The cut to next season's projected global surplus in six months — 267,000 tonnes in January to 25,000 in July. Two revisions, one weather pattern, and a market that spent the first half of the year positioned for abundance.
Well supplied today. Not next season.
The present crop is abundant and the exchanges are full; the forecast for the crop now forming has been cut to near balance under a strong El Niño. The five readings below are live; the sentence beneath them is written from them.
Editorial assessment · 26 August 2026Loading.
Rain, currency and the ocean state behind the crop now forming
Rainfall in the producing districts against the 1991–2020 normal, the currencies farmgate prices are paid in, and the Pacific state behind the drought risk.
Next-season supply risk has replaced current-season abundance as the dominant narrative.
Six forces, ranked. The design tells you which ones matter most; every one remains visible.
Editorial assessment · 26 August 2026The present crop is abundant; the next one is not — and the sources disagree about how much.
Where the beans are coming from, and how confident anyone can be. The production table is the primary evidence; balance, revision and disagreement sit beneath it.
Editorial assessment · 26 August 2026West Africa is roughly 60% of world output, and both dominant producers have guided next season lower
| Origin | 2024/25 | 2025/26 | 2026/27 f | Change | Constraint |
|---|---|---|---|---|---|
| Ivorian 2025/26 is cumulative port arrivals to 2 August, not full-season production. Ghana's 2026/27 figure is COCOBOD's internal estimate of 20 August; their public guidance three weeks earlier was 450–550 kt and StoneX has 586 kt. Indonesian and Brazilian tonnages are blank — see the note below. | |||||
Ivorian arrivals
The clearest evidence of a well-supplied present. Arrivals have run ahead of last season all year.
Ghana, delivered
A strong recovery from 597 kt. COCOBOD now guides 450–550 kt for 2026/27.
Exchange stocks
Covers only exchange-certified cocoa at US and EU delivery ports, not commercial inventory.
The surplus has not disappeared — it has been pulled forward
Net world crop minus grindings. The 2023/24 deficit was the largest in the modern record; what followed has been repeatedly revised.
2024/25 has been revised three times in twelve months — 49 kt, then 75 kt, back to 48 kt. Forward seasons are StoneX projections.
Across two revisions StoneX raised the current season from 247 kt to 422 kt while cutting next season from 149 kt to 25 kt. The surplus has not disappeared — it has been pulled forward. Production for 2026/27 is now seen down 6% at 4.83 Mt against demand near 4.75 Mt.
A source revising itself is one signal. Two sources disagreeing is another.
The gap between two estimates, relative to their midpoint. Where only one source exists there is nothing to compare — Nigeria is the clearest case.
West Africa still sets the price, but no longer accounts for all of the change
Two of the three origins below are moving fast, and the third has left the export market entirely.
A warning about Indonesian numberswhy the tonnage is blank
A warning about Indonesian numbers
Indonesia's official statistics agency put 2023 production at 641,700 tonnes, with Central Sulawesi alone at 130,800. Trade-house and exchange-facing estimates of Indonesian tradeable output run a small fraction of that. The gap is definitional — wet or unfermented farm-gate weight against fermented, dried, exportable beans — but it is large enough that quoting the official figure next to an ICCO balance would be misleading. This site leaves the Indonesian tonnage blank rather than pick a side. It is a good illustration of why origin-side measurement, done consistently, is worth paying for.
Production does not move cleanly from farm to export.
Where cocoa is grown, where it is ground, and the lanes between. The map is the evidence; the bottlenecks it cannot show — funding, payment, held stock — are what the field panel measures.
Structural — changes rarelyMap of cocoa production, export ports and shipping lanes
Amber circles are producing districts, sized by output. Teal rings are processing capacity. Solid dots are ports. Lanes show the dominant bean routes. District sizes are indicative; for published regional figures see the production atlas.
Colour is direction, not judgment. A short crop is bad news for a buyer and good news for a futures holder — the map takes no side. Dashed outline means no second source exists, so no spread can be computed.
A bean changes hands five or six times between a farm and a bar
Each hand takes a margin and carries a different risk.
Concentration is the point. Roughly three quarters of the world's grinding capacity sits with a handful of companies, and the two largest — Barry Callebaut and Olam Food Ingredients — process a substantial share between them. A farmer sells into a market with many sellers and very few buyers, which is most of the explanation for where the margin ends up.
European contraction is being partly offset by Asian and origin grinding.
Grindings are the best available proxy for chocolate demand, and the regions have decoupled.
Editorial assessment · 26 August 2026Europe's weakest second quarter since 2020; Asia up a quarter
Europe's 316 kt is the weakest second quarter since 2020, and first-half volume is down 7.8% year on year. Read the North American number with care: the NCA reporting panel changed after Q3 2025 and Hershey no longer participates, so the +7.65% is not a clean comparison.
