Analysis of the Form 20-F | Fiscal year ended 31 December 2025
FY2025, ended 31 December 2025. Reporting currency: US dollars. Accounting framework: IFRS as issued by the IASB. Amounts are in US$ millions unless otherwise stated; the document reports in thousands, and the data table published alongside this report keeps that unit. The multiples use only the closing price of US$14.14 on 31 December 2025 and the 294,931,956 shares outstanding at 31 December 2025 shown on the cover page: both were supplied to the analysis rather than taken from the document, and both refer to the same date, the last day of the fiscal year.
dLocal sits between two worlds that struggle to talk to each other. On one side are large digital companies based in the United States, Europe and China; on the other, their customers, suppliers and contractors in emerging markets where international cards work poorly, payment methods are local, and currency controls make it complicated to move money in or out. The company processes local payments through its own network of acquirers and processors and, through arrangements with financial institutions, repatriates or expatriates the funds and settles with the merchant in its currency of choice, usually dollars or euros.
Transactions come in two kinds. Pay-ins are collections: dLocal gathers money in local currency from consumers in an emerging market and makes it available to the merchant in its own country and currency after a settlement period. Pay-outs run the other way: the merchant pays in its own currency and dLocal distributes the money in local currency to suppliers, drivers, marketplace sellers, landlords or refund recipients. In both cases revenue arises per authorised transaction, as a percentage of value or a fixed amount, and a foreign exchange fee is added when the transaction crosses a currency border.
The document reports a single operating segment, "payment processing", and warns that disaggregated information exists only at the revenue level: "Disaggregated information is only reviewed at the revenue level with no corresponding detail at any margin or profitability levels". There are therefore no margins by geography or by product line, and that limit bears on everything that follows.
Revenue splits into two lines only. Transaction revenues are US$1,054.9 million, 96.5% of the total, and comprise processing fees, foreign exchange fees, instalments, advances granted to merchants, chargebacks and refunds. Other revenues, US$38.7 million, gather transactional taxes passed through, monthly minimums, transfer fees and setup fees. No revenue is recognised over time: the contract is exhausted at the level of the individual transaction and, as the document states, merchant agreements are open-ended and terminable by either party without penalty once notice has lapsed.
Revenue by type of service
| Item (US$ mn) | 2025 | % of total | 2024 | 2023 | YoY 25/24 | YoY 24/23 |
|---|---|---|---|---|---|---|
| Transaction revenues | 1,054.9 | 96.5% | 739.9 | 637.1 | +42.6% | +16.1% |
| Other revenues | 38.7 | 3.5% | 6.1 | 13.3 | +538.5% | −54.5% |
| Total | 1,093.6 | 100.0% | 746.0 | 650.4 | +46.6% | +14.7% |
Geography is the second axis, and the document builds it on the country where the merchant's end users executed the payment, noting that this does not imply the revenue is generated or taxed in that country. Latin America accounts for almost four fifths, and Brazil alone for almost a fifth.
Revenue by region
| Region (US$ mn) | 2025 | % of total | 2024 | 2023 | YoY 25/24 |
|---|---|---|---|---|---|
| LatAm | 874.1 | 79.9% | 562.2 | 492.7 | +55.5% |
| — Brazil | 207.2 | 18.9% | 152.0 | 159.0 | +36.3% |
| — Mexico | 183.0 | 16.7% | 149.2 | 116.8 | +22.6% |
| — Argentina | 161.1 | 14.7% | 85.5 | 75.1 | +88.5% |
| — other countries | 322.8 | 29.5% | 175.5 | 141.7 | +83.9% |
| Non-LatAm | 219.4 | 20.1% | 183.8 | 157.7 | +19.4% |
| — Egypt | 59.7 | 5.5% | 94.0 | 36.7 | −36.4% |
| — other countries | 159.7 | 14.6% | 89.8 | 121.0 | +77.8% |
| Total | 1,093.6 | 100.0% | 746.0 | 650.4 | +46.6% |
Two lines deserve to be read together. Argentina grows 88.5% and takes its weight to 14.7% of revenue, in a country the document treats as a hyperinflationary economy under IAS 29 and where administrative and judicial inquiries into the local subsidiary were opened in 2023. Egypt, which in 2024 was the largest market outside Latin America at US$94.0 million, falls to US$59.7: a 36.4% contraction that the document does not comment on anywhere.
At year-end the group had 1,274 full-time employees, against 1,095 and 901 in the two prior years, of whom 832 in the Americas. The document is signed by the chief financial officer alone, Guillermo López Pérez; the chief executive, Pedro Arnt, signs the attached certifications. The document explicitly names as former officers two people who appear among the defendants in the class actions: Sebastián Kanovich, formerly co-chief executive and today a director, and Diego Cabrera Canay, formerly chief financial officer.
Among the things this fiscal year introduces relative to the comparative columns, three are stated by the document itself. The US$200 million repurchase programme authorised in May 2024 was not used for a single dollar in 2025, and the treasury shares accumulated earlier were cancelled in August. The first dividend of the three years covered by the document was paid, US$150.0 million. And on 23 December 2025 a reconstruction of the board was completed: five directors left, among them co-founder Sergio Fogel, Andrés Bzurovski took the chair, and three new committees were established, including one for compensation.
Fast grower. Two figures from the document settle it: revenue grows 46.6% in a year and processed volume 59.6%, with reinvestment concentrated in working capital and capitalised software rather than in plant. It is not a stalwart, because margins are not stable but move by several points from one year to the next: the Adjusted EBITDA margin went from 31.1% to 25.3% to 25.4% across the three available years. It is not a cyclical, because the document ties results not to a price or a replacement cycle but to the growth of its own customers' volumes.
The classification changes what matters. In a fast grower, cash flow and return on capital count for more than any single year's margin, and the real question is whether volume growth translates into growth in profit per unit of volume. At dLocal it does not.
What stops a merchant from leaving? The document offers three elements, all verifiable, and none of the three is a price barrier.
The first is the network. The group operates in more than forty countries through over seventy subsidiaries, roughly three quarters of them locally incorporated "collection entities", and in December 2024 obtained authorisation as an Authorised Payment Institution in the United Kingdom, alongside its electronic money and payment institution licences in the European Union and its registration as a Money Service Business with FinCEN. Replicating that infrastructure licence by licence is slow and expensive, and this is the real moat.
The second is the cost of integration, which the document acknowledges from the opposite side when it lists among its risks merchants' reluctance to change provider: "Many potential merchant clients may have concerns about disadvantages associated with switching payment processing vendors". The same sentence that explains why a customer is hard to win explains why one is hard to lose.
The third shows up in the numbers. 98.8% of processed volume comes from merchants that have been on the platform for more than two years, and the revenue growth of 2025 comes almost entirely from existing customers: US$1,080.9 million out of US$1,093.6 total, with new merchants contributing US$12.7 million, 1.2% of revenue. The stated net revenue retention rate is 145%, against 113% the prior year.
That concentration on the existing base is also the weakness. The top ten merchants are worth 61% of revenue, and in 2025 two of them individually exceed 10%, against one in 2024 and none in 2023. dLocal's competitive advantage is real on the infrastructure side but applies to a very small number of counterparties, each of them large enough to renegotiate.
The document publishes no average prices, but it makes them derivable, because it discloses both revenue and processed volume. The ratio between the two is the effective price per dollar transacted, and its trajectory is the single most important figure in the whole filing.
| Item | 2025 | 2024 | 2023 |
|---|---|---|---|
| TPV (US$ bn) | 40.8 | 25.6 | 17.7 |
| Revenue / TPV | 2.679% | 2.917% | 3.679% |
| Gross profit / TPV | 0.987% | 1.152% | 1.566% |
In two years the price per unit of volume fell by a full percentage point, from 3.679% to 2.679%, that is by more than a quarter. Gross margin per unit of volume fell faster still, from 1.566% to 0.987%, because processing costs do not compress at the same speed as price. Put differently: for every dollar processed in 2023 the company kept a little over one and a half cents of gross profit; in 2025 it keeps less than one.
