$5.85 With Crude $28 Cheaper Than 2022 (4 September 2026): US Diesel Set an All-Time Record — and the Refining Margin, Not the Barrel, Built It

On Friday 4 September 2026 the US national average price of diesel reached $5.85 a gallon, the highest ever recorded, the Associated Press reported. West Texas Intermediate was trading near $91. At the previous record, in June 2022, WTI was between $110 and $119. Crude is roughly $28 a barrel cheaper than it was the last time the pump printed this number, and the pump has gone higher anyway — because the expensive part of a gallon of diesel is no longer the oil in it. It is the margin for converting oil into diesel, which reached $107.35 a barrel on 1 September against roughly $64 to $73 during the week of the 2022 peak. This is a refining shortage wearing the costume of an oil shock, and the two have different cures.

The previous version of this piece named three observables and said the crack spread was not one of them: distillate inventories against their five-year average, the refinery utilisation rate, and demand. All three have now reported, and all three moved the way a genuine conversion shortage moves rather than the way a risk premium moves. That is the case for treating this as a physical story rather than a headline one.

What actually happened

Diesel set its record at the end of a week in which crude rallied hard but not to new highs. Brent traded at $95.29 a barrel and WTI at $90.76 by midday in New York on Friday 4 September, each fractionally lower on the day, having gained 6.1% and 8.3% respectively over the week, according to Reuters. The proximate driver was the resumption of military exchanges between the United States and Iran in a conflict now in its seventh month; four commodity vessels transited the Strait of Hormuz on Thursday against a ten-day average of about 15, and the measurement of that waterway's throughput has itself become contested.

But the crude move is the smaller half of the story, and the comparison that shows it is the previous record at the pump. Set the two episodes side by side on the same public data series and the composition of the price inverts.

Spot crude and product prices: EIA daily spot prices; retail averages: AAA. The crack is ULSD in dollars per gallon multiplied by 42, less WTI.

In 2022 the world was short of crude and every product priced off it went up together. In 2026 the world has enough crude — US commercial inventories sit 1% above their five-year average — and is short of the plants that turn it into middle distillate. Gasoline, made in the same refineries from the same barrels, is 17% below its record. A shortage that lands on one product and not the other is not a shortage of oil.

The physical gap: harder runs, less diesel

The EIA's Weekly Petroleum Status Report for the week ending 28 August 2026, published 2 September, is the cleanest evidence available, because it reports both sides of the refinery gate in one document.

US refineries processed 17.5 million barrels a day, up 102,000 b/d on the week, at 98% of operable capacity — and distillate production fell to 5.1 million barrels a day. That is now the fourth consecutive week in which utilisation rose or held near its ceiling while distillate output declined. Gasoline output averaged 9.8 million b/d. The constraint is not how much crude the system can take in; it is what the surviving configuration of plants yields at the other end.

On the crude side, commercial inventories fell 4.5 million barrels to 424.5 million, which still leaves them 1% above the five-year average for the time of year. On the product side, distillate inventories rose 0.8 million barrels — the first build in several weeks — and remained about 14% below their five-year average. Gasoline fell 1.2 million barrels to 6% below average.

That national distillate build is where the report needs reading carefully, because underneath it one region broke a record in the other direction. Distillate stocks in PAD District 1, the East Coast, fell 1.685 million barrels to 19.318 million — the lowest weekly reading since the series began in 1990, and below the prior low of 20.966 million set in May 2022. The East Coast has little refining of its own and depends on product shipped or imported in, so it absorbs a global cargo shortage before the national average registers one. It is also the region where a heating season arrives in November. From here the regional series carries more information than the national one.

What is still physically broken

Three supply losses built this margin, and none of them is a market price that can be bid away.

Russia is the largest. It banned diesel exports on 8 July 2026 after sustained Ukrainian drone strikes cut domestic refining and triggered fuel shortages inside Russia, and on 29 August the government extended that ban to 30 September, with marine fuel and gas oils also covered. Separate restrictions run on diesel exports by non-producers and on motor gasoline until 31 January 2027, and on jet fuel until the end of November. Russian crude processing fell to roughly 3.91 million barrels a day in July, more than 1.4 million below the prior year's average. Russia supplied around 11% of global diesel in 2025, and those barrels have not been replaced.

Saudi Arabia is the most concrete. Aramco's Jazan refinery — 400,000 barrels a day — has been shut since 27 July after Houthi drone attacks damaged its gasification complex and tank farm, and its restart was pushed from 15 August to 30 August. No confirmed restart had been reported as of 4 September, which leaves the single largest identifiable swing factor in this market still absent.

