At the start of the week, a line slipped through the news feed—easy to scroll past between sanctions and strikes on oil refineries: the G7 had agreed to release 100 million barrels of oil and diesel to bring prices down. Beside it was the number that caught the eye. A gallon of diesel in the United States costs around $6.5, a record in nominal terms for the entire period of observation. Meanwhile, Brent crude is trading around $100 a barrel—well below the inflation-adjusted peaks of past crises.
So the question looks like this: if the raw material is not at a historic high, why is the product made from it at one? The answer, on the surface, seems political: Trump accuses oil companies of greed, senators demand investigations, Europe taps its reserves. But behind the arithmetic lies something more interesting. Between the well and the pump stands a refinery, and that refinery is not made of rubber. Diesel costs more than oil not because someone is marking it up, but because there is physically no one capable of making more of it.
A crack spread is the difference between the price of a barrel of crude and the price of the product made from it. The figure is often retold as “refiner greed,” though it answers to no one: market prices for the feedstock and the output determine it, and it widens precisely when there is too little product relative to crude. In normal, quiet market conditions, the diesel spread sits at $20–30 a barrel. In mid-August it broke $100 for the first time in history, and on September 14 it set another record—$118.62.
Then comes the part worth digging for. The same crack spread for Asian diesel rose to $87 from roughly $22 before the war—an almost fourfold increase. Meanwhile, U.S. refineries are operating at 97–98% of capacity—the upper limit of their nameplate capability. When a refiner is squeezing everything out of its equipment and margins are still rising, economics says one thing: the problem is not that someone is not trying hard enough. The problem is that there is nothing left with which to try harder.
This is the moment to separate two things that news coverage sticks together under the single word “oil.” Crude and fuel are different markets with different balances. The world may have enough crude and still lack diesel, because diesel has to be produced. Refining is not a lever you pull on command; it is an engineering system with a rigidly built-in output profile.
A standard complex refinery on the Gulf Coast is designed to yield roughly 60% gasoline components and 30% middle distillates—diesel and jet fuel. That is what it was built for: for decades, the American consumer burned more gasoline than diesel, and refineries were erected around that imbalance. That, incidentally, is where the familiar “3-2-1” formula comes from: three barrels of crude, two going into gasoline and one into diesel. It is not a law of nature. It is a description of one specific historical configuration.
You cannot switch a refinery over with the flick of a switch. Between heavy crude fractions and light diesel sits a hydrocracker—hydrogen, catalyst, dozens of atmospheres of pressure. It is what turns material that would otherwise become fuel oil into diesel and kerosene, adding 20–30% more finished liters to the original volume along the way. Total U.S. hydrocracking capacity, when people began counting it seriously, was around 2 million barrels a day—for a country that produces roughly 5 million barrels of diesel daily. Increasing diesel yield here is not an adjustment. It is construction: trillion-dollar investments, years of building, permits, a site.
That is why record margins do not turn into record output. The economic signal—“make more diesel”—runs into a steel wall: you cannot convert a gasoline refinery into a diesel refinery any faster than you can build a new one, and no one is rushing to build new ones in this market. There are reasons for that.
The European story is more interesting than the American one because its gap was built in historically. In the 1990s, Europe deliberately dieselized its passenger-car fleet: tax incentives and climate logic—diesel promised 15–18% lower emissions per kilometer—made it the mass fuel for cars. But Europe’s own refining sector was, and remains, “gasoline-oriented”: it has a persistent gasoline surplus and a persistent diesel deficit. Back in the 2010s, the gasoline surplus reached 800,000 barrels a day, while the diesel shortfall hit 560,000. The gap is not new. What is new is what used to patch it.
It was patched with imports, and the geography of those imports has been turned upside down over ten years. In 2013, Russia was the EU’s main diesel supplier—31%. By 2025, Saudi Arabia had taken its place with 26%, while the Gulf’s share of European diesel imports had risen roughly thirteenfold—from 2.4% to 31.6%. Europe successfully rid itself of one supplier—and immediately fell into a new dependency, this time a maritime one: through the Bab el-Mandeb Strait and the Strait of Hormuz, which became the least reliable places on the map during the war with Iran.
And this is where feedstock chemistry enters the picture—a subject almost nobody writes about because it seems boring. European refineries were built around medium-sulfur crude such as Russia’s Urals: it yields more middle distillates directly from the atmospheric column. After abandoning Russian feedstock, Europe shifted to light, sweet grades—which produce more gasoline and naphtha and less diesel. The share of medium-sulfur crude in European imports fell from 31% in 2013 to 6% in 2025, while light sweet crude rose from 39% to 51%. The refineries are the same, but what now comes through their gates is not the feedstock they were calibrated for.