The document gives one explanation, contractual and not transitory: "As the level of volume processed with some of our global merchants increases, our revenue may grow to a lesser extent due to tiered pricing in our contracts with such merchants". Prices step down with volume, so customer success mechanically reduces unit price. A second possible cause is mix, since the document states that revenue varies depending on whether a payment is cross-border or local, but it never quantifies the composition of volume and so does not allow the effect to be measured.
There is a third element, on the cost side, that the company says it does not control: "We are unable to predict if or when financial institutions will increase or decrease their fees", adding that it is not always able or willing to pass those changes on to merchants. A company that sets neither the price at which it sells, because that price declines with the customer's volume, nor the cost at which it buys, has limited pricing power. That revenue still grows 46.6% is a function of volume, not of price.
In 2025 growth is entirely organic. The cash flow statement reports no acquisition net of cash acquired, the document shows no goodwill in either balance sheet year, and the only change of perimeter declared is the incorporation of a United States subsidiary that the group itself qualifies as not constituting a business combination under IFRS 3, alongside the dissolution of a company in Costa Rica. Earlier transactions did occur — the document cites the acquisition of PrimeiroPay in the first half of 2021 and carries US$41.3 million of gross cost of acquired intangibles — but they fall outside the three comparative years.
The first acquisition of the three years covered by the document arrives after year-end, and it has a history. In June 2025 the group announced its intention to acquire AZA Finance, a cross-border payments company in Africa, and in the meantime extended it a working capital credit facility, carried on the balance sheet at a fair value of US$24.1 million. In July a third party commenced legal proceedings against AZA, the transaction was restructured to the most relevant assets, the proceeding was resolved and withdrawn in November, and on 27 February 2026 the group exercised the option, acquiring 100% of Mint Code Solutions S.A. in Cameroon and certain assets from NeWurth S.A. for US$23.0 million, settled with one dollar in cash and the release of that same credit facility.
One transparency point remains unresolved in the numbers themselves. The document states in one passage that new merchants accounted for 2% of revenue, or US$17 million; the table that splits revenue between existing and new merchants shows US$12.7 million, 1.2%. The two statements cannot both be true, and the table is the only one of the two that sums to the total.
| Item (US$ mn) | 2025 | 2024 | 2023 | YoY 25/24 | YoY 24/23 |
|---|---|---|---|---|---|
| Revenue | 1,093.6 | 746.0 | 650.4 | +46.6% | +14.7% |
| Cost of services | 690.8 | 451.3 | 373.5 | +53.1% | +20.8% |
| Technology and development | 30.7 | 25.6 | 12.7 | +19.8% | +102.6% |
| Sales and marketing | 26.5 | 21.6 | 17.1 | +22.3% | +26.3% |
| General and administrative | 118.8 | 101.2 | 70.6 | +17.3% | +43.4% |
| Total operating expenses | 175.9 | 148.5 | 100.3 | +18.5% | +48.0% |
| Personnel cost (including capitalised) | 139.0 | 108.5 | 74.8 | +28.1% | +45.1% |
Scale exists, but all of it sits above the cost of services line and none below. Operating expenses grow 18.5% against revenue at 46.6%, and their share of revenue falls from 19.9% to 16.1%: it is this, and not gross margin, that produces the 56.5% jump in operating profit. Below the line the direction is the opposite: cost of services grows faster than revenue for the second consecutive year, and gross margin falls from 42.6% in 2023 to 39.5% and then to 36.8%.
The contrast is sharp. dLocal's model scales well on overhead, because a platform processing forty billion does not cost much more than one processing twenty-five, and scales badly on variable cost, because every additional transaction carries the local financial institution's fee with it. As long as operating expenses have room to compress, operating profit can grow faster than revenue even with price falling. When that room runs out, what is left is the trajectory of gross margin.
| Item (US$ mn) | 2025 | 2024 | 2023 | YoY 25/24 | YoY 24/23 |
|---|---|---|---|---|---|
| Revenue | 1,093.6 | 746.0 | 650.4 | +46.6% | +14.7% |
| Cost of services | (690.8) | (451.3) | (373.5) | +53.1% | +20.8% |
| Gross profit | 402.8 | 294.7 | 276.9 | +36.7% | +6.4% |
| Gross margin | 36.8% | 39.5% | 42.6% | −2.7 pp | −3.1 pp |
| Technology and development | (30.7) | (25.6) | (12.7) | +19.8% | +102.6% |
| Sales and marketing | (26.5) | (21.6) | (17.1) | +22.3% | +26.3% |
| General and administrative | (118.8) | (101.2) | (70.6) | +17.3% | +43.4% |
| Impairment on financial assets | (2.2) | (0.4) | 3.1 | n.m. | n.m. |
| Other operating loss | (4.7) | (5.3) | — | −10.3% | n.m. |
| Operating profit | 219.9 | 140.5 | 179.7 | +56.5% | −21.8% |
| Operating margin | 20.1% | 18.8% | 27.6% | +1.3 pp | −8.8 pp |
| Finance income | 40.8 | 66.9 | 128.2 | −39.0% | −47.8% |
| Finance costs | (27.9) | (49.7) | (116.8) | −44.0% | −57.5% |
| Inflation adjustment | (4.2) | (6.7) | (12.5) | −36.8% | −46.9% |
| Profit before income tax | 228.7 | 151.0 | 178.5 | +51.4% | −15.4% |
| Income tax expense | (31.8) | (30.6) | (29.4) | +3.9% | +3.8% |
| Effective tax rate | 13.9% | 20.2% | 16.5% | −6.3 pp | +3.7 pp |
| Profit for the year | 196.9 | 120.5 | 149.1 | +63.4% | −19.2% |
| Net margin | 18.0% | 16.1% | 22.9% | +1.9 pp | −6.8 pp |
| Basic earnings per share (US$) | 0.68 | 0.42 | 0.51 | +61.9% | −17.6% |
| Diluted earnings per share (US$) | 0.65 | 0.39 | 0.49 | +66.7% | −20.4% |
| Weighted average diluted shares (mn) | 301.7 | 305.1 | 303.0 | −1.1% | +0.7% |
Net profit growth of 63.4% exceeds operating profit growth of 56.5%, and the difference does not come from operations: net finance results fall from US$10.5 million to US$8.7 million. It comes from tax, whose effective rate drops 6.3 points even as pre-tax profit grows 51.4%. At the 2024 rate, 2025 profit would have been US$182.4 million instead of US$196.9, and growth would have stopped at 51.4%.
The income statement shows revenue on two lines — transaction revenues and other revenues — but not their respective costs, because cost of services is a single caption. A margin split by revenue line is therefore not possible, and that is a limit declared by the document rather than a choice of this report: the segment note says explicitly that disaggregated information stops at revenue. Where decomposition is possible, on the cost of services, it says something precise: processing costs grow 53.4% and are worth US$655.0 million of US$690.8 million of total cost, 94.8%. Everything else in the cost of services, including amortisation of capitalised software, weighs less than six points.
| Item (US$ mn) | 31/12/2025 | 31/12/2024 |
|---|---|---|
| Cash and cash equivalents | 719.9 | 425.2 |
| — of which corporate cash | 424.5 | 189.0 |
| — of which merchants' cash | 295.4 | 236.1 |
| Financial assets | 99.1 | 129.3 |
| Trade and other receivables (current) | 572.0 | 496.7 |
| Other current assets and derivatives | 29.7 | 21.7 |
| Total current assets | 1,420.8 | 1,072.9 |
| Intangible assets | 74.0 | 63.3 |
| Property, plant and equipment and right-of-use assets | 7.0 | 7.0 |
| Non-current receivables, deferred tax and other | 39.3 | 28.1 |
| TOTAL ASSETS | 1,541.0 | 1,171.3 |
| Payables to merchants | 812.9 | 562.7 |
| Accrued and other payables | 41.5 | 35.0 |
| Financial liabilities | 86.9 | 50.5 |
| Lease liabilities (current and non-current) | 3.4 | 4.0 |
| Tax liabilities, derivatives, provisions | 23.5 | 28.2 |
| Deferred tax liabilities | 3.3 | 1.9 |
| TOTAL LIABILITIES | 971.5 | 682.3 |
| Total equity | 569.4 | 489.0 |
| Gross debt (financial liabilities + leases + net derivatives) | 91.7 | 57.8 |
| Net financial position as the company defines it | +727.3 | +496.7 |
| Net financial position excluding merchants' cash | +431.9 | +260.5 |
dLocal's balance sheet is made almost entirely of money in transit. Of US$1,541.0 million of assets, US$719.9 is cash, US$572.0 receivables from acquirers, processors and merchants, US$99.1 securities: US$1,391.0 million, 90.3% of assets, are short-term financial assets. On the other side, US$812.9 million of payables to merchants. Fixed assets of every kind are worth US$81.0 million, 5.3% of assets.
