Tearing apart the cocoa harvest

Cocoa.

The consensus take is comfortable: the 2025/26 surplus filled the hole, Ivory Coast port arrivals are running 20% ahead of last year, and prices are down 55% from the December 2024 panic high of $12,931 per tonne. Bloomberg calls it a recovery. Reuters filed a piece last week about how chocolate prices might start easing. Crisis over.

They're reading the rearview mirror. The windshield shows something else entirely.

COCOBOD — Ghana's cocoa regulator — just projected 2026/27 production at 450,000 to 550,000 metric tonnes. That's down from 750,000 projected for the current season. A 27 to 40% collapse in a single year. The cause isn't one storm. It's three things hitting the same tree at the same time: swollen shoot virus still eating through branches, the off-year in the biennial bearing cycle, and a Super El Niño that NOAA says has a 95% chance of reaching "very strong" classification during October–December.

The sellside is still modeling off the 2025/26 surplus. That surplus exists. That cocoa is real. And it is in the past. The next crop starts September 1, and the trees are telling you it's not there.

West Africa produces roughly two-thirds of the world's cocoa. When both Ghana and Ivory Coast flag simultaneous crop declines heading into El Niño's peak — that's not a weather scare. That's a supply valve closing on a market that never fully rebuilt its buffer.

Story off. Numbers on.

The Niño 3.4 weekly index hit +2.7°C as of August 12. That is not a typo. NOAA's median forecast for October–December is +2.66°C on the Relative Oceanic Niño Index. The Super El Niño threshold is +2.0°C RONI. The 1997–98 event — the previous record holder — peaked at +2.4°C. This one is tracking above it and still accelerating. NOAA assigns a 95% probability it reaches "very strong" during October–December.

What does El Niño do to West Africa? The mechanism is clean: warmer, drier conditions during the August–October pod development window. That window determines the main crop. Cherelle formation — the development of young pods on the tree — has already been surveyed as below average across both Ivory Coast and Ghana. The main crop runs September 1 through end of February. The damage window is now.

Ivory Coast — the world's largest producer — expects its 2026/27 main crop down more than 10%. Four pod counters and five major exporters see it landing around 1.35–1.45 million tonnes versus last season's roughly 1.6 million. Oxford Economics goes further: a full-season decline of roughly 20% when El Niño compounds with elevated fertilizer costs. Ivory Coast weekly arrivals for the week ending August 16 came in at 4,000 metric tonnes. Same week last year: 10,000. The mid-crop is dying on the vine.

And ICE certified warehouse stocks? On August 5, they reported at 3,384,965 bags — a two-year high. That sounds like comfort. But against a market consuming over 5 million tonnes a year, it's coverage measured in weeks, not months. And it was rebuilt entirely on the surplus season that's about to end.

StoneX cut its 2026/27 global surplus forecast from 267,000 tonnes in January to 149,000 in April. That was before COCOBOD's projection. Before the latest El Niño intensification. StoneX now puts the 2026/27 surplus at just 25,000 tonnes — a rounding error in a market that consumes over 5 million tonnes a year. One more downward revision and we're back in deficit.

Nigeria isn't helping either. Its cocoa output for 2025/26 was already projected down 11% year-on-year to 305,000 tonnes by the country's own cocoa association — and that was before any El Niño impact was priced in.

Here's what the machines are missing.

The sellside narrative rests on one number: the 2025/26 surplus. Ivory Coast port arrivals ran 20% ahead of last year through early August — 2.11 million tonnes cumulative. That's real cocoa. It exists. And it's the reason managed money built the short position they did. As of July's CFTC data, spec short contracts sat at 29,566. Net positioning was negative as recently as May. They're trading the recovery crop.

The shorts are trading the recovery year. The physical market is pricing the crop that isn't developing on the trees. These are not the same trade, and the gap between them is widening weekly.