The aggregate is visible in the IEA's Oil Market Report of 12 August 2026, which put global refinery crude throughput in July at 80.9 million barrels a day, nearly 5 million below year-earlier levels, and recorded diesel exports from Russia, the Middle East and Asia running 1.3 mb/d below the previous year — equivalent to about 20% of global seaborne trade in the product. The third loss is the counter-intuitive one and it is American: record distillate exports are the rational response of a refiner facing the widest margin in the world, and they are the reason a global shortage shows up on a US regional inventory number.

Demand is now the fastest-moving number

The supply side of this market has been deteriorating for six months. The demand side has started to move much faster, and it is the part that decides how long the episode lasts.

Distillate product supplied over the four weeks to 28 August 2026 averaged 3.7 million barrels a day, down 6% year on year. A week earlier the same four-week measure was down 2.2%, and a fortnight before that 0.8%. Total products supplied fell 4% and gasoline 2%. A single weekly print is noise; a run from plus 1.9% to minus 6% over five weeks is a demand curve doing what demand curves do at a record price — removing discretionary freight, deferring industrial consumption, and rationing the shortage by cost.

This matters more than any supply headline, because it works on the physical gap rather than on the premium attached to it. It is also the mechanism that ended the 2022 episode. The uncomfortable version of the same sentence is that demand destruction in diesel is demand destruction in freight, construction and agriculture — the price is being rationed by activity, not by conservation.

Where it lands: inflation, rates and the currency read

Diesel reaches consumer prices indirectly and slowly. US CPI rose 0.1% in July 2026 and the annual rate eased to 3.4% from 3.5%, with core at 2.5% and the gasoline index falling 2.9% on the month, per the Bureau of Labor Statistics. Households buy gasoline; businesses buy diesel, and it enters the basket through haulage, food distribution and anything that moves on a truck, with a lag measured in months and only to the degree firms pass it on.

Where it is already showing up is further out the curve. "All sectors of the economy are affected by diesel," Claudio Galimberti of Rystad Energy told Reuters on 4 September. "This is one of the reasons why the government bond yields in the United States are so high." That is the channel worth understanding: a business-input shock with an uncertain duration is exactly the kind of thing that lifts the compensation investors demand for holding long-dated paper, which is the same term-premium mechanism driving the long-end selloff. For the rate factor at the centre of any currency read, none of this settles the direction — a relative price shock in a business input is the category central banks have historically looked through, because tightening does not build refineries — but the duration question now has two competing answers inside the same dataset: supply that keeps getting worse, and demand that has started to fold.

The currency lesson has not changed through either phase of this episode, and it is the one most often got wrong. The price that rose is a processing margin, earned by whoever stands between crude and diesel. Canada sells the input — crude, overwhelmingly by pipeline to a single customer at a differential — and captures little of a conversion margin or a freight and insurance shock; our breakdown of the oil–CAD relationship sets out the structural reasons. The euro area is a structural net importer of both crude and diesel with a far higher diesel share in its vehicle fleet, so it takes the hit on the import bill. And the dollar feels this least directly through trade and most directly through the rates channel above. "Energy is up" is not a signal for any currency until you name which energy price moved and who sells at it.

What would change the picture

Three observables, in the order they would actually matter — and the crack spread is not one of them, because the crack is the price of the gap rather than the gap.

The first is East Coast distillate inventories. The national series can look merely tight while PADD 1 runs to a series low, and with a heating season arriving in November the regional number is now where the risk sits. Watch it against its own five-year range rather than against the national figure.

The second is conversion capacity returning. There is almost no headroom left in run rates at 98%, so realistic movement from here is downward through outages — the deferred-maintenance risk grows with every week the margin stays this wide. Jazan's restart is the nearest concrete addition and it remains unconfirmed; Russia's export ban expires on 30 September, a date that has been rolled forward twice already.

The third is whether demand keeps folding. Minus 6% year on year is a fast deterioration, and if it persists it closes the gap without any new capacity at all. That is the resolution the 2022 episode had, and it is not a benign one, because the adjustment happens in freight volumes and industrial output rather than in the price alone.

What none of these are is a forecast of the diesel price. The useful discipline is to ask which half of the barrel a given headline touches. A crude headline — an OPEC+ quota decision, a production ceiling, a strategic release — touches the input and has been close to irrelevant to the pump all year. A refining headline touches the constraint. This year the second category has done nearly all the work, and the reading order for the autumn is distillate stocks first, run rates second, the crude benchmark third — the reverse of the order most people use. You can see how the commodity factor currently sits across all eight majors on the live meter, and the wider method behind it on our about page.

Educational macro context only — not investment advice.