The net financial position deserves attention because the document gives only one version of it, and it is not the one that belongs to the shareholder. Note 30 computes net cash of US$727.3 million by adding up all liquidity; but Note 14 states that US$295.4 million of that liquidity are "freely available funds which belong to the merchants or their customers but are held by the Company". Net of that portion, the net cash attributable to the company is US$431.9 million. It remains a solid position — no medium-term debt, no covenants disclosed, financial liabilities maturing within three months — but it is worth 59.4% of the figure the document highlights.
| Item (US$ mn) | 2025 | 2024 | 2023 |
|---|---|---|---|
| Pre-tax profit adjusted for non-cash items | 279.4 | 163.1 | 200.6 |
| Change in working capital | 163.1 | (167.2) | 104.3 |
| Income tax paid | (27.0) | (28.7) | (11.5) |
| Net cash from operating activities | 415.5 | (32.8) | 293.5 |
| Capital expenditure (tangible and intangible) | (36.8) | (22.6) | (18.2) |
| IFRS free cash flow (operating cash less capex) | 378.7 | (55.4) | 275.2 |
| of which change in merchant working capital | 188.0 | (146.0) | 71.3 |
| Adjusted Free Cash Flow (company definition) | 190.7 | 90.6 | 203.9 |
| Share-based compensation | 24.1 | 23.8 | 11.9 |
| Share repurchases | — | 101.1 | 97.9 |
| Dividends paid | 150.0 | — | — |
Here lies the most important element of the whole cash analysis, and it is a rare one: management's adjusted measure is more conservative than the one derivable from the statements. Free cash flow computed the usual way, operating cash less capex, gives US$378.7 million; but US$188.0 million of those are the net increase in merchant funds temporarily held, money that does not belong to the company and that will flow out when volumes stop growing or the settlement cycle shortens. The Adjusted Free Cash Flow of US$190.7 million removes exactly that component, and the comparison with 2024 shows why it is needed: in that year IFRS operating cash was negative by US$32.8 million even though the company produced US$90.6 million of real cash.
Share-based compensation is US$24.1 million, 2.2% of revenue and 12.7% of Adjusted Free Cash Flow. The weighted average diluted share count falls 1.1%, but shares outstanding rise by 9.5 million, from 285.5 to 294.9: 1.3 million issued on option exercises and 8.2 million on warrant exercises, with no repurchase to offset them. The US$200 million repurchase programme authorised in May 2024 went unused through all of 2025 and expired; the company chose to return capital through a US$150.0 million dividend instead.
Owner earnings, computed as net profit plus non-cash items less capex and less the additional working capital absorbed in the year, give US$190.7 million if the working capital considered is only the company's own, and coincide by construction with Adjusted Free Cash Flow. With no working capital deduction at all they would be US$378.7 million. The first measure is used throughout the rest of this report, because the second incorporates other people's money. Share-based compensation is not deducted a second time, since net profit already contains it, and unrealised fair value gains on financial assets of US$14.5 million are excluded because they are not cash.
| Metric | 2025 | 2024 | 2023 |
|---|---|---|---|
| Gross margin | 36.8% | 39.5% | 42.6% |
| Operating margin | 20.1% | 18.8% | 27.6% |
| Net margin | 18.0% | 16.1% | 22.9% |
| Adjusted EBITDA margin | 25.4% | 25.3% | 31.1% |
| Revenue / TPV | 2.679% | 2.917% | 3.679% |
| Adjusted Free Cash Flow (US$ mn) | 190.7 | 90.6 | 203.9 |
| Adjusted FCF / net profit conversion | 96.8% | 75.2% | 136.8% |
| Capex / revenue | 3.4% | 3.0% | 2.8% |
| SBC / revenue | 2.2% | 3.2% | 1.8% |
| ROIC (NOPAT on year-end invested capital) | 137.7% | 49.1% | not available |
| Gross debt / Adjusted EBITDA | 0.33x | 0.31x | not available |
| Revenue per employee (US$ thousands) | 858.4 | 681.3 | 721.8 |
Return on invested capital requires a declared choice, because the usual definition does not work. If all liquidity is deducted, including the merchants', invested capital turns negative and the ratio loses meaning. The invested capital used here is therefore equity plus gross debt less corporate cash and financial assets only: US$137.6 million at the end of 2025 and US$228.4 million at the end of 2024, on year-end balances in both years. The result, 137.7%, says one thing but says it forcefully: this business does not need shareholders' capital to operate, because it works on its customers' money.
On multiples, the 31 December 2025 price and the cover-page share count at the same date give a market capitalisation of US$4,170.3 million and an enterprise value of US$3,738.5 million.
| Multiple | Value |
|---|---|
| Price / net profit | 21.2x |
| Price / diluted earnings per share | 21.8x |
| Price / Adjusted Free Cash Flow | 21.9x |
| Adjusted Free Cash Flow yield | 4.57% |
| Price / revenue | 3.81x |
| Price / book value | 7.32x |
| Enterprise value / Adjusted EBITDA | 13.4x |
| Enterprise value / operating profit | 17.0x |
| Operating profit / enterprise value | 5.88% |
| Dividend yield on the dividend paid in 2025 | 3.71% |
| Dividend yield on the dividend declared for 2026 | 1.37% |
Debt appears in the document on a single basis, carrying value, because the financial liabilities mature within three months and the document states that for them carrying value approximates fair value. The enterprise value therefore uses the carrying amount, and the question of three different measures of debt does not arise for this company.
How much of these numbers is management's opinion? The auditor points to one difficult area, but the notes contain six more that weigh as much.
Transaction revenues (Critical Audit Matter). The auditor isolates a single topic: transaction revenues, US$1,054.9 million, 96.46% of consolidated revenue. The difficulty it declares is not one of principle but of process: "the complexity of the Company's processes for setting the transaction revenues with their merchants mostly defined individually". Commercial terms are negotiated merchant by merchant, so there is no price list against which to verify a revenue figure, and the audit runs through controls testing and sample tracing to source documents. What would move this line is not an accounting estimate but the soundness of internal controls over a process fragmented across dozens of contracts: it is why the auditor assigned specialists to it.
Allowance for credit losses. It is worth US$0.9 million on gross trade receivables of US$521.7, 0.16%, and it is the note where discretion weighs most relative to the amount recorded. Note 16 states an average expected loss rate of 0.3%, up from 0.1% the prior year, and adds that no receivable is past due and that all are classified as "normal" credit risk. Note 30, however, presents the whole receivable book, US$572.0 million including advances, split by rating class with the corresponding expected loss rates, and in 2025 those rates are two orders of magnitude higher than in 2024: 5.00% on the BB class, where US$225.7 million sits, and 9.00% on the CCC class, where US$112.8 million sits, against 0.04% and 0.06% applied to the same classes a year earlier.
Applying the Note 30 rates to the amounts in the same note: 0.01 × 44.4 + 0.07 × 10.4 + 0.05 × 225.7 + 0.03 × 80.9 + 0.09 × 112.8 = US$25.0 million, that is an average expected loss of 4.38%, against US$0.9 million recorded and against the US$1.7 million that the 0.3% average rate declared in Note 16 would produce.
