The Oil & Gas Value Chain, Part 3: Downstream (2026)
Data as of 18 August 2026. The framework — what the three downstream businesses are, how a crack spread is earned and lost, how to read a refiner — is evergreen. Capacities, yields and margins are agency and company figures for their stated year (refinery data lags about a year); margins are period averages, never a day’s print — re-source them at the time of use. Worked examples are hypothetical unless a table names its company and year. Values are in US dollars unless marked. For information only; prepared with AI assistance — see the disclaimer.
Every gallon of gasoline, every tank of jet fuel and most of the plastic in the room around you began as a barrel of crude that was useless until a refinery took it apart and put it back together. Downstream is the last third of the oil and gas chain — refining, fuel marketing and petrochemicals — and the only part that sells the finished product a consumer actually buys. Unlike the producers most energy investors already own, it does not care much where crude prices go; it cares about the gap between what it pays for crude and what it sells fuels for — the whole investment case, and the reason refining is the most violently cyclical business in the hydrocarbon complex. For the company-level data behind the charts — every refiner, marketer and petrochemical producer screened by capacity, complexity and asset base — go to Metal Pilot’s downstream oil & gas stock screener.
The headline: the reader’s job is to tell a durable margin advantage from a temporary one.
The Oil & Gas Value Chain series. Part 1: Upstream · Part 2: Midstream · Part 3: Downstream · Part 4: Oilfield Services · Companion: World Oil & Gas Basins · Supplement: Oil Sands
TL;DR & Key Takeaways
- What it is: the conversion-and-sale end of oil and gas — crude refining, fuel marketing and retail, and petrochemicals. Its profit is the processing margin between crude bought and products sold.
- How it earns: on the crack spread — the margin between crude and the fuels made from it. Unlike midstream’s fee, it is uncontracted and mean-reverting, and per barrel it is thin: operating costs and depreciation absorb much of the realised margin before any profit (Section 2.6).
- The scale: global refining capacity is about 103.8 MMbbl/d; the United States runs 18.2 MMbbl/d across just 130 plants — down from 301 refineries in 1982 at roughly flat capacity.
- The scarcity signal: shortage shows up as a margin spike, not a crude shortage. When capacity is tight, the 3-2-1 crack blows out — New York Harbor diesel cracks averaged about $89/bbl in October 2022; when it is loose, refiners lose money.
- The equity behaviour: refiners amplify the margin cycle, not the oil price — merchant refiners went from losses in 2020 to record free cash flow in 2022, with shares that ran far harder than crude (Section 4.3).
- The biggest swing factor: the geography of capacity and the peak in demand. Capacity is moving east while the West closes plants, and the IEA expects refined-product demand to peak around 2027, leaving petrochemicals as the only real growth.
Figure 1. Downstream oil & gas at a glance
Figure data: U.S. EIA, the Energy Institute, IEA Oil 2025 and Wood Mackenzie; see Sections 1.7, 2.1–2.5, 3.2 and 5.
How to read this guide: new to the sector? Read straight through from Section 1. Here for the investment case? Jump to Section 4 — how to own it, how to screen it, and why the margin is both the attraction and the trap. Every technical term is defined in the Terms list at the end.
1. Downstream & the market basics
1.1 What downstream is
Downstream is the final third of the hydrocarbon value chain. Upstream finds and produces crude and gas; midstream moves, processes and stores it; downstream converts it into finished products and sells them to the end user. Where midstream sells a service and earns a fee, downstream sells a manufactured good and earns a processing margin — the difference between the cost of its raw material and the price of its output. That single distinction explains why downstream behaves like heavy manufacturing rather than like a commodity producer or a toll road.
Downstream is really three related businesses, and a company can own one, two or all three:
- Refining — the core. Refineries take crude oil and, through distillation, conversion and treating, split and rebuild it into gasoline, diesel, jet fuel, fuel oil and petrochemical feedstocks. This is the capital-intensive, cyclical heart of the sector.
- Marketing & distribution — moving finished products from the refinery gate to the customer through terminals, wholesale supply, branded fuel stations and convenience retail. This is a lower-margin, higher-stability business that monetises brand and logistics rather than conversion.
- Petrochemicals — steam crackers and downstream chemical units that turn refinery streams and natural gas liquids (NGLs) into ethylene, propylene and the building blocks of plastics, fibres and fertilisers. Increasingly the growth engine of the whole sector, and the reason a refinery is worth more when it is integrated with a chemical plant.
Where the boundary sits is a real question. The wellhead, gathering and pipeline belong to upstream and midstream; downstream begins where a molecule is transformed for sale. This guide starts at the refinery inlet and ends at the fuel pump and the plastics pellet.
Figure 2. The downstream value chain, crude to end user
Source: standard oil & gas value-chain structure (upstream–midstream–downstream); conceptual diagram.
The sub-sectors are worth naming precisely, because a downstream company is best understood as a portfolio of them, and each has its own scale metric, its own margin driver and its own risk profile. The units in the table — MMbbl/d, Mt/yr, ¢/gal — are set out in Section 1.2.
