A refinery buys one thing and sells several. It buys crude oil, and it sells gasoline, jet fuel, diesel, fuel oil and a handful of specialities. The difference between what it pays for the barrel and what it gets for the products made from that barrel is its gross margin — and in the market that margin has a name.
It is called a crack spread, from the cracking units that break heavy molecules into lighter ones.
The single most useful thing to understand about crack spreads is this: a crack spread is not a bet on the oil price. It is the price of refining capacity. It can be at a record high while crude is falling, and it can collapse while crude is rising. In 2026 both happened, in the same year.
The barrel is not one product
Crude oil is a soup of hydrocarbon molecules of different sizes. Distillation separates them by boiling point: heat the crude, and the lightest molecules vaporise first and are drawn off at the top of the column, the heaviest never vaporise at all and are drawn off at the bottom.
That is where the market’s vocabulary comes from. The light end comes off the top, the middle distillates from the middle, and the heavy end — which is properly a residue, not a distillate at all — is what is left in the bottom of the column.
One barrel in, several products out
Indicative yield from a complex refinery, by volume. Boiling ranges are approximate
Yields vary enormously by crude type and refinery configuration. A light sweet crude in a complex US Gulf refinery gives a very different slate from a heavy sour crude in a simple hydroskimmer — and that difference is the whole business.
The trade talks about "light distillates", "middle distillates" and, loosely, "heavy distillates". Strictly, only the first two are distillates. The heavy end — residual fuel oil, bitumen, coke — is what did not distil. It is atmospheric or vacuum residue. The distinction matters commercially: distillates are priced off demand for a fuel, residues are priced off what a refiner can avoid making.
Crude oil is a mixture, not a single substance. Heat it and the light bits boil off first, the heavy bits last. That is all the first stage of a refinery does — sort the mixture by weight. Petrol comes off near the top, diesel from the middle, tar from the bottom.
The three ends of the barrel
| Light end | Middle distillates | Heavy end / residue | |
|---|---|---|---|
| Main products | LPG, naphtha, gasoline | Jet fuel, kerosene, diesel, gasoil, heating oil | Residual fuel oil, bunker fuel, bitumen, petroleum coke |
| Who buys it | Passenger cars, petrochemical crackers | Trucks, trains, ships, aircraft, farms, mines, generators, home heating | Ships, power generation in a few markets, road construction |
| Benchmark quotes | RBOB (US, $/gal), Eurobob (NWE, $/t), Singapore 92 RON | ULSD / heating oil (US, $/gal), ICE gasoil (NWE, $/t), Singapore 10ppm gasoil | Rotterdam 3.5% HSFO, Singapore 380cst, 0.5% VLSFO |
| Seasonality | Peaks in the northern summer driving season | Peaks in the northern winter for heating; diesel also tracks harvest and freight cycles | Peaks in Middle East summer for power burn; bunkers are steady year-round |
| Economic sensitivity | Consumer discretionary — mileage falls when pump prices rise | The most GDP-sensitive fuel there is. Diesel demand is industrial activity | Regulation-sensitive; IMO sulphur rules reshaped this market overnight in 2020 |
| Structural direction | Long-term decline as vehicles electrify | Hardest to electrify — trucks, ships, aircraft, farm equipment. Structurally resilient | Shrinking; refiners invest to make less of it |
| Typical crack | Positive, volatile, strongly seasonal | Positive, usually the widest of the three | Usually negative — residue sells below the crude it came from |
That last row surprises people. A negative crack is not an error. Heavy fuel oil routinely trades at a discount to crude, because the refiner would rather not have made it — and the discount is precisely what makes it worth spending hundreds of millions on a coker or a hydrocracker to convert the bottom of the barrel into diesel instead. The heavy-end discount is the return on refinery complexity.
The barrel splits into three commercial families. The light end runs cars, the middle runs trucks, ships, planes and farms, and the heavy leftovers run ships' engines or become road tar.
The middle is the valuable part, because it is the hardest to replace with electricity. The heavy leftovers are usually worth less than the crude they came from — which is exactly why refiners spend fortunes on machines that convert the bottom of the barrel into the middle.
