Arcos Dorados Posted A 42.9% Comp In South America. Its CFO Put The Region's Inflation At 46-47%.

by Henry Nakamura · August 16, 2026

Earnings Reporter, Earnings Dispatch

Arcos Dorados Holdings (NYSE: ARCO), the master franchisee that operates McDonald's across Latin America and the Caribbean, released second-quarter 2026 results before the open on Thursday, August 13. The headline the coverage carried was growth: total revenues of $1,305.6 million, up 14.3% in US dollars, and systemwide comparable sales up 15.3%. Benzinga's report of the release breaks that comparable-sales figure into divisions - Brazil +5.4%, NOLAD (Mexico and Central America) -2.2%, and SLAD (the South Latin America division, whose largest market is Argentina) +42.9%. The market liked it: the shares gapped up on Thursday, and Benzinga had them 3.67% higher at $8.19 during the session.

Later that morning, on the earnings call, Jeronimo Guzman of INCA asked chief financial officer Mariano Tannenbaum a multi-part question that began with what inflation had actually been in SLAD. Tannenbaum answered: "Actually, the weighted average inflation in SLAD was around 46%-47%. Our sales were slightly below inflation. I would say in line with inflation." He then moved on to divisional margins, saying SLAD's adjusted EBITDA "grew in the quarter around $3 million or 6.6%" with the margin "stable at around 10%." We checked that passage against two independent transcript renderings - MarketBeat's and Investing.com's - and they carry it identically, word for word, including the closing qualifier. Both are third-party transcripts, not a filed document.

That exchange is the most consequential thing in the quarter, and it is not in the headline. Our arithmetic, not the company's: dividing 1.429 by 1.465 - the midpoint of the 46-47% range Tannenbaum gave - implies SLAD comparable sales were roughly 2.5% lower in inflation-adjusted terms, within a band of about -2.1% to -2.8% depending on which end of his range you use. Treat that as an order of magnitude rather than a measurement. SLAD spans a dozen markets with very different inflation rates, the comp is a sales-weighted figure while the CFO's number is his own weighted average, and neither is struck at the same moment within the quarter. It is also, notably, the same conclusion Tannenbaum volunteered unprompted: sales slightly below inflation, which he characterised as in line with it. Arcos Dorados does not publish an inflation-adjusted comparable-sales figure, and nothing here suggests it has represented the metric as anything other than what its footnote says it is.

That footnote is worth reading in full, because it is the whole reason the headline number cannot answer the question. In its first-quarter 2026 release - a 2026-period company document, so the definition is current - Arcos Dorados defines the measure this way: "this non-GAAP measure, refers to the change, on a constant currency basis, in Company-operated and sub-franchised restaurant sales in one period from a comparable period for restaurants that have been open for thirteen months or longer (year-over-year basis) including those temporarily closed." The wording is identical in the second-quarter 2025 release, so the definition has not shifted between the periods being compared. Two things follow. First, the company itself labels it non-GAAP. Second, "constant currency" here means holding the exchange rate fixed. It does not mean constant purchasing power. In a market running 46% inflation, a constant-currency comp is a nominal local-currency number, and menu-price increases carry it upward whether or not a single additional customer walks in.

The consolidated 15.3% has the same property, and the division split shows how much work SLAD is doing. Brazil's comp was +5.4% and NOLAD's was -2.2%. A consolidated figure of 15.3% therefore sits above every division except SLAD. What reads as company-wide momentum is, mechanically, Argentine and neighbouring price inflation weighted into a group average - the same shape as a year earlier, when a +12.1% systemwide comp sat on top of Brazil +0.3%, NOLAD +4.4% and SLAD +38.2%.

The dollar revenue line tells the opposite story about the same geography, and the two lines reconcile. Benzinga reports SLAD revenue up 7.0% and NOLAD revenue up 9.3%, with Brazil revenue of $520.6 million, up 25.3%. The second-quarter 2025 release - which we were able to open - gives the divisional bases: Brazil $415.4 million, NOLAD $317.8 million, SLAD $409.1 million, summing to the $1,142.3 million reported total. Our arithmetic: applying those growth rates gives about $437.7 million for SLAD and about $347.4 million for NOLAD, and $520.6m plus $437.7m plus $347.4m is $1,305.7 million against the $1,305.6 million reported. The segment percentages and the group total are internally consistent, which is one reason to believe the divisional figures in the coverage were read off the release rather than paraphrased from a summary.

So SLAD delivered a 42.9% constant-currency comparable-sales figure and 7.0% dollar revenue growth in the same quarter. Most of that gap is currency translation. The two are not struck on the same base - comparable sales measure systemwide restaurant sales, including sub-franchised units, at restaurants open thirteen months or more, while revenue is company-operated sales plus franchise fees across the whole estate - so the difference is not purely an exchange-rate effect. But nothing else disclosed in the quarter is large enough to account for a 36-point spread.

Brazil is the mirror image. A 5.4% comp turned into 25.3% dollar revenue growth, a gap running the other way, which is what a stronger real does to translated results. Set that against the year-ago quarter: in its second-quarter 2025 release Arcos Dorados reported total revenues up 2.8% in US dollars and 14.9% in constant currency - a translation drag of twelve points on the group. A year later the same mechanism is running in Brazil's favour. We did not obtain the second-quarter 2026 release's own consolidated constant-currency revenue figure and are not estimating one.