Three incompatible numbers for the same quantity. The proof that the gap arises in 2025 and is not a structural defect of the table lies in the prior year: applying the 2024 rates to the 2024 amounts gives US$0.2 million against US$0.148 recorded, the same order of magnitude. One year later the same exercise produces a gap of twenty-nine times. The document offers two partial mitigations, and neither closes the question: it states that the classes are ratings of the countries where the receivables originate rather than of the counterparties, and it states that no receivable is past due. But it still titles that column "expected loss rate" and places it alongside the amounts, which is how an expected loss calculation is presented. What would move the line is declared: the credit rating of those countries, which the company identifies as the most relevant factor. A downward revision of those ratings, on a receivable book worth 37.1% of assets, is the single accounting event that can move the income statement most.
Income tax. The effective rate of 13.9% is reconciled from Malta's 5%, not from the rate of an operating country, because the group has applied fiscal consolidation in Malta since 2024. The two lines that explain almost all of it are the effect of entities taxed at different rates, negative by US$14.0 million, and other permanent differences, negative by US$13.7 million. The latter grew 46.6% in one year and the document describes it only as "non-deductible expenses, the effect of income taxes accrued in subsidiaries and non taxable income": it is the largest and least explained line in the reconciliation. On the other hand a line for foreign withholding tax appears for the first time at US$3.8 million, which the document qualifies as a permanent difference because it gives rise to no tax credit: it is the fiscal cost of moving dividends up from the subsidiaries, and it will grow with them. There is no valuation allowance on deferred tax assets, no unrecognised tax benefits, and the Pillar Two threshold does not yet apply because consolidated revenue remains below €750 million — a threshold that current growth will cross before long.
Borrowings and the cost that cannot be found. Financial liabilities are worth US$86.9 million, all very short-dated: daily borrowings and promissory notes in Argentine pesos indexed to the BADLAR and TAMAR rates, plus an overdraft in Chile. Financial assets of US$74.5 million are pledged as security, 75.2% of the securities portfolio. Note 22 states that the cost of these borrowings in 2025 was US$20.7 million and that it is "recognized in processing cost as merchant advances finance cost", that is inside cost of services. Note 11, however, presents an identical amount, US$20.7 million, on the "other finance expenses" line within finance costs, describing it as net foreign exchange effects and fair value adjustments. The two notes place the same figure on two different lines of the income statement; since the finance costs in Note 11 sum exactly to the total shown in the income statement, it is the description in Note 22 that finds no support. The difference does not change profit, but it shifts US$20.7 million between gross margin and finance results, that is between the two measures on which this business is judged.
Contingencies. The provision recorded is US$0.4 million against 119 proceedings, with civil claims assessed as probable at about US$0.1 million. The three class actions in the United States, which grew out of a short-seller report in 2023, carry no provision at all, and the document states that it is able neither to assess the likelihood of an adverse outcome nor to estimate a range of loss. The one in the New York state courts was dismissed at first instance in March 2025 and is on appeal, with argument scheduled for 25 March 2026; the federal one in the Eastern District of New York is held in abeyance pending international service on individual defendants, which the document states was effected in early March 2026. The absence of a declared range should not be read as absence of exposure: it should be read as the reader's inability to quantify it.
Intangible assets and capitalisation. Net carrying value is US$74.0 million, 4.8% of assets, growing US$10.6 million in the year. The point is not the balance but the flow: additions in the year are US$34.5 million, of which US$34.3 are capitalised salaries. The document states that it performed no impairment test because it identified no indicators, and discloses neither remaining useful life nor expected future amortisation. What would move the line is a revision of the judgement on the recoverability of the software, which by its nature depends on the continuity of the revenue that software generates.
Related parties. Two lines change scale in a year. Costs with collection entities that are related parties go from US$0.9 million to US$27.9, thirty times over, and receivables from those entities, together with advances, reach US$24.1 million against US$6.9. The document states that all transactions take place on normal market terms, but it does not identify the counterparties and does not explain the increase. On cost of services of US$690.8 million the amount weighs 4.0%; on a line that was immaterial a year earlier, the absence of explanation is itself the information.
What is really inside the large lines? Three items exceed 5% of assets or liabilities, and all three concern other people's money.
Cash (46.7% of assets). Of US$719.9 million, US$424.5 is corporate liquidity and US$295.4 are freely available funds belonging to merchants or their customers. The net debt figure in this report is computed on the corporate portion alone, as the nature of the second requires. Inside cash there are also US$53.7 million of money market funds measured at fair value, classified as level 1.
Trade and other receivables (38.8% of assets, including the non-current portion). Of US$572.0 million current, US$520.9 are net trade receivables due from acquirers, processors, merchants and collection entities, and US$51.1 are advances and other receivables due from processors and tax authorities: it is inside this second part that US$2.1 million were written off against a third-party processor. The document declares the whole book entirely current and classified as "normal" credit risk. It embeds US$13.9 million of balances offset against payables to merchants, permitted because a legal right of set-off exists.
Payables to merchants (83.7% of liabilities). The US$812.9 million are amounts owed to merchants for pay-in transactions already processed or for pay-outs awaiting their instruction. They fall due in full within six months. They are the counterpart of the cash and the receivables: it is this line that makes dLocal's balance sheet resemble a payment institution's more than a software company's.
Intangible assets (4.8% of assets). Below the threshold, but treated here because of the capitalisation policy. Internally generated software is worth US$43.6 million net on US$94.5 of gross cost, acquired intangibles US$30.3 on US$41.3. Accumulated amortisation covers 45.5% of gross cost, against 37.5% the prior year: the base is ageing, not getting younger, which means current amortisation is not compressed relative to future amortisation. On property, plant and equipment, worth US$4.0 million, the same ratio rises from 49.9% to 55.8%. Neither category is held under finance leases: the group's lease contracts cover offices and are presented separately as right-of-use assets of US$3.0 million under IFRS 16.
Deferred tax assets. Worth US$7.7 million, 0.5% of assets, of which US$3.1 are tax losses with a maximum expiry of five years. No valuation allowance is recorded against them.
Equity. The US$569.4 million are almost entirely retained earnings, US$534.8. Share premium fell from US$186.8 million to US$7.1 because the cancellation of treasury shares in August absorbed US$199.0 million of premium until it was exhausted, charging the remaining US$2.0 to retained earnings. Treasury shares are therefore nil. Non-controlling interests are worth US$0.2 million.
Working capital must be measured in days of processed volume, not of revenue, because receivables and payables are merchant funds and not turnover: on that basis, declared once and kept throughout, receivables are worth 4.66 days of TPV and payables 7.27, with a net balance of 2.61 days of volume held by the company. A year earlier they were 6.52, 8.03 and 1.51. Both cycles have shortened, but collections far more than payments, by 1.86 days against 0.76: it is this widening of the balance, from 1.51 to 2.61 days, that produces the US$188.0 million of cash from merchant working capital.
How much of the asset base is valued on an observable price and how much on an internal estimate? Financial assets at fair value are worth US$94.8 million, of which US$64.6 level 1 and US$30.3 level 2; the document states that it holds no level 3 assets at all in either year. Everything else is carried at amortised cost. On a balance sheet made of short-dated financial instruments this is the best possible configuration, and it should be said in the company's favour.
How much of the cash produced actually belongs to shareholders? dLocal's accounts turn on this question more than on any margin.
Operating cash of US$415.5 million breaks into three blocks. Pre-tax profit of US$228.7 million adjusted for non-cash items gives US$279.4: the largest adjustments are net exchange differences of US$18.6 million, share-based compensation of US$24.1 and amortisation of US$23.9, net of US$26.3 of interest income reclassified to investing and US$14.5 of fair value gains removed as non-cash. The second block is working capital, positive by US$163.1 million. The third is tax paid, US$27.0 million.
Other people's money. Of the US$163.1 million of working capital, US$188.0 are the net movement in merchant funds: the increase in payables to merchants of US$250.1 million, less the increase in net trade receivables of US$63.7, plus US$1.6 of other taxes. The company's own working capital, by difference, absorbs US$24.9 million. The share of third-party money in IFRS free cash flow is therefore 49.6%: of US$378.7 million of apparent free cash, US$188.0 are merchant funds. This is why the rest of the report uses the Adjusted Free Cash Flow of US$190.7 million.