How a crack spread is actually quoted
The unit problem comes first. Crude trades in dollars per barrel. US products trade in dollars per gallon. European products trade in dollars per tonne. So every crack calculation starts with a conversion.
- US gallons to barrels: multiply by 42.
- European tonnes to barrels: divide by roughly 7.45 for gasoil, 8.45 for gasoline, 6.35 for fuel oil. These are density factors and they differ by product.
Once everything is in dollars per barrel, a single crack is simply the product price minus the crude price.
Refiners, though, do not make one product. They make a slate. So the market quotes composite cracks that approximate a real refinery’s output:
3-2-1
Three barrels of crude produce two of gasoline and one of distillate. The US benchmark, because it roughly matches an American refinery's gasoline-heavy slate.
(2×gasoline + 1×distillate − 3×crude) ÷ 3
5-3-2
Five barrels give three of gasoline and two of distillate. A slightly more distillate-weighted version, used by the Dallas Fed among others.
(3×gasoline + 2×distillate − 5×crude) ÷ 5
6-3-2-1
The European structure: six barrels of Brent give three gasoline, two gasoil and one fuel oil. It includes the heavy end, because European refineries are less complex.
(3×gaso + 2×gasoil + 1×HSFO − 6×Brent) ÷ 6
The arithmetic, worked
Take mid-2026 levels: Brent at USD 92 a barrel, gasoline at USD 3.57 a gallon and diesel at USD 4.17 a gallon.
That is the number that made refining shares double in 2026. On the same inputs the 3-2-1 works out at USD 66.33 a barrel — the structures give similar answers, which is why arguing about which one to use is usually a waste of time.
Individual cracks across the barrel
Product price less Brent, USD per barrel, indicative mid-2026 levels
The shape is what matters, not the exact levels. Middle distillates lead, gasoline follows, the residue trades at a discount. In a weak year all four bars are shorter but the ordering is usually the same.
A crack spread is a refinery's gross margin: what it sells the fuels for, minus what it paid for the crude. Because a refinery makes several products at once, the market quotes a blended version — three barrels in, two petrol and one diesel out. In mid-2026 that margin was about USD 68 a barrel, roughly four times normal.
Why product supply and demand is a different market
This is the part worth slowing down for, because it is where most of the analytical value sits.
Crude supply and demand is a story about wells and cargoes: how much OPEC+ produces, how fast US shale grows, how much sits in floating storage, whether a strait is open. Product supply and demand is a story about plants and end uses: how much refining capacity exists and where, what configuration it has, what is down for maintenance, and how much of each specific fuel people are burning this month.
The refinery is the bridge between the two, and the crack spread is the toll charged for crossing it. When the bridge is congested, the toll goes up — no matter what is happening on either bank.
Four states of the world
What happens to the crack, depending on which side is tighter
2026 sat in the bottom-right box even though crude was also tight — because the relative loss of refining capacity exceeded the relative loss of crude supply. The crack responds to the difference, not to either level.
Crude and fuel are two different markets with a factory sitting between them. Crude prices are about wells and tankers. Fuel prices are about refineries and what people are actually burning. The crack spread is the toll for crossing between the two — and when the bridge is congested, the toll rises whatever is happening on either bank.
What 2026 is telling us
The mechanics above explain an otherwise strange year. Crude was expensive — Brent averaged around USD 94 in the second quarter — and yet refining margins still went to records. Both happened for the same reason and to different degrees.
On the crude side, roughly 8.3 million barrels a day of Gulf production was shut in, and the IEA expects global supply to fall 4.3 mb/d across 2026.
On the product side, the losses were proportionately larger:
- Ukrainian strikes cut Russian refinery runs to their lowest in two decades.
- Gulf refining capacity was disrupted alongside crude production.
- Seven US refineries have closed permanently since 2019, removing about 1.2 million barrels a day of capacity that simply is not there to be restarted.
- Together, permanent closures and conflict damage removed an estimated 4.5 million barrels a day — 5.4% of world capacity — in the second quarter alone.
