signal it is broken.