Table 1. Downstream sub-sectors and their headline metrics
| Sub-sector | Stage | What it does | Headline metric |
|---|---|---|---|
| Refining | Conversion | Splits and rebuilds crude into fuels and feedstock | Capacity (MMbbl/d); Nelson complexity; utilisation |
| NGL fractionation / feedstock supply | Feed | Supplies ethane, propane and naphtha to crackers | Feedstock cost; frac spread (NGL value less gas cost) |
| Fuel marketing & wholesale | Distribution | Supplies branded and unbranded fuel to resellers | Marketing margin (¢/gal); volume |
| Convenience retail | Distribution | Sells fuel and merchandise at branded stations | Sites; fuel volume; merchandise margin |
| Petrochemicals (olefins) | Conversion | Cracks feedstock into ethylene and propylene | Cracker capacity (Mt/yr); operating rate |
| Petrochemicals (polymers/derivatives) | Conversion | Makes polyethylene, polypropylene and derivatives | Capacity; integration margin |
| Lubricants & specialties | Conversion | Base oils, greases, waxes, asphalt | Base-oil capacity; premium margin |
| Renewable / low-carbon fuels | Conversion | Renewable diesel, sustainable aviation fuel (SAF), biofuels | Capacity; biofuel-credit (RIN) value |
Source: sub-sector taxonomy from the Metal Pilot project model; capacity conventions per U.S. EIA reporting definitions.
1.2 Units & measurement conventions
Downstream inherits oil’s units and adds a few of its own. This report follows the industry’s Roman-numeral convention: M = thousand, MM = million. Crude and product flows are quoted in barrels per day — MMbbl/d (million) and Mbbl/d (thousand). A barrel (bbl) is 42 US gallons (about 159 litres).
The distinction that trips up newcomers most in refining is calendar-day versus stream-day capacity. Barrels per calendar day (b/cd) is what a plant can process averaged over a year, allowing for maintenance and downtime — the figure EIA reports. Barrels per stream day (b/sd) is the maximum when running flat out, typically about 6% higher. A refinery quoted at 631 Mbbl/cd and 665 Mbbl/sd is the same plant measured two ways, and confusing them overstates capacity.
Petrochemicals switch units entirely: cracker and polymer capacity is measured in million tonnes per annum (Mt/yr) of product (ethylene, propylene, polyethylene), not barrels. And margins are quoted in dollars per barrel for refining (the crack spread) and cents per gallon for marketing.
Table 2. Units used in this report
| Unit | Meaning | Typical magnitude | Conversion |
|---|---|---|---|
| MMbbl/d | Million barrels per day (flow) | US refining capacity: 18.2 | 1,000 Mbbl/d = 1 MMbbl/d |
| Mbbl/d or b/cd | Thousand barrels per (calendar) day | A single refinery: 100–650 | Stream day ≈ 6% higher |
| $/bbl | Dollars per barrel (refining margin) | 3-2-1 crack: ~$17 long-run average; $5–40 range | The crack-spread unit |
| ¢/gal | Cents per gallon (marketing margin) | Retail fuel margin: ~35–40 (2020–25) | 42 gal = 1 bbl |
| MMBtu | Million British thermal units (energy) | Valero’s 2025 gas price: $3.04/MMBtu | The natural-gas price unit |
| NCI | Nelson Complexity Index (dimensionless) | ~1–2 (topping) to 10+ (deep conversion) | Higher = more conversion |
| Mt/yr | Million tonnes per annum (petchem flow) | A world-scale cracker: 1–2 | — |
| RIN | Renewable Identification Number (credit) | One per gallon of biofuel | Compliance currency |
| bbl | Barrel of crude or product (stock/flow) | 1 bbl = 42 US gal | ≈ 159 litres |
Source: unit conventions per U.S. EIA and standard refining/petrochemical practice; magnitudes as cited in Sections 1.5 and 2–3; retail fuel margin per NACS (OPIS data).
1.3 Inside the refinery
Crude oil is a mixture of thousands of hydrocarbons, and it is worthless until it is separated into specific products. A refinery does that in a fixed sequence of units grouped into five stages — Figure 3 is the overview — and the value it captures comes from how completely it can turn the cheap, heavy bottom of the barrel into expensive, light products.
Figure 3. Inside the refinery — the processing sequence
Source: standard refinery process sequence per U.S. EIA refining primers; conceptual diagram.
Table 3 opens each stage into its units, in the order the barrel meets them: what each unit is fed, what it physically does, what it sends on, and the lowest rung of the configuration ladder (Section 1.5) on which it normally appears.