A note on the reporting framework, because it changes what these documents are. Arcos Dorados is a foreign private issuer incorporated in the British Virgin Islands. It files 6-Ks, not 10-Qs, and it does not file quarterly financial statements on the US domestic schedule. But it is not an IFRS filer: its fiscal 2025 annual report on Form 20-F, which we read on EDGAR, states that "We prepare our consolidated financial statements in accordance with accounting principles and standards generally accepted in the United States, or U.S. GAAP." The same 20-F documents Argentine peso depreciation against the dollar of 40.8% in 2025, 27.5% in 2024 and 357.4% in 2023 - the backdrop against which a 40-plus percent nominal comp should be read.

The margin story is real and separable from the inflation arithmetic. Adjusted EBITDA was $126.8 million against $110.1 million in the year-ago quarter as reported in the second-quarter 2025 release - growth of 15.2%, matching Benzinga's figure and the 9.7% consolidated margin it reports, up about 10 basis points. Two aggregator write-ups of the call put the growth at "about 20%"; that does not reconcile with either the prior-year base or the margin, and we have not used it. Brazil carried the improvement, at $75.8 million of adjusted EBITDA on $520.6 million of revenue. Tannenbaum said Brazil's adjusted EBITDA margin "expanded 180 basis points to 14.6%" and attributed it to lower food and paper costs, sales growth above inflation, operating leverage, lower G&A following a restructuring completed late last year, and a stronger Brazilian real. That reproduces: our arithmetic puts $75.8m over $520.6m at 14.56%, the year-ago $52.9m over $415.4m at 12.74%, a difference of 182 basis points.

NOLAD went the other way on all three lines: comparable sales -2.2%, dollar revenue +9.3%, adjusted EBITDA -18.3%. Revenue up in dollars while comps fall is again translation, this time out of a firmer Mexican peso. Tannenbaum explained the profit decline in operating terms: NOLAD's margin fell 110 basis points excluding income from a restaurant transaction in the prior-year quarter, he said, "primarily because lower operating leverage outweighed improved food and paper costs."

One arithmetic point worth naming, because it is easy to misread as a hole. The divisional adjusted EBITDA figures do not sum to the consolidated total - and the company itself supplies the bridge. Its releases carry an explicit "Corporate and Other" line: -$24.6 million in the second quarter of 2025 and -$23.0 million in the first quarter of 2026. Our arithmetic, applying the reported divisional growth rates to the year-ago bases, gives Brazil $75.8m, SLAD about $43.2m and NOLAD about $33.7m, or roughly $152.7 million against a reported consolidated $126.8 million, implying a Corporate and Other line of about -$26 million. That is in line with the two figures the company has actually published, and the SLAD estimate is corroborated by Tannenbaum's own remark that SLAD adjusted EBITDA grew about $3 million, or 6.6%, at a margin near 10%. It is not an unexplained gap; it is a segment the company reports, and it sits outside every divisional margin quoted above.

The earnings-per-share line needs a label the coverage did not consistently supply. Net income was $45.0 million, against $22.6 million a year earlier - up 99%. The figure that ran in the headlines was $0.22, described by Benzinga, Investing.com and Zacks as beating a $0.15 consensus, with Zacks explicitly noting it is "adjusted for non-recurring items" and Investing.com calling it adjusted EPS. It is not the GAAP basic number. Our arithmetic: the fiscal 2025 Form 20-F reports 130,663,057 Class A and 80,000,000 Class B shares outstanding, about 210.7 million in total; $45.0 million over that is about $0.21. The year-ago quarter, on the same basis, reported basic and diluted EPS of $0.11 on $22.6 million - so GAAP EPS roughly doubled while the reported adjusted figure exactly doubled. The distinction matters because the beat that ran in the headlines was a non-GAAP number measured against a non-GAAP consensus. We could not read the second-quarter 2026 release's own EPS table to confirm the reported basic and diluted lines, and we are not asserting them.

The consensus figures themselves disagree, which is its own caution. Benzinga describes revenue of $1.306 billion as a beat against a $1.290 billion estimate; MarketBeat's panel puts the estimate near $1.28 billion, and Zacks scores the same revenue as a 1.99% beat. Quiver Quantitative describes the identical $1,305,630,000 as a $10 million miss against a $1,315,597,530 estimate - and its EPS estimate is $0.14, not the $0.15 the others use. Each provider looks internally coherent; they are simply different estimate panels. Take one provider's EPS and revenue lines together, or neither. No single provider's numbers should be mixed with another's.

What we could not read: the second-quarter 2026 6-K exhibit itself. We searched EDGAR and found no 2026 second-quarter 6-K in the index; the company's investor-relations quarterly-results and financial-news pages both returned "no files for the selected year" when we checked on Sunday, August 16; and the release has not surfaced on the wire service that carried the company's first-quarter 2026 and fourth-quarter 2025 results. Every 2026 quarterly figure in this piece is therefore attributed to a named secondary account of the August 13 release, while the metric definition, the divisional bases, the Corporate and Other line and the accounting framework come from company documents we did open - the first-quarter 2026 release, the second-quarter 2025 release and the fiscal 2025 Form 20-F. The divisional comparable-sales splits, including the 42.9%, rest on Benzinga alone. We are willing to run them for three reasons: the segment revenue figures reconcile to the group total within $0.1 million; Benzinga's divisional adjusted EBITDA growth rates reproduce exactly against the divisional bases in the filed 2025 release; and its SLAD figure of +6.6% is repeated verbatim by the CFO on the call. Readers should still treat the divisional comps as single-sourced.

The next disclosure that could settle the real-versus-nominal question is the third quarter. The question to carry into it is narrow and answerable, and it is not an accusation: does SLAD's comparable-sales figure clear the division's own inflation rate, or does the growth in that division continue to be, as its chief financial officer put it, in line with prices rather than ahead of them?