Origin grinding is taking share
Côte d'Ivoire processed 161 kt domestically in Q2, up from 125 kt a year earlier, as producing countries push value addition onshore. Over time this moves the demand signal upstream, away from the European association data the market has always relied on — a slow erosion of the market's own visibility.
Grindings detailQ2 and H1 2026 by reporting body
| Region | Q2 2026 | y/y | H1 2026 | H1 y/y |
|---|
Cocoa has recovered sharply from the February low, and speculative length is thinning.
Futures price → speculative positioning → administered farmgate price → what the farmer receives. The thesis line is rewritten from the live series on load.
Written from live dataThe multi-year arc
The season clock in chapter 7 shows the same series inside the current season, against the events that moved it.
ICE New York
London Sep 2026
Off record high
Off 2026 low
Positioning
Loading from the weekly CFTC report.
Administered prices are creating funding and cross-border distortions
Prices at origin are administered, not discovered. That makes the gap between official and world price the most informative number in West African cocoa — and the driver of leakage.
| Origin | Main crop | Revised | Change |
|---|---|---|---|
| Reports of the Ivorian mid-crop cut vary between CFA 1,200 and 1,300 per kg. Both countries hold unsold stock — roughly 100 kt in Côte d'Ivoire, 50 kt at Ghanaian ports — and Abidjan committed CFA 280 bn to buy it back at the guaranteed price. | |||
The season calendars are merging
Ghana and Côte d'Ivoire have agreed to align the 2026/27 marketing year to a single 1 September – 31 August window and to harmonise guaranteed farmgate prices. Staggered seasons and divergent fixed prices are precisely what has historically driven physical smuggling across the shared border, so alignment is aimed squarely at closing that arbitrage.
EUDR lands mid-season
Plot-level geolocation becomes mandatory for large and medium operators on 30 December 2026, with micro and small enterprises following on 30 June 2027. The bind is that farmgate prices were cut 28.6% and 57% this season, so producers are being asked to fund mapping compliance out of sharply reduced income — with a Commission simplification review still capable of moving the requirements.
Arrivals and pod counts will decide whether Ghana's downgrade is real.
The gap between the well-supplied present and the forecast shortfall closes on a known schedule. Upcoming events first; the full season record beneath.
Editorial assessment · 26 August 2026What resolves next
The season clock
Cocoa runs on a hard annual cycle. Drag the hand, or the slider, to move through the 2025/26 season and see what was actually known on any date.
The line is the ICCO daily price, averaged by month. Hollow marks are individual ICE New York settlements — a different basis, which is why they disagree: the monthly series bottoms in March at $3,240, the lowest daily close was $2,952 on 24 February. Ticks along the bottom mark the events above.
What the farmer gets.
The question most people ask first, and the one the price series cannot answer on its own. Foundational rather than current — it sits here so it does not interrupt the market argument above.
Structural — changes rarelyIncome, the living income differential and child labourthe long read
A cocoa farmer in Ghana or Côte d'Ivoire typically works two to four hectares and sells at a price set by a marketing board, not negotiated. That price is announced per season, so a farmer carries none of the upside when world prices triple — and, as the 28.6% Ghanaian cut in February 2026 showed, is exposed when they fall.
Roughly 6% of the retail price of a chocolate bar reaches the farmer, on widely cited estimates. The living income differential — $400 a tonne, introduced jointly by Ghana and Côte d'Ivoire in 2019 — was designed to close that gap. It is paid on top of the world price and has been unevenly realised: buyers have at times offset it by negotiating down the origin differential, which is a separate line in the same contract.
The currency point matters here too. A farmgate price fixed in cedi loses real value as the currency depreciates, without anyone announcing a cut. The CFA franc, pegged to the euro, cannot move that way. Two farmers in neighbouring countries can see the same nominal price behave very differently.
Child labour is the issue most associated with cocoa, and the numbers are contested but large: a 2020 NORC study for the US Department of Labor estimated about 1.56 million children engaged in child labour in cocoa production across the two countries. Most is hazardous work on family farms rather than trafficking, which matters because the remedies differ — household income and school access, not only enforcement.
The honest position is that this is a poverty outcome before it is an enforcement failure. A household earning below a living income relies on family labour, and certification schemes have repeatedly been found to lift compliance paperwork faster than incomes.
Where we sit. Harmattan does not audit farms and does not certify anything. What the field panel measures — realized farmgate prices, payment delays, and whether buyers have the cash to pay at all — bears directly on household income, and that is the contribution we can honestly claim.
Every figure, its date, and what we may do with it.
Live feeds are fetched once a day at build time and stamped with the time of the last successful run. Hand-entered figures carry their citation where they appear. The full ledger — links, licences, usage status and lag — is on its own page.
Structural — changes rarely