Table 3. How a barrel is refined — the units in processing order
| Unit | Feed in | What happens | Streams out | Ladder tier |
|---|---|---|---|---|
| Desalting | ||||
| Desalter | Raw crude, wash water | Heated to roughly 65–175 °C and passed with wash water through a high-voltage electrostatic field; the water settles out, taking the salts and sediment that would corrode the plant. | Desalted crude; brine to effluent treatment (Table 7) | All |
| Distillation | ||||
| Atmospheric distillation (CDU) | Desalted crude | Heated to roughly 340–370 °C and flashed into a column near atmospheric pressure; each cut condenses at its own boiling range. Nothing is chemically changed. | Gas and LPG, naphtha, kerosene/jet, gasoil (diesel), atmospheric residue | All |
| Vacuum distillation (VDU) | Atmospheric residue | Re-distilled under reduced pressure, so the heavy molecules boil off before they get hot enough to crack. | Vacuum gasoil (VGO), vacuum residue | Cracking |
| Conversion | ||||
| Fluid catalytic cracking (FCC) | VGO; some residue in a residue FCC | A fluidised catalyst at roughly 450–510 °C breaks long molecules into shorter ones; the coke it lays down is burned off in a regenerator that also heats the unit. | FCC gasoline, light cycle oil, propylene and butylene, slurry oil | Cracking |
| Hydrocracker | VGO, cycle oils | Cracking over a catalyst in hydrogen at roughly 70–140 bar breaks and saturates the molecules in one step, so the products come out low in sulphur. | Diesel, jet and naphtha; light gases | Cracking |
| Delayed coker | Vacuum residue | Thermal cracking with no catalyst: heated to roughly 480–510 °C, the residue soaks in drums for about a day until it splits into liquids and solid coke. A visbreaker is the milder alternative. | Coker naphtha and gasoils (to hydrotreating), gas, petroleum coke (Table 7) | Deep conversion |
| Catalytic reformer | Heavy naphtha, hydrotreated first | A platinum catalyst in a train of furnaces and fixed-bed reactors rearranges the naphtha into aromatics, raising its octane sharply. | Reformate (high-octane blendstock), hydrogen, LPG | Hydroskimming |
| Isomerisation | Light naphtha, normal butane | A platinum or aluminium-chloride catalyst at roughly 110–170 °C turns straight-chain molecules into branched, higher-octane ones. | Isomerate; isobutane for alkylation | Hydroskimming (optional) |
| Alkylation | Isobutane; propylene and butylene from the FCC | A sulphuric- or hydrofluoric-acid catalyst joins the light olefins to isobutane, building larger branched molecules. | Alkylate (high-octane blendstock); spent acid to regeneration (Table 7) | Cracking |
| Treating | ||||
| Hydrotreaters | Naphtha, kerosene, diesel, FCC feed, coker streams | Hydrogen over a catalyst at up to about 70 bar turns sulphur into H₂S and nitrogen into ammonia, bringing each stream to specification — 15 ppm sulphur for ULSD, a 10 ppm average for Tier 3 gasoline. | Treated streams to blending; H₂S and sour water to the sulphur plant | Hydroskimming |
| Hydrogen and sulphur plants | Natural gas; H₂S from the hydrotreaters | A steam methane reformer tops up the hydrogen the reformer makes; an amine unit captures the H₂S and a Claus unit turns it into sulphur. Both are sized to the slate’s sulphur (Section 1.5). | Hydrogen, elemental sulphur, CO₂ (Table 7) | Hydroskimming |
| Blending | ||||
| Gasoline blending | Reformate, alkylate, isomerate, FCC gasoline, butane | Blended to octane, vapour-pressure and sulphur specification; in the US shipped as RBOB or CBOB blendstock and finished with ethanol at the terminal. | Finished gasoline grades | All |
| Distillate and jet blending | Hydrotreated kerosene and gasoils, hydrocracker distillate | Diesel is blended to cetane, cold-flow and sulphur specification; jet fuel, a treated kerosene cut, to tighter freeze-point and flash-point limits. | ULSD, heating oil, jet fuel | All |
| Fuel oil and specialty pool | Vacuum residue, slurry oil, cutter stock | Whatever conversion has not consumed is cut to a viscosity specification, or sent to asphalt and lubricant base-oil units. | Residual and marine fuel, asphalt, lubricant base oils | All (shrinks up the ladder) |
Source: unit sequence, feeds and products per U.S. EIA — refining crude oil and the unit-by-unit descriptions and operating conditions in the OSHA Technical Manual, Section IV Chapter 2 — Petroleum Refining Process; sulphur limits per US EPA diesel fuel standards and US EPA Tier 3 fuel standards; ladder tiers follow Table 6. Temperatures and pressures are typical ranges converted from the OSHA figures, not a specific plant, and the order is conceptual — real refineries omit, duplicate or re-sequence units (the naphtha hydrotreater, for one, sits ahead of the reformer).
Separation only sorts. Desalting and distillation return exactly the cuts nature put in the crude — the straight-run yields of Table 5. For a heavy barrel such as Cold Lake Blend that leaves about 61% as vacuum gasoil and residue, far too much heavy, low-value material to sell as it stands.
Conversion is where the money is made. The conversion units rebuild those heavy fractions into gasoline, diesel and jet, and a refinery with plenty of them can run cheap, heavy, sour crude and still make a light-product slate — which is the entire source of its margin advantage.
Treating and blending are the price of the specification. Every stream is desulphurised before it can be sold, and what the treaters strip out becomes the second production line of Section 1.6. What leaves the gate is gasoline, diesel, jet fuel, fuel oil, LPG and petrochemical feedstock — each a family of grades set out in Section 2.2.
