Everyone watches South American production. The acreage, the rain, the CONAB and WASDE revisions — that machinery is well covered. What gets far less attention is what happens to that grain before it reaches a port. Argentina crushes over 40 million tonnes of soybeans a year. Brazil now diverts roughly a sixth of its corn crop into ethanol. Both are growing faster than the crops that feed them, and both quietly subtract from the exportable surplus the market actually trades. This is what we found trying to measure them — including the series we had to derive ourselves, and the one number we still cannot source.

Two claims on the same continent's grain The Setup

The two stories are not the same story, which is worth saying up front because it is tempting to bundle them as "South America." Argentina's crush is a soybean question: it is the world's largest exporter of soybean meal and oil, so Argentine crush is the meal supply that decides how hard a soy-oil fuel bid can pull crush anywhere else. Brazil's corn ethanol is a corn question: it is domestic demand that converts a record crop into something less than a record export program. Different crops, different mechanics, read by different questions.

And Argentina's half of it is not running flat out. USDA's Buenos Aires post estimates installed crushing capacity at about 67 MMT against a crush near 42 — roughly 60% utilisation, with "abundant idle capacity" in its own words. Argentina already crushes about 85% of its own crop and imports beans from Paraguay on top of that, so what limits the crush is bean availability, margin, and whether the farmer will price — not plant. It is worth holding that in mind through the pace numbers below: a shortfall in Argentina is a selling story, and the equipment to crush a great deal more is already standing.

Argentina: the marketing year we had to find The Method

Argentina's agriculture secretariat (MAGyP) publishes monthly oilseed crush back to 1965 — grain crushed, oil produced, meal produced, by oilseed, in tonnes. It is a government series, published as one spreadsheet, and it runs about a month behind. That makes it more current than the US crush series, which lands roughly two months after the month it describes.

We wanted to measure it the way we measure US crush: against the annual figure WASDE forecasts. But that requires knowing which twelve months WASDE means by Argentina's marketing year. Rather than assume, we tested the candidates — sum MAGyP's months across each window and see which one reproduces WASDE's number.

October–September, to six hundredths of a percent, across eighteen marketing years. That is not a good correlation — it is an identity. It is also not the window you would guess from USDA's own country tables, which state Argentina's local soybean marketing year as beginning in April; the world supply-and-use table those crush numbers live in is aggregated on a different year. Testing the candidates rather than reading the label off the country table is what found it. WASDE's Argentine crush number is MAGyP's monthly series summed. We had already found the same thing in the United States, where NASS crushings reconcile to the WASDE crush line to within rounding on every complete year. In both countries the monthly series is not a proxy for the annual forecast; it is the accumulating numerator of it.

Which means the pace question is answerable. Ten months into a marketing year you can ask whether the industry is actually pressing at the rate the forecast requires, and what its own run-rate implies the year will finish at.

The part where the two countries stop being comparable The Caveat

Here is the finding we think matters most, and it is a limitation rather than a signal. We back-tested the run-rate estimate in both countries — at each month of the year, how far off was the implied full-year number? Averaged across history, that error is the noise floor. A deviation smaller than it means nothing.

Argentina's band runs roughly four times the American one. Argentine crush swings with drought and with export-tax-driven farmer selling in a way US crush simply does not — the 2022/23 drought cut it from 38.8 to 30.3 million tonnes and back. So the same-sized deviation carries considerably less information there, and early in a marketing year the Argentine read is close to uninformative. Two numbers that look alike on a page are not equally trustworthy, and we would rather publish the band than let the resemblance do the talking.

As of the July data, both are boring in the same way: the US running −0.37% against a ±0.52% band, Argentina +0.53% against ±2.16%. Both on pace. Both nulls. That is a perfectly good answer, and it is the answer most of the time.