But the hidden vulnerability isn't just weather. It's the farmgate pricing collapse. In March, Ivory Coast slashed the mid-crop farmgate price from CFA 2,800/kg to CFA 1,200/kg — a 57% cut. When you pay farmers less than production cost, they stop spraying, stop fertilizing, and start letting marginal trees die. That's not a one-season decision. That's structural underinvestment in the tree stock that compounds for years.

I've watched this pattern before in other tree crops. Palm oil did it in 2020. Rubber did it earlier. The timeline is always the same: high prices drive overproduction, the crash guts farmer income, underinvestment hollows out the next cycle, and three years later you're staring at a supply hole that nobody saw coming because it didn't show up in a single quarterly report. It showed up in ten thousand farmers deciding not to buy fertilizer.

And the fertilizer shortage is the part nobody is connecting. West African cocoa is a heavy fertilizer user, and the export bans we've been tracking — China's industrial sulfuric acid export ban effective May 1, Russia's industrial sulfur export ban extended through December, the Hormuz disruption choking seaborne supply — have pushed input costs well beyond what a CFA 1,200/kg farmgate price can absorb. The farmer math doesn't work. When farmer math doesn't work during a Super El Niño stress event, you lose not just next year's pods. You lose branches. Branches don't grow back on a nursery tree's timeline.

European grindings hit a 17-year low in Q1 — down 7.8% to 325,895 tonnes. North America down 3.8%. Demand destruction from $12,000 cocoa is real. But it's masking how tight the next season gets. When demand recovers — and it will, because chocolate is the last consumer discretionary that gets cut — it recovers into a supply hole.

The Q1 grindings number will be cited as evidence that demand is too weak to drive a squeeze. Don't buy it. Demand for chocolate doesn't disappear — it reprices. Hershey and Mondelēz are both on record discussing reformulation and package downsizing, not demand collapse. They're buying time. When they stop buying time and start buying beans for Q1 2027 delivery, the physical market finds out how thin the pipeline really is.

The trigger is September 1. That's when the 2026/27 main crop officially begins in West Africa. The first hard data — initial port arrivals from the new season — hits in October.

If the early cherelle surveys hold — and El Niño intensifying into peak season makes them worse, not better — this is how it resolves. First, the surplus consensus evaporates. StoneX, ICCO, and the trade houses revise to deficit. Price pushes through $6,000. The 29,566 contracts of managed money short start covering. Second wave: the physical buyers — the Hersheys, the Mondelezes, the Barry Callebauts — who held off on forward cover because they believed the surplus story, realize the window is closing and scramble for Q1 2027 delivery. That's demand re-entering a supply vacuum.

Cocoa did this in 2024. The price went from $4,000 to nearly $13,000 in under nine months. I know a cocoa trader who covered a short in March 2024 at $8,600 that he'd put on at $3,200 three months earlier. He still describes it as the most educational loss of his career. The mechanics are identical every time. Only the trigger and the calendar change.

Where does the capital go? Not into the broad agricultural commodity ETFs. Those are diluted with grains and oilseeds that may actually benefit from this El Niño pattern — Brazil is expecting a record coffee crop, and South American soybeans look strong. The edge, if there is one, is in companies sitting at the bottleneck between bean supply and chocolate manufacturing. Barry Callebaut and Olam Food Ingredients (ofi) are among the largest global cocoa processors. Their margins expand when raw cocoa prices rise faster than forward contracts they've already locked in. They are the levered play on a surplus-to-deficit flip that the physical market is already signaling.

The counterargument deserves its sentence: if Q4 rains arrive on schedule and El Niño underperforms the models, the cherelle signal could be a false alarm and the surplus holds. That's the bet the shorts are making. I respect it. But the ocean temperature is +2.7°C and climbing, COCOBOD doesn't make public forecasts this bearish for fun, and ICE warehouse stocks — even at a two-year high — amount to weeks of consumption, not months.

Financial layer says surplus. Physical layer says deficit. The 2026/27 season starts in ten days. In my experience, when these two disagree on tree crops, the trees don't lie. They just take until October to prove it.