The single number that captures how much conversion a refinery has is the Nelson Complexity Index (NCI) — a measure of secondary-processing capacity relative to simple distillation. The scale runs from a plant with almost no conversion capacity up to a modern deep-conversion refinery well into double digits — the configurations behind those numbers, and the crude each one can actually run, are set out in Section 1.5. The US Gulf Coast has some of the world’s most sophisticated refineries (U.S. EIA), which is why they can profitably run discounted heavy crude and out-earn simpler plants elsewhere. Complexity, not size, is the durable competitive moat in refining.
1.4 The crude slate
The barrel a refinery buys is not standard, and the differences are where much of refining margin comes from. Crude is graded on two axes. API gravity measures density: light crude (high API) is rich in the naphtha and middle fractions that make gasoline, diesel and jet, so it is easy to refine and trades at a premium; heavy crude (low API — Canadian oil sands, Venezuelan Merey, Mexican Maya) is thick with residue that a simple plant cannot upgrade, so it trades at a discount. Sulphur splits crude into sweet (low sulphur, cleaner, easier to meet fuel specs) and sour (high sulphur, which must be stripped in hydrotreating and corrodes equipment) — sour barrels are discounted too. The oil guide covers these quality axes from the producer’s side; what matters downstream is the cost of turning each grade into finished product.
The economics fall straight out of those two axes, and they are the reason complexity pays. A simple refinery — little beyond distillation, the hydroskimming configuration of Section 1.5 — can only run light-sweet crude profitably, so it pays the premium and earns a thin margin. A complex, deep-conversion refinery buys cheap heavy-sour crude and uses coking, hydrocracking and hydrotreating to make the same premium light products, capturing the discount as margin. That light–heavy gap is the single largest prize in refining, and the reason US Gulf Coast refiners spent decades and billions building the world’s most complex fleet to run discounted heavy imports.
Processing a heavier, more sour barrel genuinely costs more, and the cost is concrete: it consumes far more hydrogen (made on-site from natural gas, so heavy-crude economics are partly a gas-price bet), more energy and catalyst, and more maintenance, and it leaves more low-value residue and petroleum coke. But when the light–heavy differential is wide, that extra cost is dwarfed by the feedstock saving — which is why a complex refiner’s margin tracks crude differentials (WCS versus WTI, Maya versus the Gulf benchmark) as closely as it tracks the crack spread. When those differentials compress — heavy supply tightens, or light shale floods in — the complex refiner’s advantage narrows, one of the sector’s quieter risks (Section 4.5).
Table 4. Crude grades and their refining economics
| Crude grade | API gravity | Sulphur | Example | Price vs benchmark | Refinery needed |
|---|---|---|---|---|---|
| Light sweet | High (>38°) | Low (<0.5%) | WTI, Brent, Bonny Light | Premium | Simple (hydroskimming) |
| Light sour | High | Moderate | Arab Extra Light | Slight discount | Hydrotreating |
| Medium sour | Medium (25–35°) | Medium–high | Arab Light, Urals | Discount | Moderate–complex |
| Heavy sour | Low (<25°) | High (>1%) | WCS, Maya, Merey | Steep discount | Complex (coking/hydrocracking) |
Source: crude-quality conventions per U.S. EIA — crude oils have different quality characteristics and standard assay practice; API/sulphur bands and differentials are indicative and vary by cargo and market.
One slice of that heavy-barrel conversion happens before a refinery ever sees the crude. Canadian upgraders — coking and hydroconversion plants bolted onto oil-sands operations — turn bitumen into synthetic crude oil that arrives at the refinery gate as a light, sweet feed: Syncrude Sweet Premium averages 33.0° API and 0.19% sulphur, lighter and sweeter than WTI. Every upgraded barrel is therefore one taken out of the discounted heavy pool that complex refiners compete for — and with roughly 48% of Alberta’s raw bitumen upgraded in 2024, a share its regulator expects to fall toward 44% by 2034, the balance arrives as diluted bitumen looking for coking capacity, most of it on the US Gulf Coast. How an upgrader works and why a producer builds one is covered in the upstream guide.
What density does to the barrel. API gravity is not just a price label — it decides what comes out of the distillation column before a single conversion unit touches the crude. A crude assay splits a barrel into boiling-range cuts, and the straight-run yield — what distillation alone recovers — shows the gap plainly. Table 5 takes one crude from each of four basins in the World Oil & Gas Basins guide, and its two ends are the two sides of the WCS–WTI spread. WTI Light, the light end of the Texas WTI stream the Permian dominates, arrives four-fifths light and middle product; Cold Lake Blend, diluted bitumen from the Alberta oil sands, delivers under two-fifths, with almost a third of the barrel left as vacuum residue, the tar-like bottom that boils above 550 °C. Between those two sits the vacuum gasoil (VGO), too heavy for fuel as it is but the main feed a cracking unit turns into gasoline and diesel.
Sulphur is a separate axis. The middle pair holds density roughly still and moves only the second axis. Guyana’s Liza and Abu Dhabi’s Upper Zakum — an Arabian Basin barrel of the same medium sour type as Ghawar’s Arab Light — distil almost identically, 58% and 60% light and middle product with about a sixth left as residue, yet Upper Zakum carries 3.5 times the sulphur. A refinery pays for that difference in hydrotreating, hydrogen and sulphur recovery, not in conversion capacity. Metals follow neither axis: sweet Liza carries nearly twice Upper Zakum’s nickel and vanadium.
