Mid-Cycle EBITDA: Valuing Refiners Through the Cycle

LTM EBITDA Fails at Crack Extremes

The same refinery can look cheap and expensive on the same multiple inside a single year, without anything about the business changing. That is a denominator problem, not a valuation signal.

A refiner buys crude and sells fuels, so its earnings track the gap between the two. The Gulf Coast 3-2-1 crack is the standard measure of that gap: the margin on three barrels of crude turned into two barrels of gasoline and one of diesel, quoted per barrel of crude run. When the crack is wide, last-twelve-months (LTM) EBITDA swells and EV/EBITDA prints cheap. When it compresses, trailing EBITDA collapses and the identical stock screens expensive.

FY2025 shows how far apart the ends of that swing sit. Marathon Petroleum filed $16.87/bbl refining and marketing (R&M) margin. PBF Energy filed $7.72/bbl ($8.77/bbl before special items) on 832.9 mb/d of throughput (thousand barrels a day, the industry’s unit for refinery scale), in a year hit by a fire at its Martinez plant.

Mid-cycle EBITDA replaces the trailing denominator with a normalised one: a planning crack, a through-cycle utilisation rate, and the costs that stand between a margin and an earnings line. Capture rate explains how much of an indicator crack a given system actually keeps; mid-cycle EBITDA is what you value once that is settled.

Planning Inputs (June 2026)

Input	Value	Where it comes from
USGC 3-2-1 planning crack	$25/bbl	Set below the spot indicator, because cracks mean-revert
Utilisation default	92%	The EIA CY2025 US average of 92.0%
RIN cost, where a filer gives no split	$3.50/bbl	Below both of Valero’s recent prints, $3.75/bbl in FY2024 and $5.85/bbl in FY2025
EV/EBITDA band	~5–7×	Working convention, not a filed standard
EV/EBITDA anchor	6.0×	Midpoint of that band

The crack assumption is the one that carries the whole build, so it is worth knowing why it sits where it does. Spot Gulf Coast 3-2-1 was around $36.70/bbl in June 2026 and the futures strip implied nearer $46. The planning mark is set roughly a third below spot because refining margins revert faster and harder than almost any other commodity margin: new capacity, restarted units and demand response all bite within a year or two. Anchoring a ten-year valuation to a wide screen crack is the single most reliable way to overpay for a refiner. When spot cracks fall below the planning mark, the same logic runs in reverse and the mark is the generous one.

Utilisation is less contested. US operable refining capacity was 18,424 mb/d as of 1 January 2025 and the system ran at 92.0% across CY2025. The default matches that national average rather than any one filer’s rate, since the good operators beat it (MPC 94%, PSX 94%, VLO an implied ~93.4%) and a normalised build should not assume you own a good operator.

From Crack to Earnings per Barrel

The commonest mistake in a mid-cycle build is to run the indicator crack straight through as if it were earnings. It is not. A $25/bbl crack does not become $25/bbl of EBITDA, or anything close.

Two things stand in the way. The first is that no refinery earns the indicator. The Gulf Coast 3-2-1 prices one region, one crude grade and one yield pattern, and a real system runs none of those exactly. FY2025 filed margins against that $25/bbl mark came in at 67% for MPC, 49% for Valero, 44% for Phillips 66 and 31% for PBF: about half for a large merchant system, and a very wide spread around it. The second is cost. Wages, energy, catalyst, maintenance and head office all sit between a processing margin and an earnings line, and for a big US refiner they run somewhere around $5–7/bbl.

Put those together and the build is short:

Line	$/bbl
Planning USGC 3-2-1 crack	25.00
Realised refining margin, at roughly half the indicator	12.50
Less cash operating cost and corporate overhead	(6.50)
Mid-cycle EBITDA per barrel	6.00

Valero calibrates every row of that. It filed $12.29/bbl refining margin in FY2025, near enough half the planning crack, and $5,273M of adjusted Refining operating income on roughly 1.09 billion throughput barrels, which is $4.84/bbl. The $7.45/bbl gap between the two is operating cost plus depreciation. Operating income is struck after depreciation, so add it back and Valero’s refining EBITDA per barrel was comfortably above the $6.00 used below: the assumption is a plain merchant mark, not a heroic one.

Two cautions on the middle row. Capture that low is measured against a national indicator, and is not the same ratio as a filer’s own reported capture rate, which runs against its internal benchmark and sits far higher. And RIN compliance is already inside that middle row rather than missing from it, because the half-the-crack ratio is measured off filed margins that are themselves struck after compliance. Taking Valero’s $5.85/bbl FY2025 renewable volume obligation cost off the $12.50 would count it twice. Deduct it separately only where the margin you started from was struck before compliance; $3.50/bbl is a reasonable placeholder where a company does not disclose the split.