Nor can refining be ramped up quickly for another reason: there is simply less of it. EU capacity fell from 13.8 million barrels a day in 2000 to 12.0 million in 2025—a 13% drop. Seven of the union’s twenty-seven countries have no refining at all, and production is concentrated in five or six states. Britain, whose refineries at Grangemouth and Immingham have shut down, now imports more than half its diesel. When such a system loses external supplies, it has no internal route to make up the difference.
Now for the G7 and the 100 million barrels—the figure with which this text began. Its least obvious detail is that tapping the strategic reserve and getting diesel are not the same thing. The U.S. Strategic Petroleum Reserve (SPR) stores only crude oil: roughly 40% sweet and 60% sour, in salt caverns 2.5 kilometers underground. Not because that is more convenient, but because when it was planned in 1977, the decision was that the industry would refine whatever was needed itself, while crude was cheaper to buy, store, and transport—and unlike finished fuel, which ages and changes its properties, it does not spoil over time.
In other words, the SPR is a barn crammed with grain beside a mill that is precisely what the system lacks. Releasing 50 million barrels of crude oil—the second half of the French proposal—gives the market not a liter of diesel until someone refines that oil, and refining is the bottleneck. There is exactly one real reserve of fuel itself, and it is tiny: the NorthEast Home Heating Oil Reserve in the United States—one million barrels of ultra-low-sulfur diesel, created in 2011 to heat New England. One million for a global market losing 1.6 million barrels a day.
Which is why the second half of the G7 deal—50 million barrels of diesel specifically, with “front-loaded” deliveries in the first twenty days—is the only part that works as intended. A reserve treats the symptom: it buys the market time, but creates not one ton of new capacity.
A shortage has its mirror half—demand that does not respond to price the way it is supposed to. Normally, rising fuel prices cool consumption and the market balances itself. That does not happen with diesel, and to understand why is to understand who burns it.
Diesel is the fuel of things that do not stop on command. Trucks hauling everything, including food to the store. Combines and tractors at the height of harvest. Construction equipment, mining, maritime shipping. And—a new layer that did not exist in the calculations ten years ago—backup diesel generators at data centers: centers training and serving AI models install powerful diesel generators for outages because an hour of downtime at a hyperscale facility costs absurd money. This is demand that does not look at price: a generator is insurance, and insurance is bought regardless of the spread.
Electrification is gradually eating into gasoline demand from passenger cars, but diesel has no close electric substitute at scale. Gasoline can give ground gracefully; diesel and jet fuel cannot, because what they set in motion cannot be switched to a battery in a single winter. Demand that refuses to bend meets supply that falls in steps—and that is not a recipe for a temporary spike. It is the description of a lasting premium.
There is a layer to this picture that usually does not make business news, because it is about metal. American refineries have held utilization above 95% for eleven consecutive weeks—something that has happened only three times since 1990. A similar push in 1998, twenty-four weeks at the limit, was followed by payback: by mid-October, utilization had collapsed to 86%, because several plants had to be urgently taken offline for repairs after months of operating at full tilt.
The logic is brutal. Every day offline means lost output at margins twice the decade average, so preventive maintenance is postponed. Already, some repairs scheduled for the second quarter have been pushed to late 2026 and even 2027. The longer a refinery operates beyond its normal regime, the greater the chance that one day it will shut itself down—and then that will be lost capacity, not deferred capacity. In a system where the global refining shortage is estimated at nearly 2 million barrels a day, an unexpected failure at one large refinery hits prices harder than any political statement.
The paradox is that success itself is the source of risk. The more expensive the margin, the stronger the temptation to run equipment into the ground; the more severe the wear, the closer the breakdown. The market pays refiners precisely to bring their own failure nearer.
Put the numbers together. Global refining in July stood at 80.9 million barrels a day—roughly 5 million below a year earlier. Combined diesel and gasoil exports from Russia and the Gulf were down almost 1.6 million barrels a day from the start of the year. U.S. distillate inventories on September 11 stood at 107.9 million barrels—the lowest for this time of year since records began in 1982. China, which had served as a balancing force for years, simply halted fuel exports in October to preserve its domestic market.
And over all this lies the original structural loss, which came before today’s wars: according to the International Energy Agency, since the start of the pandemic the world has shut around 3.5 million barrels a day of crude-distillation capacity, while bringing 2.5 million in new refineries online; a net loss of 1 million barrels. The U.S. market alone lost around 1.235 million barrels of refining capacity in roughly fifteen months—more than 6.5% of the country’s operable atmospheric-distillation capacity. Refineries close forever: permits, capital, and trained operators do not return to a vacated site, and every shutdown is a one-way door.
There is a telling detail here, too, one that rarely gets noticed. In North America, available diesel storage capacity for lease rose sharply—from 11 million barrels in June to 13 in October, a four-year high. It sounds like good news, but it means the opposite: tank farms are empty because there is nothing to store. The market is not building inventory—it has forgotten how to accumulate it.