Table 5. Straight-run yields of four crudes, light sweet to heavy sour
| Cut | Becomes, after processing | WTI Light | Liza | Upper Zakum | Cold Lake Blend |
|---|---|---|---|---|---|
| Origin | — | Texas — WTI stream, Permian-led | Guyana — Stabroek, deepwater | Abu Dhabi — Arabian Basin, offshore | Alberta oil sands — diluted bitumen |
| API gravity | — | 47.4° | 31.9° | 33.4° | 19.5° |
| Sulphur | — | 0.07% | 0.59% | 2.09% | 3.87% |
| Nickel + vanadium | — | 0.2 ppm | 39.1 ppm | 21.3 ppm | 201.0 ppm |
| Butane and lighter | LPG, refinery fuel gas | 3.2% | 1.6% | 3.1% | 2.4% |
| Naphtha (C5–150 °C) | Gasoline blendstock, reformer and cracker feed | 34.9% | 18.4% | 18.3% | 12.9% |
| Middle distillates (150–370 °C) | Jet fuel, kerosene, diesel, heating oil | 41.9% | 38.1% | 38.1% | 23.7% |
| Vacuum gasoil (370–550 °C) | Feed for the FCC or hydrocracker | 15.9% | 26.7% | 24.0% | 29.4% |
| Vacuum residue (550 °C+) | Coker feed, asphalt, residual fuel oil | 4.1% | 15.2% | 16.5% | 31.6% |
| Light and middle cuts combined | Product that needs no conversion | 80.0% | 58.1% | 59.5% | 39.0% |
Notes to Table 5
- Each cut is named by its boiling range. The four crudes run from light sweet to heavy sour: WTI Light light sweet, Liza medium sweet, Upper Zakum medium sour and Cold Lake Blend heavy sour.
Source: ExxonMobil crude assay summary reports — WTI Light – Export (February 2026), Liza (May 2024), Upper Zakum (January 2026) and Cold Lake Blend (October 2023). Yields are % by volume of whole crude, grouped from the assays’ atmospheric and vacuum cuts; butane and lighter is taken by difference. ExxonMobil gives WTI Light’s origin only as Texas: it is the light end of the WTI stream, above the 40–44° API band of the WTI Midland specification, and WTI is produced mainly in the Permian (S&P Global). Cold Lake Blend is diluted bitumen, so part of its naphtha is blended-in diluent rather than native to the bitumen. Basin settings per the World Oil & Gas Basins guide. An assay is a single-sample snapshot; individual cargoes vary.
Figure 4. What distillation alone gets from four crudes, light sweet to heavy sour
Figure data: Table 5.
What conversion does with it. Those straight-run numbers are the starting point, not the result: the conversion units of Section 1.3 turn most of the vacuum gasoil into gasoline and diesel and, in a deep-conversion plant, coke the residue too, so even the Cold Lake barrel ends up overwhelmingly light product. What the table adds is the cost of getting there — it scales with density and sulphur, in conversion capacity, hydrogen, and the coke and sulphur left behind (Section 1.6) — and the nickel-and-vanadium row, because those metals concentrate in the residue and poison cracking catalyst.
Density also changes the volume that comes out. The products are less dense than the crude they are made from, and hydrocracking adds hydrogen to the molecules, so a refinery sells more barrels than it buys: US refineries recorded a processing gain of 5.9% in 2025 — about 44.5 gallons of product from every 42-gallon barrel of input. Because the gain comes from that density gap, a heavy barrel that is converted deeply produces more of it than a light barrel that barely needs converting.
1.5 Refinery types
A refinery is not a general-purpose machine, and this is the point most often missed about the sector. Two plants of identical nameplate capacity can be entirely different businesses, because what a refinery can profitably do with a barrel is fixed by the units standing behind the distillation column — and those were chosen, at vast expense, for a particular kind of crude. The industry recognises four standard configurations, and the Nelson Complexity Index is the one-number summary of where a plant sits on that ladder.
Table 6. Refinery configurations and the crude each can run
| Configuration | NCI | Units beyond distillation | Crude it runs economically | Fuel-oil tail | Where it survives |
|---|---|---|---|---|---|
| Topping | ~1–2 | None — atmospheric distillation only | Light sweet | 40–50% | Remote markets; asphalt and specialty plants |
| Hydroskimming | ~2–4 | Catalytic reforming, hydrotreating | Light sweet to light sour | 30–40% | Shrinking — the configuration most closures come from |
| Cracking | ~5–9 | Vacuum distillation, FCC or hydrocracker | Medium sour | 10–15% | The typical inland US and European refinery |
| Deep conversion | ~10+ | Delayed coker or residue hydrocracker | Heavy sour | Under 5% | US Gulf Coast; the Asian and Middle East export complexes |
Source: configuration conventions per U.S. EIA — refining crude oil and the downstream charge-capacity categories reported in the U.S. EIA Refinery Capacity Report; complexity bands and fuel-oil shares are indicative and vary plant by plant.
