|
Part I
The Mechanism
|
Coffee.
The USDA just printed a record global production forecast: 189.7 million bags for 2026/27. Bloomberg ran the number. Reuters ran the number. Every sell-side desk in London dutifully updated their "surplus" models and called it bearish.
Meanwhile, ICE certified arabica stocks just fell to approximately 229,000 bags. That's down from approximately 730,000 bags at this point last year — a 69% drawdown. Thirty-nine consecutive sessions of declines ran through mid-August before the streak finally broke. The deliverable supply sitting in exchange warehouses right now wouldn't cover two days of global consumption.
This is the record crop that was supposed to fix everything. Brazil alone is projected at 66.7 million bags by CONAB — the largest harvest in its history. And yet the physical coffee is not showing up where it needs to be. The harvest is behind schedule. The ports are damaged. The shipping lanes are rerouted.
Everyone is staring at the production forecast. Nobody is watching the delivery chain.
|
Part II
The Diagram
|
Headlines off. Logistics on.
Brazil's 2026/27 arabica harvest was 86% complete as of August 12. Last year at the same date: 95%. The five-year average for Brazil's total coffee harvest: 94%. Heavy rains through June and July knocked beans to the ground, slowed mechanical harvesting, and created quality concerns that are causing Brazilian lots to fail ICE certification at elevated rates. Record crop or not, what hasn't been picked can't be shipped, and what fails grading doesn't enter the deliverable supply.
Then Colombia. A 7.4-magnitude earthquake hit western Colombia on August 10, shutting down Buenaventura — the Pacific port that handles roughly 60% of Colombia's coffee exports. Heavy cargo vehicles resumed partial movement by August 12, but road inspections are ongoing and there's no timeline for full capacity. Colombia is the world's second-largest arabica producer. Losing 60% of its export throughput, even temporarily, into a market already running on fumes is the kind of event that turns a tight market into a rationed one.
European port stocks are confirming the squeeze from the demand side. The European Coffee Federation reported green coffee inventories declined 3.6% year-over-year in April to 6.82 million bags — still near the nearly two-year low of 6.60 million bags hit in March. Destination stocks are what roasters actually draw from. When those run low, procurement shifts from routine to scramble.
And the shipping layer is still broken. The Strait of Hormuz has been effectively closed since February 28, disrupting Asia-origin trade lanes. Meanwhile, ongoing Red Sea diversions mean Vietnamese robusta — normally the relief valve when arabica tightens — is taking two extra weeks and costing more to reach European roasters. The substitution mechanism is not functioning at normal speed or cost.
The USDA says there's a surplus. The warehouses say there isn't. One of them is wrong, and warehouses don't have PR departments.
|
Part III
The Weak Link
|
Here's where the dashboard watchers are going to get hurt. They're pricing the crop they can see. They're not pricing the crop that hasn't flowered yet.
El Niño has officially formed. NOAA is giving greater than 90% odds of a "very strong" event by November–January — that would rank among the most powerful in the modern record. Brazil's critical flowering window for the 2027/28 crop opens in September and runs through October. If El Niño delays the spring rains in Minas Gerais during that window, the damage isn't just a lower yield number in twelve months. It's a structural hit to the tree's reproductive cycle that can suppress output for two consecutive seasons.
I've watched this movie before. In 2015, El Niño hammered Brazilian robusta — nearly 40% of production was lost, while arabica was largely spared. The market didn't price the full impact until months later, when the crop-survey data finally confirmed what the weather stations had been screaming. The sell-side was still publishing surplus forecasts while the trees were failing.
The managed money longs are already elevated — north of 42,000 contracts by late March, and they've stayed crowded. But the market is long for the wrong reason. They're playing the 2026/27 harvest delay trade. They're not positioned for the possibility that the 2027/28 crop gets compromised before the 2026/27 beans are even in the warehouse. That's a very different trade with a very different duration.
This is where it gets dangerous. If El Niño delivers, the market has to simultaneously reprice the current crop (delayed), the near-term delivery chain (broken), and the next crop (threatened). Three legs collapsing at once. I've seen two-legged collapses in coffee before. Three is a different animal.
|
Part IV
The Chain Reaction
|
Walk through the sequence.
Colombia's Buenaventura stays partially impaired through September. Brazil's harvest limps to completion two to three weeks late. ICE stocks continue to bleed. That alone keeps arabica in the $3.00–3.50 range. Uncomfortable but manageable for most roasters.
Then NOAA's September update drops and confirms the "very strong" classification. The flowering-window data starts trickling in from Minas Gerais. If the rains are late — even by two weeks — the crop scouts start cutting 2027/28 estimates. That's when the second wave hits. Roasters who were waiting to buy on the "surplus dip" realize there is no dip. They hit the bid for forward coverage. The curve inverts violently.
Where does capital go? Not into broad agricultural commodity ETFs — those are diluted with grains, livestock, and oilseeds that have their own dynamics. The edge, if there is one, is in companies positioned as toll booths on the delivery chain. Origin-country exporters and cooperatives with physical inventory already in hand. Integrated roaster-traders whose forward books lock in today's price against tomorrow's scarcity. Companies whose margins expand precisely when the spot market panics.
The consensus says record crop, surplus, prices come down. The exchange warehouses, the port in Buenaventura, the harvest monitors in Minas Gerais, and NOAA's Pacific buoys all say something else. In my experience, when the logistics chain and the weather models agree against the forecast, the forecast loses. It just takes one more data point to prove it.