Tax. Cash tax paid, US$27.0 million, sits 14.9% below tax on an accrual basis, US$31.8 million. The difference is explained by the net deferred tax benefit of US$0.8 million and by the timing gap between accrual and payment, visible in income tax payable falling from US$19.7 million to US$18.1. In 2024 the gap ran in the same direction but was far narrower, US$28.7 million paid against US$30.6 accrued, because that year included the accrual of the US$4.5 million Uruguayan tax settlement, paid in 2025.
Capex. The US$36.8 million of investment covers the entire capital spend: the document states it has no commitments to purchase property, plant and equipment, no new lease was entered into in either year, and there are no off-balance sheet arrangements. But US$34.3 of the US$36.8 million are capitalised salaries, that is personnel cost that does not pass through the income statement. That share is worth 29.8% of all the group's salaries and wages, against 23.4% the prior year. Had the company capitalised the same percentage of salaries in 2025 as in 2024, US$7.4 million more of cost would have flowed through the income statement, and operating profit would have been 3.4% lower.
More than three quarters of the growth in apparent free cash flow is explained by merchant working capital. On the correct measure, Adjusted Free Cash Flow goes from US$90.6 million to US$190.7, more than doubling, but it remains below the US$203.9 million of 2023, when revenue was 59.5% of today's.
The balance sheet is presented for two years only, so every calculation that requires invested capital rests on two points rather than a trend. That is a limit of the document and should be read as such.
Per-share values. Computed on each year's weighted average diluted share count.
| Per share (US$) | 2025 | 2024 | 2023 |
|---|---|---|---|
| Revenue | 3.624 | 2.445 | 2.147 |
| Net profit (diluted) | 0.65 | 0.39 | 0.49 |
| Adjusted Free Cash Flow | 0.632 | 0.297 | 0.673 |
| Equity | 1.887 | 1.603 | 1.502 |
Revenue per share grows 48.2% against 46.6% for revenue in absolute terms, because the weighted average diluted count fell 1.1%: the shareholder gained ground rather than losing it. The comparison that matters, however, is a different one: cash per share in 2025, US$0.632, remains below that of 2023, US$0.673, while revenue per share grew 68.8% over the same two years. Two years of top-line growth have not yet brought the shareholder back to the cash per share he started from.
Return on capital and on incremental capital. ROIC goes from 49.1% to 137.7%, but the incremental figure cannot be computed meaningfully: between the two years NOPAT grows by US$77.3 million while invested capital falls by US$90.9, because corporate liquidity rose more than equity. A negative denominator produces a number with no economic meaning, and that should be said rather than written down. The substance is that in 2025 dLocal produced more operating profit while employing less of its own capital, which is the signature of a model that lives on the float of other people's funds.
Reinvestment rate and implied growth. Reinvestment in the year is US$35.4 million: US$10.5 of capex in excess of amortisation, no acquisitions, and US$24.9 of the company's own working capital absorbed. That is 18.7% of NOPAT, and multiplied by ROIC it gives implied growth of 25.7%. Operating profit instead grew 56.5%. The difference, more than thirty points, does not come from invested capital: it comes from operating leverage on overhead and from the recovery of operating margin after a depressed 2024, and to a lesser degree from the tax rate. It is growth that cannot be repeated through the same mechanism, because operating expenses cannot fall below a certain share of revenue.
Self-funding. Over the three available years the company produced US$485.2 million of cumulative Adjusted Free Cash Flow and returned US$349.0 million through repurchases and dividends, raising externally only US$6.4 million from option exercises and US$102.2 net of very short-term debt in Argentina. dLocal funds its own growth: it does not depend on the availability of capital markets, and this is one of its most solid qualities.
Capital intensity. Cumulative capex over the three years is US$77.7 million against US$466.5 of cumulative net profit, 16.6%. The capex to amortisation ratio is 1.40 in 2025, 1.32 in 2024 and 1.49 in 2023: consistently above one, so capacity is being expanded, but on bases so small that the figure says more about salary capitalisation than about real investment.
Operating return on enterprise value. Operating profit over enterprise value gives 5.88%. It is the return that would accrue to a buyer of the whole company before capital structure and before tax surprises, and it is the only multiple that stays legible for a company whose effective rate moved more than six points in a year.
On research productivity the document permits one observation: technology and development spending over the three years is US$69.0 million through the income statement plus US$70.8 capitalised, against revenue growth of US$443.2 million over the same period. Half of this company's technology spending does not pass through the income statement.
Some of these comparisons do not apply to a payment processor and should be declared as such: there is no inventory, and trade payables in the industrial sense do not exist, because the "trade payables" line is money owed to merchants rather than credit obtained from a supply chain. Four comparisons remain, and three of them exceed the five-point threshold.
| Comparison | Growth A | Growth B | Gap |
|---|---|---|---|
| Net trade receivables against revenue | +13.9% | +46.6% | −32.7 pp |
| Loss allowance against gross receivables | +478.4% | +14.1% | +464.3 pp |
| Capitalised cost against expensed salaries | +72.9% | +24.1% | +48.8 pp |
| Revenue against operating cash flow | +46.6% | n.m. | not computable |
The first gap is favourable and should be read that way: receivables grow far less than revenue, a sign that the collection cycle has shortened, consistent with days of TPV falling from 6.52 to 4.66. There is no trace of sales made by loosening payment terms.
The second runs the other way but calls for care: the loss allowance grows almost sixfold while gross receivables grow 14.1%, which is the opposite of the classic signal of an allowance thinning out to borrow profit from the future. Here the allowance has been strengthened. What remains, as set out in 3-bis, is that even after that strengthening coverage is 0.16% against declared expected loss rates of up to 9% on some classes.
The third is the one to watch. Capitalised salaries grow 72.9% while those that stay in the income statement grow 24.1%: capitalisation running at three times the pace of expense shifts cost from the income statement into the future, and its share of total salaries has risen more than six points in a year.
The fourth comparison, between revenue growth and operating cash flow growth, is not usefully computable because operating cash in 2024 was negative by US$32.8 million and a ratio on a negative base means nothing. On the adjusted measure, Adjusted Free Cash Flow grows 110.5% against revenue at 46.6%: cash runs ahead of revenue, but only because the 2024 base was less than half the 2023 one.
The sector is cross-border payment services into emerging markets. Three standard metrics of this report are not meaningful and are not computed: inventory and its turnover, because there is none; the cash conversion cycle in its usual form, because receivables and payables represent client funds rather than the company's commercial cycle; and net debt to EBITDA as a measure of financial risk, because the company is in a net cash position and the debt it does carry is a daily peso facility used to advance funds to merchants, that is an operating instrument rather than leverage.
The metrics that matter are instead those of volume and float.
| Sector metric | 2025 | 2024 | 2023 |
|---|---|---|---|
| TPV (US$ bn) | 40.8 | 25.6 | 17.7 |
| Take rate (revenue / TPV) | 2.679% | 2.917% | 3.679% |
| Gross profit per unit of volume | 0.987% | 1.152% | 1.566% |
| Days of TPV in receivables | 4.66 | 6.52 | not available |
| Days of TPV in payables to merchants | 7.27 | 8.03 | not available |
| Net float in days of TPV | 2.61 | 1.51 | not available |
| Merchant funds held (US$ mn) | 295.4 | 236.1 | not available |
| Net revenue retention | 145% | 113% | not available |
| Share of top ten merchants | 61% | 62% | 60% |
| Merchants above 10% of revenue | 2 | 1 | 0 |
| Revenue from new merchants (US$ mn) | 12.7 | 13.1 | 21.9 |
Two readings impose themselves. The first is that the take rate falls while volume accelerates, and neither curve shows any sign of turning: the model grows by selling more volume at a lower unit price, and profit per unit of volume compresses faster than price. The second is that the contribution of new customers is fading in absolute terms, from US$21.9 million to US$13.1 to US$12.7 over three years, while growth concentrates on an ever narrower base.
