DLO — DLocal Limited

Deep-value analysis — DLocal Limited (NASDAQ: DLO)

Date: 09/03/2026 · Reference price: … (yfinance, last session) · Market cap: … mn · Regime: REFRESH on top of the thesis of 08/04/2026

What I carry over and what I re-derive. This is a refresh under the REFRESH regime, triggered by a new filing — the 6-K of 08/13/2026 (Q2 2026 results, accession 0001846832-26-000031) — and by a falsifier that remained unverifiable in the mechanical triage. Carried over unchanged from the 08/04/2026 report: the description of the business model and the moat, the dual-class governance structure, the history of the two short-seller episodes (Muddy Waters November 2022, an anonymous report February 2025), the unanimous dismissal of the class action in April 2026, Pedro Arnt’s biography and mandate, the AZA/Mint Code acquisition in Cameroon, the FY2021-FY2025 annual financial series and the unqualified audit opinion from PricewaterhouseCoopers Uruguay dated 03/18/2026. Fully re-derived, because the new filing moved them: the FCF bridge, the cash position and net debt, gross margin and take rate, customer concentration, receivables quality, 2026 capital allocation and — as a consequence — all five valuation models and the Monte Carlo simulation.

Three errors from the previous report that I explicitly correct (they weren’t reading errors, but definitional ones, and all pull in the same direction): (1) “net cash ~…-900 mn” counted merchants’ money as belonging to the shareholder — actual corporate cash was … mn at 12/31/2025 and is … mn at 06/30/2026; (2) the DEEPIN marker from the previous run was given in thousands, not millions, and the Monte Carlo simulation came out with a median MOS of … — a 1,000x scale error that made the marker unusable; (3) the “forward P/E ~13.5x, PEG <0.5” pillar rests on a consensus EPS of ~… for 2026, incompatible with the … actually realized in the first half. All three are redone below, with figures from the filings.

Executive summary

The thesis, in one sentence: DLocal is a business growing volume spectacularly and cash not at all. The second quarter of 2026 delivered the company’s largest TPV in history — …, … year-over-year, the seventh consecutive quarter of growth above 50% — and, in the same quarter, a payment volume equal to all of 2023 (…). Revenue rose to … mn (… YoY), gross profit to a record … mn (…), net profit to … mn (…). And yet, the measure the company itself uses to calibrate its dividend — Adjusted Free Cash Flow — was … mn in the first half of 2026, versus … mn in the first half of 2025: down 5.6%, while revenue grew 55.4% and volume 83%. This isn’t a quarterly blip; it’s the third consecutive year in which the company’s structural cash generation fails to exceed the 2023 level (… mn), even though revenue has doubled since then (from … mn to … mn TTM).

Estimated value: a range of … – … per share, with the model median at … The five triangulated models give margins of safety from … (EPV Greenwald, the no-growth floor) to … (bull DCF), with a median of …. The Monte Carlo simulation across 20,000 scenarios, starting from my base-case assumptions, gives a median intrinsic value of … and a …% probability of undervaluation — practically a coin flip. The current price of … implies, by reverse construction, an owner-earnings growth of 26.2% in year one, decelerating to ~11.8%, at a 12% cost of capital. That’s exactly my bull-scenario assumption, not the base case. In other words: the market isn’t miscalculating, it’s betting the optimistic scenario is the likely one.

What broke from the old thesis. Of the seven falsifiers stated in August, two have fallen and one is within 1.8 percentage points of the threshold. Broken outright: “balance sheet practically debt-free, net cash ~…-900 mn” — of the … mn of total cash at 06/30/2026, only … mn is corporate cash; the remaining … mn is merchants’ money, held by the company but owed to them, and shows up on the other side of the balance sheet in … mn of trade payables. Additionally, on 08/12/2026 — one day before reporting, after the financial statements were approved by the Board — the company signed a … mn unsecured senior facility at Term SOFR …, maturing August 2029, repayable in 11 quarterly installments of … mn. Broken as an artifact: the “forward P/E 13.5x, PEG <0.5” pillar — with … diluted EPS realized in H1 and a pace of …/quarter, the 2026 fiscal year is heading toward ~… EPS, i.e. a forward P/E of 21.6x, not 13.6x. Near-broken: gross margin fell to 31.8% in Q2 2026 (from 38.6% in Q2 2025 and 35.3% in Q1 2026), less than two points from the 30% threshold at which the falsifier would trigger; the gross-profit/TPV ratio fell to 0.72% from 1.07% a year ago.

What holds, and holds solidly. Growth is real and accelerating, not decelerating: NRR of 153% (the fifth consecutive quarter above 140%), TPV Retention 188%, TPV guidance raised to 60… for 2026. Return on capital remains exceptional — ROE of 41.3% on equity of … mn. The 16.1% operating margin in Q2 is above the 15% threshold (though the half-year figure is 15.9%, i.e. right at the limit). And the business isn’t a scheme: the auditor is PwC, the opinion is unqualified, and the class action about exactly this topic — the take-rate trend — was unanimously dismissed by the New York Appellate Division in April 2026.

Verdict: WATCH, with a lowered entry threshold. The company is real, the moat is debatable but not nonexistent, and unit economics are good. The problem isn’t the quality of the business, it’s the price paid for a cash conversion that hasn’t materialized in three years. At … you’re buying the bull scenario. A risk-reward ratio deserving capital in a concentrated portfolio appears somewhere below …-13.00 — i.e. at the level where the company itself bought back its own shares in the first half of 2026, at an average price of … — where the base case turns positive and the EPV floor is no longer at …. The signal that changes everything, in either direction, is the 11/11/2026 reporting: if H2 2026 Adjusted FCF confirms the operating leverage promised for the second half, the bull thesis becomes the base case; if H2 repeats H1’s pattern (revenue …, cash …), then what you have isn’t a scale-effect platform, it’s an intermediary buying its own volume.