World coffee production is forecast at a record 189.7 million bags and the composite price has fallen 25 percent in seven months. Inside that record, robusta is rising too — Brazil down six hundred thousand bags, Indonesia down a million. One country is covering for both.

Here is what a roaster is about to hear from everyone. Record crop. Record exports. Prices off a quarter since New Year. Buy later, buy cheaper, stop worrying.

Most of that is true, and I am not going to pretend otherwise. U S D A’s July circular has world production up 10.8 million bags to a record 189.7 million, world bean exports up 11.9 million to a record 131.4 million, and this sentence, which is the one everybody will quote: coffee prices “dropped 25 percent in the last 7 months as additional supplies became available.”

The reason is Brazil, and it is a real reason. Arabica there is forecast to rebound 9.5 million bags to 47.5 million, ending five years of underperformance — timely rain during flowering in September and October, near-record yields in Minas Gerais, where over seventy percent of Brazil’s arabica grows.

Then read the next sentence in the same paragraph, which almost nobody will.

Brazil’s record is an arabica record. Its robusta went the other way.

And it is not alone. Indonesia is forecast down a full million bags to 11.4 million — and the circular is specific that essentially all of that is robusta, off 1.0 million to 10.0 million, on excessive rainfall that disrupted flowering and cherry formation in the lowlands of southern Sumatra and Java where about seventy-five percent of it is grown. Indonesian arabica is flat.

So the two largest robusta origins after Vietnam are both down, for unrelated weather reasons, in the same forecast year. Here is who is absorbing that.

Vietnam’s gain does not cover the other two, and the circular is candid about why the gain exists at all: “Recent high prices allowed coffee growers to increase expenditures on fertilizers and other inputs to boost yields.” That is a yield bought with last year’s price. It is not a structural expansion, and it is the sort of thing that reverses when the price that paid for it falls twenty-five percent.

I want to be careful here, because I am reading a supply forecast and forecasts are not prices. So I went and looked at what the market actually did yesterday, which is a different kind of evidence and available to anyone.

The I C O publishes a daily composite indicator and four group indicators. This is August 3, and it is a single day — I am showing it because of the shape, not the size.

Every arabica group down. Robusta up. On the day, in the same direction the supply forecast points.

One day proves nothing on its own and I would not build a buying decision on it. What it does is stop you dismissing the forecast as a paper exercise. The tape and the circular are not arguing.

So what does a roaster actually do with this.

Kept honest, and there is a lot to keep honest in a commodity piece. These are forecasts. The 2026/27 numbers are U S D A projections approved by the World Agricultural Outlook Board in July, not observed harvests, and the circular itself shows how badly Brazilian arabica has surprised in both directions since 2021/22. One day is one day. The August 3 indicator moves are a single session; I am using them for direction, not magnitude, and a day-on-day change of under two percent is well inside normal noise. The I C O groups are not a clean arabica-robusta split — three of the four are arabica types and one is robusta, so “three down, one up” is three arabica groups, not a weighted market. “Prices dropped 25 percent” is U S D A’s characterisation of the I C O monthly composite over seven months, and I did not independently recompute it. I have not modelled differentials, which is where a roaster actually buys — indicator prices are not what lands in your contract, and origin differentials can move against the index. Nothing here is a price forecast and I am not making one; robusta supply being tighter than arabica supply is a statement about bags, and bags are one input into a price among several. Finally, I did not verify Vietnamese conditions independently — the fertiliser explanation is U S D A’s, reported as theirs.

Sources, both primary, both fetched this morning, August 4, 2026. (1) United States Department of Agriculture, Foreign Agricultural Service, Coffee: World Markets and Trade, July 2026, approved by the World Agricultural Outlook Board / U S D A — retrieved with a browser user-agent (per house rules for apps.fas.usda.gov) and extracted with pypdf. Source of: Brazil’s 2026/27 combined harvest forecast of a record 71.9 million bags, up 8.9 million; arabica rebounding 9.5 million to 47.5 million after a five-year period of underperformance; the Minas Gerais flowering and yield detail and the “over 70 percent” arabica concentration; the robusta passage quoted verbatim above, including the 600,000-bag retreat to 24.4 million and the Espirito Santo detail; Brazil bean exports up 11.0 million to a record 45.0 million; world production up 10.8 million to a record 189.7 million; world bean exports up 11.9 million to a record 131.4 million; global consumption at a record 179.7 million; ending stocks rising a second consecutive year to 26.3 million while remaining “below the long-term average”; the 15.5 million bag decline in global ending stocks between 2020/21 and 2024/25; the statement that coffee prices “dropped 25 percent in the last 7 months as additional supplies became available (as measured by the International Coffee Organization (ICO) monthly composite price index)”; Vietnam up 800,000 to a record 32.5 million with the quoted fertiliser-expenditure explanation; Indonesia down 1.0 million to 11.4 million with robusta off 1.0 million to 10.0 million on southern Sumatra and Java flowering disruption and arabica flat at 1.4 million; Colombia up 900,000 to 13.4 million; Ethiopia up over 500,000 to a record 12.1 million on varietal replacement across over half its cultivated area; and Central America and Mexico up nearly 600,000 to 17.7 million with Honduras rebounding 500,000 to 6.0 million. (2) International Coffee Organization, ICO Indicator Prices — August 2026 (I-CIP), the daily indicator workbook, fetched this morning. Source of the August 3 closes and day-on-day changes quoted above in US cents per pound: I-CIP 283.11 (−0.8%), Colombian Milds 374.87 (−1.7%), Other Milds 350.86 (−1.8%), Brazilian Naturals 314.34 (−1.4%), Robustas 186.11 (+1.7%). August 3 is the only session posted for the month so far. No trade press supplied any fact in this edition.

