A surplus sounds like a gift if you are standing at the drinking end of it. More coffee, cheaper coffee, better coffee for the same money. That is not what happened the last time the world had far too much coffee, and the record of what did happen is the most useful document in coffee economics, because the mechanism that produced it has never been dismantled.
When there is too much coffee, the surplus is not sold cheaply. It is bought up, warehoused, destroyed or converted into something else, and the cost lands on the grower rather than on the drinker. Jonathan Morris records the scale of it in Coffee: A Global History: 75 furnaces built for the purpose, running from 1931 to 1939, and something close to 80 million bags destroyed, a volume equal to about three years of what the whole world drank at the time. The 1937 figure alone is usually given as more than 17 million bags against world consumption near 26 million, and that pairing is worth reading as an order of magnitude rather than an audited number. What the furnaces could not take, the railways did. Beans were blended with tar until locomotive boilers would burn them, which at least got some traction out of the crop on its way to the same end.
Why burn it instead of selling it
Demand for coffee is inelastic. People who drink two cups a day do not start drinking four because the price halved, and people who drink none do not start because it got cheap. So an extra ton of coffee on the market does not find an extra ton of drinkers. It finds the same drinkers and lowers the price on every ton already in the chain.
Gavin Fridell, in Coffee, quotes the journalist Mark Pendergrast on the year that makes this cleanest. Set Brazil's 1938 trade with the United States beside its 1937 trade: the volume went up by 300 million pounds, and the money came back down by roughly $3.15 million. More volume, less revenue. That single comparison explains why a producing country with market power would rather set fire to a warehouse than clear it, and it is the arithmetic at the bottom of every commodity trap since.
The Brazilian instrument had a name and a start date. Valorization was the name: the state buys its own harvest and sits on it, so the price at the port never has to absorb the whole crop at once. Brazil reached for it first in 1906, a year when its output hit 20.2 million bags and its share of everything the world grew ran near 85 per cent, both figures from Morris. The tool came back in 1917, in 1921 and in 1926. It works, in the narrow sense that any state can hold a price up for as long as its treasury is willing to keep buying. It breaks the way every version of it breaks. Fridell's account has the 1929 crash landing on a government already holding bags by the tens of millions and no longer able to carry them, left with a choice between pushing them into a falling market and destroying them. Brazil destroyed them.
The instant coffee in your cupboard is a disposal project
The commission went out in 1929, from Brazilian officials to Nestlé, and it reads as industrial policy because that is what it was: build us a coffee product that will take up the surplus and keep on a shelf. Morris credits Nestlé's Max Morgenthaler with six-plus years of work before the result was drinkable. Nescafé reached the market in 1938. Then the war came, and by Morris's account the US military took up nearly all of what the factories could make.
That is the honest origin of soluble coffee. Not convenience, and not a breakthrough somebody had been waiting for. A glut needed somewhere to go, and a food company was commissioned to build the somewhere. It is worth holding that next to how instant is marketed now.
What the burning did not do
It did not raise the price durably, and it did not reach the farmer. Destroying stock lifts the price only while the destruction continues, because the trees that produced the surplus are still standing and will produce again next year. Supply destruction treats the inventory and leaves the planting untouched, which is why Brazil had to keep doing it for nine consecutive years rather than once. It also did nothing about the structural cause, which was that a high price in one decade recruits planting that arrives as fruit in the next, long after the price that justified it has gone.
And the money never travelled downward. A price defended at the port is a price defended for whoever owns the coffee at the port. Hermann Sielcken, the German-born merchant who had led the syndicate financing the first valorization scheme, was called to answer for it before a US congressional committee in 1912. The dispute was between a producing state, a merchant syndicate and an importing government. The people who picked the cherries were not represented in the room and did not appear in the settlement.