Worked Example: Illustrative Merchant Refiner

Line	Value
Rated capacity	1,000 mb/d
Utilisation (default)	92%
Throughput	920 mb/d (~335.8M bbl/yr)
Mid-cycle EBITDA per barrel	$6.00
Mid-cycle EBITDA	~$2,015M
Net debt (1.5× mid-cycle EBITDA)	~$3,023M
EV at 6.0× anchor	~$12,090M
Equity (EV − net debt)	~$9,067M
Shares	300M
Implied share price	~$30.22

The per-barrel margin is where the answer actually lives. Hold the multiple and the balance sheet still, and moving it to $5.00/bbl takes the same system to roughly $23.50 a share, while $7.00/bbl takes it to about $37. A sixth off the margin costs more than a fifth of the equity, because the debt does not move with it. That is why the margin build deserves the effort and the multiple deserves an argument, not a decimal place.

EV/EBITDA Anchors

Refiner equity turns on normalised processing margin, so the multiple belongs on mid-cycle EBITDA. The band is ~5–7×, anchored at 6.0×. It sits below where fee-based midstream businesses trade, and for a reason worth holding onto: a pipeline collecting a tariff has a much shorter distance between its good years and its bad ones than a refinery does, so more of its earnings can be capitalised.

The band is a working convention, not a filed standard, and the two ends of it are three turns apart on the same company. On the illustrative refiner above, 5× gives ~$23 a share against 7× at ~$37, the same spread the margin assumption produced, since a multiple and a margin move enterprise value in exactly the same way. Nobody should be arguing about 6.0× versus 6.2×.

The comparison that matters is not two multiples but two denominators. Take one refiner and compute EV/EBITDA on its trailing year, then on mid-cycle EBITDA at the $25/bbl planning crack. In a wide-crack year those two numbers can be several turns apart, and only one of them is telling you about the business. Keep both in the model, clearly labelled, and treat the gap between them as a measure of where the cycle is, not of whether the stock is cheap.

Leverage as the Risk Offset

A multiple tells you what the business is worth at mid-cycle. Leverage tells you whether it gets there. Net debt ÷ mid-cycle EBITDA is the screen, and it uses mid-cycle in the denominator for the same reason the valuation does: measured against a trough year every refiner looks distressed, and against a peak year none of them do.

Net debt ÷ mid-cycle EBITDA	Reading
Below 1.5×	Conservative
1.5–2.5×	Normal
Above 2.5×	Stretched for a merchant refiner

Run the screen and the absolute debt figures stop being informative. Valero carries ~$6.0B of net debt and PBF ~$1,620M, a difference of nearly four times. But at the same illustrative $6.00/bbl, Valero’s 2,988 mb/d system generates around $6.5B of mid-cycle EBITDA and PBF’s 832.9 mb/d around $1.8B, so both land near 0.9×. On this screen they are the same company.

Where they differ is the distance from mid-cycle to trough. A smaller, less complex system with fewer regional outlets loses more of its margin in a bad year, so PBF spends more time a long way below its own normalised number, and its ratio is the one that stops looking comfortable first. That is what “high beta to cracks” means in practice, and no leverage table shows it. Pair the screen with a trough case.

Phillips 66 does not fit the screen at all. Its refining segment produced $2,338M of adjusted EBITDA in FY2025 against $3,773M from midstream, $2,046M from marketing and $845M from chemicals, so refining is under a third of the group and its $18.6B of net debt is not a refining ratio. Value an integrated name by parts and mid-cycle the refining piece on its own; a refining-only build understates the equity badly, and a consolidated one silently applies a refining multiple to fee-based midstream earnings that deserve a higher one.

When Mid-Cycle Beats NAV Logic

An oil producer is a wasting asset: the barrels run out, so you value what is left with a NAV or EV/DACF build against a price deck. A refinery is a factory. It has no reserves to deplete and, maintained, it processes barrels indefinitely. What it owns is a margin, and a margin is a thing you normalise rather than deplete. That is why the two ends of the oil chain need different valuation machinery, and why mid-cycle earnings do for a refiner what a reserve report does for a producer.

The whole sequence, then: set the planning crack at $25/bbl and utilisation at 92%, work down to a realised margin per barrel using capture rate, take out cash operating cost and RIN compliance if the filed margin has not already, multiply by throughput, then apply 5–7× and check the leverage screen against the same denominator.

The result you are looking for is a disagreement. A refiner that screens cheap on its trailing year and merely fair on mid-cycle is not cheap; it is being priced off a wide crack that will not last. The one worth the work is the reverse case.