Three kinds of plant sit off that ladder because they were built for a different job. Condensate splitters run ultra-light feed that a conventional crude unit is not designed to handle, and are common in the Middle East and on the US Gulf Coast. Lubricant and asphalt refineries deliberately buy heavy crude, because there the residue is the product rather than the problem. And crude-to-chemicals complexes — the Jamnagar model in Section 3.1 — are engineered to send a far larger share of the barrel to a steam cracker than the 5–15% a conventional refinery yields.
The harder point is that complexity buys the ability to run discounted crude, not any crude. Every refinery works within a designed crude envelope — an assay band it was engineered for — and four constraints hold it there:
- Hydraulics. Every column and unit is sized for a particular distribution of cuts. Feed a coking refinery light shale crude and the naphtha overhead floods while the coker and the FCC run starved — which is why Gulf Coast refiners kept buying heavy imports right through the shale boom instead of simply switching to cheap domestic light crude.
- The sulphur and hydrogen ceiling. How sour a plant can go is usually capped not by its coker but by its sulphur-recovery and hydrogen capacity, both sized years earlier for a different slate.
- Acid and metals. High total acid number (TAN) crudes corrode carbon steel through naphthenic acid attack, so a plant without upgraded metallurgy carries a hard TAN limit; nickel and vanadium in heavy barrels poison FCC catalyst.
- Blend compatibility. Mixing incompatible crudes drops asphaltenes — the heaviest, tar-like molecules in crude — out of solution and fouls heat exchangers, so even two permitted grades cannot always be run together.
In practice a refinery therefore runs a crude basket inside that envelope, re-optimised with a linear-programming (LP) model — the optimiser that picks the most profitable crude mix and product slate the plant’s units allow — as differentials move, and trials an unfamiliar grade in small percentages before committing to a cargo.
None of which makes the top of the ladder unconditionally better. A deep-conversion refinery running light sweet crude is wasting the billions it spent on conversion units and can earn less on that barrel than a simple plant would — the configuration only pays while the light–heavy differential is wide enough to fund it, which is the risk developed in Section 4.5. Configuration also decides what a region has to trade: the US fleet is weighted toward catalytic cracking and therefore gasoline; Europe’s leans more on hydrocracking for diesel, yet its diesel-heavy demand still outruns it, so Europe exports surplus gasoline and imports diesel — the trade pattern in Section 2.1.
1.6 Byproducts
Crude is not pure hydrocarbon. Every barrel also carries sulphur, nitrogen, organic acids, metals (nickel and vanadium above all), salt and water, and fuel specifications leave almost no room for them — US diesel may hold no more than 15 ppm of sulphur, against several percent in a heavy sour crude. So a refinery runs a second, less visible production line that carries the contaminants out, partly sold, partly burned on site and partly pure disposal cost. It matters to the margin as much as to the environment: coke, sulphur and most other byproducts sell for far less per barrel than crude, so every percent of the barrel that leaves as one pulls realised margin below the benchmark crack (Section 2.6), and the capacity to handle them often caps which crude a plant can buy.
The sulphur chain. Hydrotreaters and hydrocrackers react the sulphur in each stream with hydrogen, turning it into hydrogen sulphide (H₂S) gas; an amine unit scrubs that gas out, and a sulphur recovery unit running the Claus process converts it to liquid elemental sulphur — 92–97% of it on its own, 99% or more with tail-gas treatment. The tonnage is large: the United States produced about 8.1 Mt of sulphur in 2025 — 7.6 Mt of it recovered as elemental sulphur, with refineries the largest source — out of about 84 Mt of world output.
Where sulphuric acid comes in. Most recovered sulphur ends up as acid in someone else’s plant — about 90% of all sulphur consumed is used as sulphuric acid, above all to make phosphate fertiliser — and a few refineries skip the solid step with a wet sulphuric acid unit that makes commercial acid directly (Table 7). Refiners are also acid consumers: sulphuric acid is the catalyst in alkylation, the unit that joins light gases into high-octane gasoline blendstock, and the spent acid goes back to regeneration plants.
How much sulphur a barrel carries — a hypothetical illustration. A hypothetical 200 Mbbl/d refinery running only Cold Lake Blend (3.87% sulphur, about 149 kg a barrel) takes in roughly 5.8 kg of sulphur per barrel — some 1,150 tonnes a day. The same plant running WTI Light (0.07%) would take in under 0.1 kg a barrel — some sixty-five times less. That is the physical meaning of “sour”, and why a plant’s sulphur-recovery capacity, not its coker, so often caps how heavy a slate it can buy (Section 1.5).
Carbon, gas, hydrogen and metals. The coker leaves petroleum coke — 4.6% of US refinery output in 2025 — sold by grade into cement kilns and power plants, aluminium-smelter anodes and the graphite electrodes of electric-arc steel furnaces. Still gas, the methane, ethane and hydrogen every unit throws off, was another 3.9% and is mostly burned in the refinery’s own furnaces. Complex plants make or buy extra hydrogen from a steam methane reformer (SMR), whose CO₂ is vented or, at a few sites such as Valero’s Port Arthur, captured. The nickel and vanadium in heavy crude end up in the coke, fuel-oil ash and spent catalyst, where they become a small secondary metal supply.