Every roaster is about to be told the same thing. Record crop, record exports, prices off 25% since New Year. Relax.

Most of it is true. USDA's July circular has world production up 10.8 million bags to a record 189.7 million, world bean exports up 11.9 million to a record 131.4 million, and the line everyone will quote: prices "dropped 25 percent in the last 7 months as additional supplies became available."

Brazil is the reason, and it's a real one. Arabica there rebounds 9.5 million bags to 47.5 million, ending five years of underperformance. Timely rain during flowering, near-record yields in Minas Gerais.

Now read the next sentence in that same paragraph, which nobody will:

"Robusta production is forecast to retreat 600,000 bags to 24.4 million following last year's record harvest. Cooler temperatures combined with periods of elevated rainfall to lower yields in Espirito Santo, where approximately 70 percent of Robusta coffee is grown."

Brazil's record is an arabica record. Its robusta went the other way.

And Indonesia is down a full million bags — essentially all of it robusta, 10.0 million now, on rainfall that disrupted flowering in southern Sumatra and Java. Indonesian arabica is flat.

So the two biggest robusta origins after Vietnam are both down, for unrelated weather, in the same year.

Who's absorbing it? Vietnam, up 800,000 to a record 32.5 million. Alone. And it doesn't cover the other two.

USDA is candid about why that gain exists: "Recent high prices allowed coffee growers to increase expenditures on fertilizers and other inputs to boost yields."

That's a yield bought with last year's price. Not a structural expansion. Exactly the kind of thing that reverses after the price that funded it falls a quarter.

Forecasts aren't prices, so I went and looked at what actually traded. ICO daily indicators, August 3:

Colombian Milds −1.7%
Other Milds −1.8%
Brazilian Naturals −1.4%
Robustas +1.7%

Every arabica group down. Robusta up. Same direction the forecast points.

One session proves nothing and I wouldn't buy on it. What it does is stop you filing the circular under paper exercise. The tape and the forecast aren't arguing.

Also worth seeing what the composite did with that: the I-CIP printed −0.8%. It averaged a rise into three falls and reported the fall. If you're tracking the headline index, that split is invisible to you.

What I'd actually do:

If you run robusta anywhere — espresso base, a percentage in the house blend, the value tier that lets your single origins be expensive — that's the line not participating in the relief. Model arabica down and robusta flat-to-firm rather than applying one number across the book.

The good news is real and it's arabica-shaped. Colombia +900,000. Ethiopia to a record 12.1 million after replacing over half its cultivated area with higher-yielding varieties. Central America and Mexico up nearly 600,000. If your book is mild-heavy, this is the loosest market in years and you should use it.

The thing keeping me from calling it a glut: ending stocks rise to 26.3 million bags and USDA still says they're "expected to remain below the long-term average." Global stocks fell 15.5 million bags between 2020/21 and 2024/25. Two good years haven't refilled that. A record crop into a thin cupboard behaves differently than one into a full cupboard.

And the question for your importer this month isn't where prices are going. It's: what share of my contracted volume is robusta, which origin, and is it Vietnam. If it's Vietnam for effectively all of it, you're single-sourced on the only robusta origin currently growing.

Kept honest: these are July forecasts, not harvests, and Brazilian arabica has surprised hard in both directions since 2021/22. One day of indicator data is direction, not magnitude, and sub-2% daily moves are ordinary noise. Three of the four ICO groups are arabica types, so "three down one up" is three arabica groups, not a weighted market. The 25% figure is USDA's characterisation of the ICO monthly composite; I didn't recompute it. I haven't modelled differentials, which is where you actually buy. And none of this is a price forecast — tighter robusta supply is a statement about bags.

FIRST SOURCE · one verified original-source finding, composed for one reader · this edition: coffee roasters — written for the people who have to price a blend, not an index