Brazil: the series that had to be derived The Gap

Brazilian corn ethanol is the fastest-growing claim on any grain crop we track, and for most of its history it is not published as open structured data. Brazil's petroleum regulator, ANP, once published exactly the right dataset — feedstock tonnes into ethanol, monthly, by state, with corn as its own category. Its own data inventory now marks that panel descontinuado. CONAB reports the corn figure, but only for the current and coming season, in the spreadsheet attached to each cane bulletin.

So for the history we derived it, as the difference between two things that are published:

A residual is a nervous kind of number: it inherits both sources' errors and shows you none of them. So the question is not whether it looks plausible but whether it can be falsified. The strongest test we found was not the headline at all — it was the states where the answer must be zero. São Paulo is the heart of the cane industry and makes no corn ethanol. If our residual is really measuring corn, São Paulo should come out at nothing, on a very large number.

Against CONAB's published figures the derived national number lands within about 2% for the two most recent closed seasons. It also has a known bias we will state plainly: for 2025/26 it reads 7.4% light, and the entire gap sits in São Paulo and Maranhão — two states CONAB says make no corn ethanol at all, where CONAB's cane estimate and ANP's recorded output disagree with each other. In the states that actually make the stuff, the residual is within a few percent. So it is sound for the national growth arc and should not be quoted state by state where CONAB publishes. Where CONAB does publish, we use CONAB and label it as such.

What the number turns out to be The Finding

Brazil's corn ethanol reached 10.2 billion litres in 2025/26. At roughly 410 litres per tonne that absorbs about 24.8 million tonnes of corn — near enough 17% of the 143 million tonne crop CONAB is forecasting. CONAB's own forecast for 2026/27 is 11.9 billion litres, another 17% higher.

The consequence is simple to state and easy to miss: a record Brazilian corn crop and a record Brazilian export program are no longer the same statement. The grind sits between them and is growing faster than the crop. For anyone modelling global corn balances off Brazilian production, that wedge is now large enough to matter on its own.

What we still can't answer Open

Three things we went looking for and did not find, which we would rather publish than quietly drop.

• Farmer selling paceStill open
  The share of the Argentine crop sold with a firm price is one of the most interesting numbers in that market — unpriced farmer length is supply overhang that has not hit the board yet. USDA’s own reporting makes it the binding variable, and shows how violently it moves: farmer sales ran 52% above the prior year in late 2025, the strongest in four months since MY2018/19, then inverted to only 22% of the crop priced by mid-2026, the lowest share since at least MY1994/95, with producers holding beans as an inflation hedge. Same country, same crop, opposite behaviour inside eight months. We still cannot source it as a series: it comes from the Rosario exchange, not a statistical agency, and we have not found it in Argentine government open data.
• Argentine idle capacityPartly answered
  We drafted this section without a capacity number. USDA’s Buenos Aires post supplies one: an estimated 67 MMT of installed crushing capacity against a crush near 42, which it calls “only about 60 percent of utilization” with “abundant idle capacity.” That reframes the Argentine read — a shortfall there is a selling story, not a plant story, and the capacity to crush far more already exists. Two caveats keep it an estimate rather than a series: it is attaché research from industry contacts rather than a survey, and the same report concedes that “no numbers exist for Argentina’s sunflower crush capacity.” We quote the order of magnitude, not a decimal place.
• Where the corn grind bindsStill open
  Brazilian corn ethanol has grown every year we can measure. Nothing in the data tells us where it stops — whether it is bounded by corn availability in Mato Grosso, by domestic fuel demand, or by nothing yet. The forecast says another 17%. It does not say what happens after.

Reading it

• Crush pace
  Ahead of the required pace means demand is tightening the balance sheet faster than the forecast assumes; behind means the crush line carries air. Check it against the band before believing it.
• Meal, not oil
  Argentine crush matters to the soy complex mainly through meal. A loud oil bid with weak meal clearing is throttled crush.
• Brazilian corn
  Subtract the grind before treating a production number as an export number.
• Labels
  Where a figure is derived rather than reported, we say so on the chart itself.