Table 7. Refinery byproducts — where they come from and where they go
| Byproduct | Removed in / made by | What it is | Where it goes | Scale or benchmark |
|---|---|---|---|---|
| Elemental sulphur | Hydrotreaters to amine unit to Claus sulphur recovery unit | H₂S converted to liquid or solid sulphur | Sulphuric acid plants to phosphate fertiliser, chemicals, metal leaching | US 8.1 Mt produced (7.6 Mt recovered elemental), world 84 Mt (2025e) |
| Sulphuric acid (made) | Wet sulphuric acid unit, instead of solid sulphur | Commercial 98% H₂SO₄ | Fertiliser, chemicals, mining | Up to 99.9% sulphur removal |
| Spent sulphuric acid (consumed) | Alkylation unit, where acid is the catalyst | Acid diluted with water and organics | Regeneration plants, back as 93–99.2% acid | 2.5–5 Mt/yr reclaimed (Veolia estimate) |
| Petroleum coke | Delayed or fluid coker | Solid carbon from vacuum residue | Fuel grade to cement, power; anode grade to aluminium; needle to graphite electrodes | 4.6% of US refinery output (2025) |
| Still gas | Distillation, crackers, coker | Methane, ethane and hydrogen | Burned as refinery fuel | 3.9% of US refinery output (2025) |
| Hydrogen and CO₂ | Catalytic reformer; steam methane reformer | H₂ for hydrotreating; CO₂ from the SMR | H₂ used on site; CO₂ vented or captured | Port Arthur: 4 Mt CO₂ captured by October 2017 |
| Nickel and vanadium | Concentrate in residue, coke, fuel-oil ash and spent catalyst | Metals from heavy crude | Metal recyclers to vanadium chemicals and alloys | US secondary vanadium ~5,700 t (2023) |
| Process water, caustic and sludge | Desalter, steam stripping (sour water), caustic washes, tank cleaning | Water carrying H₂S and ammonia; brine, used caustic, oily solids | Sour water stripped, its H₂S to the sulphur plant; the rest treated or re-run | A disposal cost, not a revenue |
Source: diesel sulphur limit per US EPA diesel fuel standards; Claus chemistry and recovery efficiencies per US EPA AP-42 §8.13; sulphur tonnage, sources and end use per USGS Mineral Commodity Summaries 2026 — sulfur; WSA per Topsoe; spent-acid regeneration per Veolia; coke and still-gas shares per U.S. EIA refinery yield, coke grades per CRS R43263; Port Arthur capture per US DOE; secondary vanadium per USGS Mineral Commodity Summaries 2024 — vanadium; sour-water composition per Bryan Research & Engineering. The sulphur-per-barrel figure is a hypothetical illustration from the Table 5 assay.
1.7 Margins & the crack spread
Downstream has no single benchmark price the way crude has Brent, because its profit is a spread, not a level. The refiner buys crude at one price and sells a slate of products at another, and the difference — after energy and operating cost — is the margin. The market’s shorthand for it is the crack spread.
The most-quoted version is the 3-2-1 crack spread: the gross margin from turning three barrels of crude into two barrels of gasoline and one barrel of distillate. It is a proxy, not a refiner’s actual margin (real refineries make a wider slate and run more complex configurations), but it captures the sector’s fortunes in one number. Across 2010–2021 the US Gulf Coast 3-2-1 averaged about $17/bbl (author’s arithmetic on EIA spot prices, Table 15); it can fall below $5 in a weak market and spike above $40 in a disruption.
Crucially, this margin is uncontracted and mean-reverting. There is no take-or-pay, no reservation charge, no twenty-year firm contract. A refiner is exposed to the daily balance of refining capacity against product demand, which is why the same asset can earn a fortune one year and lose money the next (the 2020–22 swing is walked in Section 2.5). That is the opposite of midstream’s fee stream, and it is the single most important thing to understand about the sector.
The other two businesses earn differently. Marketing captures a retail and wholesale margin — a few cents per gallon between the refinery-gate product price and the pump price, plus convenience-store merchandise profit that is far steadier than fuel. Petrochemicals capture an integration margin — the spread between cheap feedstock (US ethane, in particular) and the price of ethylene and polymers, which runs on its own cycle driven by global cracker capacity.
Table 8. How the three downstream businesses earn
| Business | What it sells | Margin captured | Who bears price risk | Cyclicality |
|---|---|---|---|---|
| Refining | Gasoline, diesel, jet, fuel oil | Crack spread ($/bbl) | Refiner (fully) | Very high |
| Feedstock supply / fractionation | Ethane, propane, naphtha | Frac spread | Processor | High |
| Fuel wholesale & marketing | Branded/unbranded fuel | Marketing margin (¢/gal) | Shared | Medium |
| Convenience retail | Fuel + merchandise | Fuel margin + merchandise | Retailer | Low–medium |
| Petrochemicals (olefins) | Ethylene, propylene | Integration/cracker margin | Producer (fully) | High |
| Polymers & derivatives | Polyethylene, polypropylene | Chain margin | Producer | High |
| Renewable fuels | Renewable diesel, SAF | Spread + RIN and low-carbon-fuel (LCFS) credits | Producer + policy | Policy-driven |
Source: margin taxonomy from company filings; crack-spread convention per U.S. EIA.