The document does not provide several metrics that would be relevant for this sector and that other operators publish: the number of transactions, the split of TPV between pay-in and pay-out, the split between cross-border and local-to-local, approval rates, volumes by country. It defines pay-in, pay-out and local-to-local without ever quantifying them. This is a level of disclosure below what the model would warrant, and it should be recorded as such.
The risks that follow come from the document's risk factors and contingency notes, ordered by severity, distinguishing those already realised from those still potential.
Compression of unit price (realised). The take rate went from 3.679% to 2.679% in two years and gross profit per unit of volume from 1.566% to 0.987%. The document states the contractual cause: tiered pricing means customer growth reduces price. This is not a future risk but a dynamic that has been running for at least two years.
Merchant concentration (realised and worsening). The top ten are worth 61% of revenue and two of them individually exceed 10%, against one in 2024 and none in 2023. The document warns that losing one of these relationships, or a material reduction in volume, would damage the business. Contracts are open-ended and terminable without penalty once notice has lapsed.
Class actions in the United States (potential, unquantified). Three suits, arising from a short-seller report in 2023, no provision, and an explicit statement by the company that it cannot estimate a range of loss. The first was dismissed at first instance and is on appeal; the second is held in abeyance pending international service. The related legal costs appear among the "other non-recurring costs" excluded from Adjusted EBITDA in all three years.
Exposure to Argentina (realised in the accounts, potential on the regulatory side). Argentina is worth 14.7% of revenue and grows 88.5%. It is treated as a hyperinflationary economy, with a US$4.2 million charge in the income statement, and the group's operating funding is in Argentine pesos indexed to BADLAR and TAMAR. The document recalls that administrative and judicial inquiries into the local subsidiary were opened in 2023, that they seek no penalties at this stage, and that there were no developments in 2025.
Credit quality of receivables from processors and acquirers (potential, with a precedent already realised). Receivables are worth US$598.0 million including the non-current portion, against an allowance of US$0.9. The document states that it bears credit risk toward agents, processors and acquirers, and that it is not insured against credit losses. There is already a precedent: in 2025 US$2.1 million due from a third-party processor were written off and the residual carrying value of US$5.6 million was remeasured to fair value, with legal proceedings under way to recover it.
Currency and exchange control risk (realised). Net exposure across the top five currencies is US$34.0 million and a 10% move would produce US$3.4 million of income statement effect: small relative to profit. The real risk is not the measured one but the operational one: the model consists of converting and transferring currency in countries that impose controls, and net exchange differences adjusted in the cash flow statement are worth US$18.6 million.
Regulation and licensing (potential, permanent). The group operates in more than forty countries with dozens of local licences and three subsidiaries subject to stated minimum capital requirements, all met. The document devotes one of the longest blocks of its risk factors to this, warning that it may not be in full compliance with all applicable regulations on financial services, exchange controls and anti-money laundering.
The collapse of Egypt (realised, unexplained). Egyptian revenue falls 36.4%, from US$94.0 million to US$59.7, after more than doubling the year before. The document shows the figure in the regional table and comments on it nowhere. A market worth 5.5% of revenue that loses more than a third in a year without a line of explanation is a disclosure risk before it is an operating one.
Working capital as a source of cash (potential). The document itself lists among its risks that working capital needs may grow beyond its capacity to generate cash. The other side of that coin is that US$188.0 million of the cash produced in 2025 are merchant funds: a slowdown in volume, or a shortening of settlement times imposed by customers, would reverse that flow.
Losing a merchant to a customer that builds its own rails (potential). The document devotes a risk factor to merchants reducing or ending the relationship, and the contractual structure makes it immediate: open-ended agreements, no penalty, short notice. It is worth reading alongside a figure already known: the revenue brought by all new merchants in 2025 is US$12.7 million, less than an eighth of what each of the top ten customers brings on average.
What emerges is not a financially fragile company, because it has no debt and US$431.9 million of its own liquidity. It is a company exposed on two fronts the balance sheet does not protect: the price at which it sells, which falls by contract, and the number of counterparties able to make it fall faster.
The comparison is not possible. The document does not state which risk factors were added or removed relative to the prior year, and the previous year's 20-F was not supplied to this analysis. Reconstructing last year's list from memory would mean inventing it, and this report does not do that.
Two risk factors are nonetheless worth flagging for their content, without attributing to the document a claim of novelty it does not make. The first concerns artificial intelligence and contains the clarification "we do not use AI in our financial reporting processes": a sentence a company writes when it believes someone is asking. The second is titled "Our insurance policies are not sufficient to cover all claims" and describes a layered insurance programme including directors' and officers' liability and cyber cover: for a company with three open class actions, it is the factor to reread first next year.
Who controls dLocal, and with what rights? The answer lies in a dual-class share structure that the document describes precisely.
There are class A shares, listed on Nasdaq with one vote each, and class B shares with five votes each, converting automatically into class A on transfer. At year-end class A number 165.9 million and class B 129.1: 43.8% of the capital controls 79.6% of the votes. The two founders, Sergio Fogel and Andrés Bzurovski, each hold 37.8% of the class B, that is together roughly 60% of total voting power while owning roughly 34% of the capital. The document adds that class B holders have the right to maintain their proportional interest by purchasing additional class B shares if class A shares are issued. General Atlantic, the principal financial shareholder, holds 28.1% of the class A and 15.8% of the capital, and has its own director on the board, Luiz Ribeiro; on 3 September 2025 it sold 15,000,000 class A shares in a registered secondary offering in which the company was not a seller.
The board was reconstructed on 23 December 2025, when five directors — among them co-founder Sergio Fogel, who remains the largest holder of class B — stepped down and two independents joined. The document describes the outcome as a nine-person, majority-independent board with five independent members, and names nine; in the section on directors, however, it states that there are eleven members. Of the two statements, the nine-person board appears twice and is the only one supported by the list. The chairman is Andrés Bzurovski, co-founder; chief executive Pedro Arnt sits on the board; Sebastián Kanovich, formerly co-chief executive and today a defendant in the class actions, sits on the board and on two committees.
As a foreign private issuer, dLocal relies on five exemptions from Nasdaq's governance requirements, substituting Cayman Islands practice, and the document lists them: independent directors do not hold executive sessions; the compensation committee is not composed solely of independent directors; nor is the nominating committee; shareholder approval is not required for issuances of securities, including those serving incentive plans; and the requirements on disclosure of third-party director compensation and on distribution of annual and interim reports do not apply. On the compensation committee the document gives two versions: the recent developments section places it among the three committees established with the December 2025 board reconstruction, while the compensation section writes "Effective beginning in 2026, we have instituted a new Compensation Committee". On either reading, for all or nearly all of the year under review, the pay of executive directors did not pass through a dedicated committee.
On compensation the document states the limit explicitly: Cayman Islands law does not require individual disclosure and the company states it has not made it public elsewhere. Only the aggregate for directors and officers is known: US$22.5 million in 2025, US$19.8 in 2024 and US$9.0 in 2023. Of the US$22.5 million, US$17.1 is share-based.
Related party transactions are approved by the audit committee or the board under a stated policy. The auditor is Price Waterhouse & Co. S.R.L. of Buenos Aires, in office since 2020, with unqualified opinions on both the financial statements and internal control; total fees are US$2.5 million, of which US$0.4 for non-audit services, 14.9% of the total. In 2025 the board approved an additional US$125 thousand of audit fees for the 2024 financial statements, an overrun on the amount originally agreed.
Capital allocation. Over the three available years the company directed US$77.7 million to capex, nothing to acquisitions, US$199.0 to share repurchases and US$150.0 to dividends. The balance sheet shows no goodwill in either year, and no write-down of acquired assets appears in any of the three. The first extraordinary transaction of the period arrives in February 2026 and is worth US$23.0 million, settled almost entirely by offsetting a credit facility already extended to the seller: a structure that reduces the outlay but that also means the group had financed, for US$24.1 million, the party it then bought, and had done so while that party was a defendant in proceedings brought by a third party.