Meanwhile the refineries still standing ran flat out: US utilisation above 95%, US Gulf Coast above 97%. There was no spare capacity left to bid for. Global crude throughput in July was 80.9 mb/d, nearly 5 mb/d below a year earlier. Inventories drained — US petroleum stocks fell 178 million barrels from the start of the year to 1.53 billion, and OECD stocks hit ten-year lows.
That is what a record crack is: not a signal that oil is expensive, but a signal that the ability to turn oil into fuel is scarce.
There was one more source of demand that most coverage missed. With Asian LNG trading around USD 22 per MMBtu — roughly double the energy-equivalent price of fuel oil — industrial and power buyers across Asia switched out of gas and into oil products and coal. That switched demand did not vanish; it arrived at the door of an already overloaded product market. The gas crisis and the refining squeeze were not two separate stories. See The Atlantic–Pacific LNG Arbitrage for the switching arithmetic that pushed those buyers across.
The consumer bore it. US retail diesel reached USD 5.64 a gallon by mid-April, 62% above early January; gasoline hit USD 4.25, up 45%. Diesel led, as it usually does when the constraint is refining rather than crude, because diesel is the hardest cut to substitute away from.
Record refining profits were not a sign that oil was expensive. They were a sign that the ability to turn oil into usable fuel had become scarce. War damaged refineries, and seven American ones had already shut for good since 2019 — and a demolished refinery does not come back. The plants still standing ran flat out, and diesel led the way up, because diesel is the fuel nobody can substitute away from.
How cracks are used
A refiner hedging its margin
- The exposure
- Buys crude today, sells products in four to eight weeks. The margin is at risk over that window.
- The hedge
- Sell the crack: buy crude futures, sell gasoline and distillate futures in the 3-2-1 or 5-3-2 ratio. Locks in a known margin.
- What is left
- Basis risk — the refinery's actual yield is never exactly 3-2-1, its crude is not exactly the benchmark grade, and its location differential moves independently.
A consumer hedging a fuel bill
- The exposure
- An airline, a shipping line or a trucking fleet is short the product, not short crude.
- The hedge
- Hedging with crude alone leaves the crack unhedged — which is exactly the exposure that hurt fuel buyers in 2026, when the product moved far more than the crude.
- The lesson
- Hedge the thing you actually buy. Crude is a convenient, liquid proxy right up until the moment the crack does the moving.
If you buy diesel, hedging with crude oil is not really hedging. In 2026 the crude price rose, but the diesel price rose far more — and the gap between them was where all the pain sat. A crude hedge does not cover that gap. Airlines, shipping lines and trucking fleets learned this the expensive way.
What to watch from here
The EIA has Brent averaging USD 87 in 2026 and USD 69 in 2027 as Gulf barrels return. If that happens, the crude side normalises. Whether the crack normalises with it depends on something slower-moving: capacity.
Conflict damage can be repaired. Refinery closures cannot be undone — closed refineries are dismantled, and permitting new ones in OECD markets is close to impossible. So the honest read is that the crude spike of 2026 is likely temporary while some part of the structural refining tightness is not.
Three indicators are worth watching, in order:
- Refinery utilisation. Sustained readings above 93–95% mean no cushion. It is the tightest single indicator of a coming margin spike.
- Days of forward cover on distillate stocks, rather than absolute inventory. Ten days versus twenty days of demand cover matters more than the headline barrel count.
- The diesel–gasoline differential. When diesel pulls away from gasoline, the constraint is industrial and structural. When gasoline leads, it is usually seasonal and will fade.
The oil price spike will probably fade as Gulf production comes back. The refining squeeze is more stubborn, because you can repair a damaged refinery but you cannot un-demolish a closed one. Watch how hard refineries are running: once they are above about 95%, there is no cushion left, and the next outage goes straight through to the price at the pump.
Sources and further reading
- Oil Market Report, August 2026 — International Energy Agency
- Energy Indicators, August 2026 — Federal Reserve Bank of Dallas
- Energy Indicators, May 2026 — Federal Reserve Bank of Dallas
- Short-Term Energy Outlook, August 2026 — U.S. Energy Information Administration
- Refinery capacity and utilisation data — U.S. Energy Information Administration