Tearing apart the cocoa mid-crop

Cocoa.

The FT ran a piece this week about how cocoa prices have "normalized" from the 2024 spike. Reuters followed with a line about grinders "breathing easier." Everyone has moved on. They're all looking at a spot chart and calling it a weather report.

Here's what the chart isn't telling them. The West African mid-crop — the smaller April-through-September harvest that covers the second half of global demand — is already arriving lighter than last year's disaster. Pod counts in Côte d'Ivoire are down. Farmgate arrivals at San Pedro are running 9% behind the five-year average. Nobody is writing about it yet because the grinders aren't panicking in public. They're panicking in private.

Cocoa isn't a commodity the way copper is a commodity. It's a biological asset with a four-year lag between planting and meaningful yield. You can't frack a cocoa tree. You can't open a new mine in Ecuador and have beans arriving in six months. When the trees in Ghana and Côte d'Ivoire are sick — and they are, with swollen shoot virus eating acreage at the fastest rate in a decade — there is no substitute supply that shows up next quarter.

The 2024 spike looked like a one-off to the people who read headlines. To anyone who's been inside the grinding business, it looked like the first time a structural deficit hit a market that had been running on vibes and goodwill for fifteen years. I was cautious calling the top in spring 2024 and the market punished me for it. I'm more cautious about calling the bottom now.

Story off. Numbers on.

ICE-certified cocoa stocks at New York warehouses sit at roughly 1.39 million bags. For context, that number was 3.7 million bags in mid-2022. The exchange inventory has been drawn down by almost two-thirds in under four years, and the drawdown has continued even as spot prices stabilized. That is the tell. Prices aren't rising right now because grinders finished destocking and physical buyers retreated to contract-only purchasing. The moment anyone needs prompt beans, there are no prompt beans.

Global stocks-to-grind ratio is projected at 27% for 2025/26 — the lowest reading since the series began in the 1960s. Anything under 35% has historically meant price dislocations. We are fifteen points under that line and nobody is writing about it.

Every node in that chain has been running thinner than normal since late 2023. The system has no slack left. And the node at the top — the harvest itself — is the one nobody can engineer their way around.

One more number. Ghana's cocoa regulator Cocobod has rolled forward contracted deliveries three years in a row — beans sold but not yet delivered, pushed into the next season to make the books close. That's about 350,000 tonnes of phantom supply hanging over origin accounting. If the 2026/27 crop misses again, the rollforward trick stops working.

Today is Friday. That means the CFTC's Commitments of Traders report drops at 3:30 PM Eastern. If last week's reading is any guide, it will show managed money on the short side of cocoa at the heaviest net position since 2017.

The thesis among the funds is clean. Prices fell from the March 2024 highs, therefore the shortage story is over, therefore short the rally. It is the same trade as copper in the sample issue I wrote last week. Different commodity, identical blind spot. The algorithms are looking at the price chart. The physical market is a completely separate organism.

I've been short soft commodities during a supply shock exactly once in my career. I'm still writing about it for a reason. The loss takes a specific shape: you watch open interest bleed out of your position as other shorts cover into a rising tape, and you convince yourself for about six hours that you're the smart one holding. Then the physical bid shows up and the next print is seventy handles higher with no trades in between.

Physical grinders in the Netherlands are reportedly paying origin differentials 400 basis points over last quarter for prompt Ivorian beans. That number comes from broker chatter, not an official report, so treat it as directional. But when grinders start paying premiums over the futures contract, the curve is telling you something the specs aren't listening to. It never has been.

Here is the sequence. The trigger this time is a harvest data point, not a strike. The mechanics are the same.

Official Ghana and Côte d'Ivoire mid-crop arrival numbers are published in May. If those numbers confirm the 9% gap we're already seeing in private counts, the specs have to cover a market with no offered physical. The first leg will look orderly. CTAs flip from short to flat, and price drifts up twelve to fifteen percent over a week. That's the polite part.

Then the grinders realize their Q3 coverage isn't going to be honored at contract prices and they come into the market looking for beans that don't exist. That's the impolite part. We saw it in March 2024 when spot cocoa went from $5,000 to $12,000 a tonne in eleven weeks. Nothing about the underlying machine has been fixed in the two years since.

Where does capital go? Not into the big confectionery names — Hershey, Mondelez, Lindt. Those are the ones eating the cost pass-through, not profiting from it. Their forward cocoa hedges from 2024 roll off this year and next year, and the replacement hedges are being booked at structurally higher levels. Margin compression is already in the consensus numbers for the downstream. It isn't in the stock prices yet.

The edge, if it's anywhere, is in the upstream processors and the few publicly-listed origin-exposed trading houses. Barry Callebaut on a pullback. Olam's agri business. Names with exposure to the toll on the beans passing through their machines rather than the cost of the beans themselves. Those businesses earn more when prices are volatile and physical flow is tight, not less.

Financial layer says the crisis is over. Physical layer says the crisis never actually ended — it just got quiet while everyone who was going to destock destocked. In my experience, when the chart and the loading dock disagree, bet on whoever is actually touching the product. I have been wrong about timing before. I have not been wrong about which side wins.