Share repurchases. The comparison that matters is between price paid and year-end price. In 2024 the company repurchased 11,583,705 shares for US$101.1 million, that is US$8.72 per share, under the US$200 million programme authorised in May of that year; in 2023 it had repurchased US$97.9 million under the previous programme, which it closed having exhausted it. In 2025 it repurchased none, despite having the remaining US$98.9 million of authorisation available until 31 May and despite closing the year with US$424.5 million of corporate liquidity. The average price paid in 2024, US$8.72, set against the US$14.14 at the close of 2025, says those repurchases proved well timed in hindsight; and it makes the decision to let the remainder lapse unused, replaced by a dividend, harder to follow. The new US$300 million programme was authorised on 13 March 2026, after year-end.
Net dilution. In 2025, 1.3 million shares were issued on option exercises and 8.2 million on warrant exercises, and none were cancelled against them: the August cancellation concerned treasury shares repurchased in prior years. Shares outstanding therefore rise from 285.5 million to 294.9, up 3.3%, while against share-based compensation of US$24.1 million the repurchase was nil. In 2025 the buyback neither returned capital nor plugged the dilution: there was none. The weighted average diluted count falls 1.1% only because it reflects the 2024 cancellations, not because the share base shrank.
Accounting aggressiveness. The company excludes ten items from Adjusted EBITDA. Two deserve to be summed over time. Share-based compensation is US$24.1 million in 2025 and US$59.8 across the three years: a real cost worth 11.0% of 2025 operating profit. "Other non-recurring costs" are US$0.1 million in 2025, US$1.6 in 2024 and US$1.7 in 2023, US$3.4 million in total, and the document states they relate to costs of addressing the short-seller allegations and the class actions "in 2023, 2024 and 2025". An item a company calls non-recurring that appears in all three comparative columns is recurring; the amount is small, however, and the point concerns method more than magnitude. It should be added that among the amortisation excluded from the adjusted measure there are US$20.8 million that the document places inside cost of services: Adjusted EBITDA is therefore not computed on the same basis as gross margin, and comparing the 25.4% adjusted margin with the 36.8% gross margin crosses that line twice.
Governance signals inside the document. Internal control over financial reporting is declared effective by management and audited without qualification, with no material weakness. There were no auditor changes and no restatements of prior years. The document discloses no pledges over directors' shares. Two structural elements already described remain: the absence of a compensation committee for all or nearly all of the year, and the fact that three of the four committees are not composed solely of independent directors.
On compensation and incentives the document allows nothing further. It does not state the metrics on which awards are built, nor the horizon, nor whether those metrics coincide with the items excluded from Adjusted EBITDA. All that is known is that directors and officers received US$17.1 million of share-based compensation out of a group total of US$24.1: 70.9% of all equity compensation went to that group of people, and their total pay is worth 11.5% of net profit.
In this document does management admit an error, a project closed, a target missed, a write-down?
It admits three negative facts, all of them accounting. The first is the write-off of US$2.1 million of receivables from a third-party processor, with the clarification that "The Company, together with its external legal counsel, continues to believe that it has good prospects of recovering the written-off balance": the admission comes with the reassurance in the same sentence. The second is the fair value remeasurement of the residual claim on the same processor, now carried at US$5.6 million. The third is the settlement of the Uruguayan tax matter, where the company states it chose to settle "despite having strong arguments for its position".
None of the three is accompanied by any reflection from management on why it happened. And on what a shareholder would actually want to hear — the compression of the take rate, the contraction in Egypt, the unused repurchase programme, the thirtyfold increase in related-party costs — the document is silent or confines itself to showing the figure. Management's commentary on 2025 is a list of percentage changes with their proximate cause, without a single sentence acknowledging a result below expectations.
One element runs the other way and should be acknowledged: the definition of Adjusted Free Cash Flow the company chose for its dividend policy removes merchant funds from the cash generated. It is a choice that reduces the base on which dividends are calculated and makes the measure more conservative than the free cash flow derivable from the statements. A management wanting to look more generous could have distributed on the larger number.
A write-down recorded in the accounts without a line of explanation in the commentary is not candour but bookkeeping. Here the write-downs are present, the explanation is not, and the real transparency is concentrated in a technical definition that almost nobody reads.
The significant accounting policies are standard IFRS applied without unusual options, and the document declares no change of policy, restatement or reclassification in the year. The treatment of Argentina under IAS 29, which produces a negative adjustment of US$4.2 million, is the only treatment outside the ordinary, and it is required by the standard.
The accounting choice with the largest impact is the principal role. dLocal presents revenue gross of the fees paid to local financial institutions, and the document lists six justifications, among them bearing credit risk toward processors and acquirers and having full margin risk on each transaction. The consequence is that revenue of US$1,093.6 million includes US$655.0 million of processing costs that pass through to third parties: the gross margin of 36.8% is the real economic figure, and the gross revenue overstates the size of the business. It is a legitimate and reasoned presentation, but anyone comparing this company's turnover with another's needs to know what is being compared.
The second choice that moves profit is salary capitalisation, already discussed: US$34.3 million in 2025, 29.8% of all salaries, against 23.4% the year before. The third is the placement of borrowing costs, where Notes 11 and 22 contradict each other over US$20.7 million.
The auditor identifies a single Critical Audit Matter, the audit of transaction revenues, treated in 3-bis. The opinion on internal control is unqualified.
On the distinction between IFRS and adjusted measures, the company's position is unusual and should be described precisely. Adjusted EBITDA and its margin are presented as IFRS measures, because the company qualifies them as the profit measure of its single segment under IFRS 8; Adjusted Free Cash Flow is instead openly declared non-IFRS. There are ten exclusions from Adjusted EBITDA, the most significant of which is share-based compensation.
On tax, the effective rate of 13.9% differs from the nominal rate because the group consolidates fiscally in Malta at 5%, and the reconciliation starts there. The point to note is that the document shows two different rates for the same year: 11.5% in Note 12 and 13.9% in management's commentary. Tax of US$31.8 million over pre-tax profit of US$228.7 gives 13.89%, so it is Note 12 that does not tie. For the other two years the two sources agree with each other and with the arithmetic.
The overall picture of earnings quality is good on cash and weak on editorial coherence. There is no deferred revenue to roll forward, no goodwill to test, no level 3 assets, liabilities are almost all short-dated, and profit converts to cash at 96.8% on the correct measure. Against that, the document contains at least seven discrepancies between its narrative and its statements that an attentive reader finds without difficulty: the two tax rates; TPV growth given as 53.4% when the volumes disclosed give 59.6%, and which instead coincides exactly with the growth in processing costs; net revenue retention given as 145% in one place and 144% in another; revenue from new merchants given as 2% and US$17 million against the 1.2% and US$12.7 million of the table; 2024 sales and marketing expenses cited as US$4.5 million instead of US$21.6; a board declared to have eleven members with nine named; and the date of the declared dividend, given as 13 March 2025 in Item 8 and 13 March 2026 in Note 32 for a payment due in June 2026.
To these can be added four more, two of which touch the statements: the balance sheet titles the entire US$99.1 million line "financial assets at fair value through profit or loss" while Note 15 classifies US$88.2 of it at amortised cost; the tax reconciliation in Note 12 gives US$32.595 million of current tax less US$0.841 of deferred benefit, that is US$31.754, against the US$31.752 shown in the income statement; the offsetting note applies two different amounts to 2024, US$7.8 million on the asset side and US$17.0 on the liability side; and the 2024 repurchase programme is dated 13 May by one note and 14 May by Item 16E.
None of these discrepancies changes profit, and the statements as a whole tie line by line. But a set of accounts is also an act of communication, and a dozen discrepancies between narrative and statements says something about the editorial control of a company with over a billion of revenue that has already faced three class actions over the quality of its public statements.
The new US$300 million repurchase programme, authorised on 13 March 2026 and expiring on 19 March 2027. It is worth 7.2% of market capitalisation at the year-end price, and corporate liquidity of US$424.5 million covers it in full. The previous programme, however, went unused for a full year, so authorisation is not execution.
The acquisition in Cameroon, completed on 27 February 2026 for US$23.0 million. The document states it serves to strengthen the group's position in cross-border payments and expand its presence in Africa, and that purchase price allocation is in progress with intangibles expected to be recognised for customer contracts, platform, intellectual property and licences. It is the company's first growth by acquisition in the period.