The last thing it did not do was end. A century on from 1906, Brazil went back to the same tool: Fridell records an announcement in August 2013 that the state would buy three million bags and hold them back to support the price. Whether it has been reached for again since, we have not checked, and we are not going to imply a continuous policy record we cannot show.
The 20th century tried the same idea with more countries in the room
The International Coffee Agreement, running from the early 1960s to 1989, was the only mechanism in modern history that raised prices for producers across the board, and the reason it worked is the reason it is hard to revive. Fridell puts the export ceiling near 47 million bags of green coffee bound for the North American and Western European markets, divided into national allocations. Importing countries were inside the agreement and had agreed, in effect, to pay more. Consumers were never asked, and would probably have said no if they had been.
It leaked constantly. Coffee sold cheaply into non-member markets found its way back into member markets, and the underlying overproduction never stopped. When the agreement collapsed in 1989 the floor went with it. Between 1998 and 2002 the ICO composite indicator fell from 109 US cents a pound to under 48 cents, and robusta from 83 cents to 28, figures Morris sets out in his chapter on the crisis. He also tracks what it did on the ground. Guatemala lost half the people employed in coffee. In Colombia the trees came out and coca went in, because coca paid.
The number that matters in that sequence is not the average price, and Fridell shows why with the arithmetic. Take the quota years, 1963 to 1989, against the two decades that followed, 1990 to 2011: the mean composite price sits at roughly 94 cents a pound on both sides of the line. Near-identical averages, completely different experiences. What changed was the path between them. Volatility is what ruins a smallholder. It removes the ability to plan a crop cycle that takes five years to come good, and the ability to borrow against next season. A steady 94 cents and a 94-cent average reached by swinging between 48 and 300 are entirely different products from the farm's point of view, and only one of them supports the seasonal, multi-year rhythm a coffee tree actually runs on.
Today the problem is the opposite one, which is the point
The market of the last two years has not been a glut. The ICO composite indicator averaged 354.32 US cents a pound in February 2025, the highest monthly figure in the series, with a daily peak of 375.00 cents on 13 February 2025. That beat the previous nominal record of 305.13 cents set in March 1977. By 24 July 2026 the composite had come back to 274.72 cents, which is about 27 per cent off the peak and still roughly 47 per cent above where it sat in March 2024.
Meanwhile the supply picture has turned. USDA's Foreign Agricultural Service forecast a record world crop of 189.667 million bags for 2026/27, with ending stocks recovering to 26.283 million. That is the setup for the old cycle to run again, and if it does, the history above is the guide to who absorbs it. What is genuinely different this time is that the growing area itself is moving, and the ground suitable for arabica is being squeezed upward, which is a constraint no incinerator or quota ever had to price in.
Two cautions on reading that. First, the 2024 to 2026 price regime may have broken the assumptions older coffee economics ran on, and a forecast surplus in a market this reactive to weather headlines is a forecast, not a fact. Second, the surplus arriving does not mean your bag gets cheaper. US retail ground roast coffee went from $5.964 a pound in March 2024 to a peak of $9.723 in April 2026 and was $9.457 in June 2026, only 2.7 per cent off the top, all figures as published by the Bureau of Labor Statistics and correct as of August 2026. Retail followed the green price up over fourteen months and has barely begun to follow it down, which is the same asymmetry that governs every bag you buy.
What to do with this
Nothing about the incinerators changes how you brew. It changes how you read a price.
When you next see a headline about coffee prices falling, the useful question is not what your supermarket will charge in a month. It is who is currently holding the surplus, because that is who decides whether it reaches the market at all. In 1937 the answer was a government with 75 furnaces. Today it is a mix of exporters, exchange warehouses and roasters, and none of them has any obligation to pass a lower green price to you or a higher one to the grower.
The practical version, for anyone who wants their money to behave differently: buy from roasters who state what they paid for the lot and to whom, and treat everything else as unverified. A farm name on a bag signals a great deal less than most people assume, and a price with a date attached to it signals a great deal more.