Residual operating leverage on overhead. Operating expenses fell from 19.9% to 16.1% of revenue in one year. If that compression continued even partially while volumes grow, operating profit would keep growing faster than revenue even with the take rate falling.
The United Kingdom licence obtained in December 2024, which the document calls "a significant advancement" of the group's regulatory framework. The document does not describe its commercial effects, and they cannot be quantified from here.
The dividend as declared policy. The company has committed to distributing 30% of the prior year's Adjusted Free Cash Flow. On the 2025 result that means US$57.2 million declared for 2026, and the mechanism ties distribution to a measure that excludes merchant funds.
None of these five depends on an assumption outside the document, and that is their merit; none of them, however, acts on the price per unit of volume, which is the variable the rest depends on.
The document contains no numerical margin targets, investment plans or business plan targets: as the group states in its trend information section, it flags no quantified expectation beyond 31 December 2025. Comparison is therefore possible only for commitments that appear in the comparative columns and the notes.
| Item | Stated | Delivered | Variance |
|---|---|---|---|
| US$200 million repurchase programme (May 2024) | authorised through 31 May 2025 | US$101.1 million repurchased in 2024, none in 2025 | US$98.9 million unused at expiry |
| US$100 million repurchase programme (December 2022) | up to US$100 million by July 2023 | concluded 30 June 2023 having repurchased US$100 million | none |
| Dividend policy (30% of prior year Adjusted FCF) | 30% of US$90.6 million for 2024 = US$27.2 million | US$150.0 million paid in 2025 | +US$122.8 million above policy |
| Dividend policy applied to 2025 | 30% of US$190.7 million = US$57.2 million | US$57.2 million declared on 13 March 2026 | none |
| Uruguayan tax matter | US$4.5 million accrued in December 2024 | paid in 2025, no further tax impact | none |
| Minimum capital requirements of licensed subsidiaries | three thresholds stated in euro, dollars and reais | declared met in both years | none |
| Margin, capex or volume targets | not present in the document | — | comparison not possible |
The dividend line deserves a reading. The stated policy would have produced US$27.2 million; US$150.0 million were distributed, 5.5 times as much. For 2026 the company returns exactly to the formula. 2025 was therefore a year of extraordinary distribution, and anyone using that dividend to estimate a forward yield would be wrong by roughly two thirds.
If three years from now this investment had produced a permanent loss of capital, what must have happened? Three mechanisms, all built on figures from the document.
First: a renegotiation by one of the two merchants above 10%. The mechanism is contractual and already described by the document: prices step down with volume, contracts are open-ended and terminable without penalty after notice, and two customers individually exceed 10% of revenue while the top ten are worth 61%. The figure that measures it today is that 61%, together with the US$12.7 million of revenue brought by every new merchant in 2025: if one of the two large ones renegotiates or leaves, the contribution of new customers cannot replace it. The observable signal that would anticipate it is the take rate: a decline faster than the 0.24 percentage points of 2025 would indicate that renegotiation is already under way.
Second: the return of credit risk on processors, on a book covered at 0.16%. The mechanism is declared by the company: dLocal bears credit risk toward agents, processors, acquirers and collection entities, is not insured against credit losses, and must still settle with the merchant if a processor fails. The 2025 precedent shows the mechanics in full: of US$8.3 million due from a third-party processor in contractual instalments, the first was collected in December 2024, the second and third were not, US$2.1 million were written off and US$6.2 were reduced to US$5.6 of fair value. The figure that measures the exposure today is the ratio of the US$0.9 million allowance to gross trade receivables of US$521.7, that is 0.16%, against the US$25.0 million that the Note 30 expected loss rates applied to the whole receivable book would produce. The observable signal is migration of receivables toward the worse rating classes: in 2025 the CCC class already holds US$112.8 million, where a year earlier it held US$13.0.
Third: the reversal of the float. The mechanism is that almost half the cash produced in 2025 is merchant money held for 2.61 days of volume, against 1.51 the prior year. That balance grows as long as volumes grow and as long as merchants accept those settlement times. If volumes stopped growing the contribution would fall to zero; if large customers obtained faster settlement it would turn negative, exactly as happened in 2024, when working capital absorbed US$167.2 million and IFRS operating cash was negative. The figure that measures it is that difference of 2.61 days; the observable signal is days of TPV in payables to merchants, today 7.27.
No probability is assigned to these scenarios, because the document provides no basis for estimating one.
What is left for the shareholder if growth stops?
| Item | Total (US$ mn) | Per share (US$) |
|---|---|---|
| Tangible equity | 495.5 | 1.680 |
| Net working capital | 454.8 | 1.542 |
| Current ratio | 1.47x | — |
| Corporate liquidity and financial assets | 523.6 | 1.775 |
| Corporate liquidity net of gross debt | 431.9 | 1.464 |
| Share of assets valued on observable prices or at amortised cost | 100% | — |
Asset protection exists but is thin relative to the price. Tangible equity per share is worth US$1.68 against a quotation of US$14.14: 11.9%. Net corporate liquidity per share is worth US$1.46, 10.4% of the price. A shareholder buying at that level would be paying almost nine tenths of the price for the stream of future earnings, and almost nothing for the assets.
If revenue stopped growing and margins returned to the average of the three available years, that is an operating margin of 22.2%, liquidity would protect survival but not price: the company would face no solvency problem, having no medium-term debt, but the residual value to the shareholder would remain tied almost entirely to its ability to keep generating profit. The item that would genuinely protect is liquidity; the one that would not is tangible equity, too small relative to the market capitalisation. The document provides no liquidation value and this report does not estimate one.
The numbers in this document describe a company that grows fast, consumes no capital, carries no debt and converts almost all of its profit into cash. They also describe a company whose selling price per unit of service has fallen by more than a quarter in two years, whose growth concentrates on a handful of ever larger customers, and whose generated cash is half other people's money. The two descriptions do not contradict each other: they are the same business read from two sides.
At US$14.14 the market capitalises the company at US$4,170.3 million, that is 21.2 times net profit and 21.9 times the cash actually available. The cash yield is 4.57% and the operating return on enterprise value 5.88%: not a price that embeds euphoria, but not one that offers protection either. That price requires volumes to keep growing enough to offset the fall in unit price, because that is exactly what happened in 2025: volume grew 59.6%, the take rate fell, and the result was still revenue growth of 46.6%.
The thesis holds under two conditions. The first is that take rate compression stays slower than volume growth: in 2025 volume grew thirteen percentage points faster than revenue, and that gap is precisely the price given away. The two curves run in opposite directions and the document offers no reason to expect a reversal, because tiered pricing is written into the contracts. The second is that operating leverage continues: in 2025 it was leverage, not gross margin, that produced the jump in operating profit, and it has a natural limit.
The thesis does not hold if one of the two merchants above 10% renegotiates, because the contribution of new customers is US$12.7 million a year and cannot replace it; or if volumes decelerate, because that would simultaneously stop revenue growth and the float's contribution to cash, as already happened in 2024. At the 31 December 2025 price a margin of safety does not emerge from the document's numbers: it emerges from the quality of the balance sheet, which protects against ruin but not against the price paid.
Metrics to monitor in the coming quarters:
The first two of these thresholds measure the same thing from two sides, and it is the thing that decides the thesis: whether dLocal is selling scale or giving it away.
Report based exclusively on DLocal Limited's Form 20-F for the fiscal year ended 31 December 2025, approved by the board of directors on 17 March 2026 and filed with the SEC on 18 March 2026. Multiples are computed on the closing price of US$14.14 on 31 December 2025 and the 294,931,956 shares outstanding at 31 December 2025, both supplied to the analysis as data external to the document. This report is for analytical purposes and does not constitute investment advice.
All amounts in US$ millions unless otherwise stated. Calculations are performed on the thousands disclosed by the document and rounded for presentation.
The full data table behind this report — every line item with its source in the filing and the arithmetic checks — is published separately: DLocal FY2025 data table.