2. The refining network
2.1 Where the capacity is
Refining geography is a record of where fuel demand grew and where governments decided to build. For decades the United States and Europe dominated; today the map has tilted decisively toward Asia and the Middle East. Global refining capacity is about 103.8 MMbbl/d, and the striking fact is that China (about 18.8 MMbbl/d in 2025) has moved past the United States (about 18.2 MMbbl/d) to become the world’s largest refiner — the result of two decades of state-directed build-out, while the US fleet shrank.
Table 9. Refining capacity by country, 2025
| Country | Capacity | Note |
|---|---|---|
| China | ~18.8 MMbbl/d | Now the world’s largest; still adding capacity |
| United States | ~18.2 MMbbl/d | Falling as plants close |
| Russia | ~6.7 MMbbl/d | Nameplate; sanctioned, lower effective utilisation |
| India | ~4.9 MMbbl/d | Fast-growing; anchored by Jamnagar |
| South Korea | ~3.6 MMbbl/d | Export refiner; high complexity |
| Saudi Arabia | ~3.3 MMbbl/d | Expanding via integrated refining-petchem |
| Japan | ~3.1 MMbbl/d | Mature, gradually rationalising |
Source: Energy Institute Statistical Review of World Energy 2026, “Refining capacity”, 2025 (thousand barrels daily, shown here in MMbbl/d; world 103.8 MMbbl/d). The country values were confirmed through a secondary compilation of the same Energy Institute data (Energy Factbook), because the Review’s data file sits behind an email wall; the Review’s own text agrees (US 17.5% and India about 4.7% of global capacity). Atmospheric distillation basis; Russia’s nameplate overstates effective throughput under sanctions.
Figure 5. Top refining countries by capacity, 2025
Source: data in Table 9.
Why location decides the margin. Where a refinery sits sets both sides of its spread. A plant with access to discounted local or pipeline-delivered crude buys below the waterborne benchmark (the supply advantage screened in Section 4.2); a plant far from its customers pays freight to reach them, which is why shipping rules shape US coastal markets (Section 2.7); and the export complexes of Asia, the Middle East and the US Gulf are built to ship product abroad rather than to supply the market next door — who makes and ships each product is Section 2.3.
2.2 The product slate
A refinery’s economics are set by its yield — the mix of products it makes from a barrel. In the United States, motor gasoline, distillate (diesel and heating oil) and jet fuel together make up about 87% of refinery yield (measured, as EIA does, against crude and other inputs), with gasoline the largest share and distillate second. The rest is a long tail of lower-value or specialty streams: still gas, LPG, petrochemical feedstocks, residual fuel oil, asphalt, lubricants and petroleum coke.
The slate is not fixed. Refiners shift it toward whatever product is paying best, within the limits of their equipment. In 2024, US refineries produced a record-high share of jet fuel as air travel recovered, while the gasoline yield fell to its lowest share since 2015 — a live example of the slate flexing toward demand. The long-run direction matters for the investment case: gasoline is a declining product, while jet fuel and petrochemical feedstock are where the durable demand sits (Section 5).
Table 10. Refined products — types, grades and US yield, 2025
| Product family | Main types and grades | Made mainly from | Main uses | US yield, 2025 |
|---|---|---|---|---|
| Motor gasoline | Regular (87 octane), midgrade (89–90), premium (91–94); RBOB and CBOB blendstocks; summer and winter grades | Reformate, FCC gasoline, alkylate and butane; ethanol added at the terminal | Cars and light trucks | 45.9% |
| Distillate fuel oil | Ultra-low-sulphur diesel (≤15 ppm); No. 2 heating oil; marine gasoil | Middle-distillate cut, hydrotreated; hydrocracker and FCC streams | Trucks, rail, farm and construction machinery, ships, heating | 30.0% |
| Jet fuel and kerosene | Jet A (freezes at −40 °C or below, US); Jet A-1 (−47 °C or below, international); heating kerosene | Kerosene cut, hydrotreated; hydrocracker | Aviation; heating | 11.1% |
| Petroleum coke | Fuel grade; calcined anode grade; needle coke | Coker, from vacuum residue | Cement kilns, power plants, aluminium anodes, graphite electrodes | 4.6% |
| Still gas | Methane, ethane and hydrogen mix | Light ends from every unit | Refinery fuel | 3.9% |
| LPG and gas liquids | Propane, butane, isobutane; refinery-grade propylene | Light ends; FCC; reformer | Heating, petrochemical feed, alkylation, gasoline blending | 3.6% |
| Asphalt and road oil | Paving and roofing grades | Vacuum residue | Roads, roofing | 2.0% |
| Residual fuel oil | HSFO (3.5% sulphur, ships with scrubbers); VLSFO (0.5%); ULSFO (0.1%, emission-control areas); No. 6 fuel oil | Residue and heavy cracked streams | Ship bunkers; some power and industrial boilers | 1.9% |
| Petrochemical feedstock | Naphtha; other oils (gasoils); aromatics such as benzene, toluene and xylenes | Naphtha cut; reformer | Steam crackers, plastics, fibres | 1.4% |
| Lubricants and waxes | Base oils Group I, II and III; paraffin wax | Vacuum gasoil, solvent-refined or hydroprocessed | Engine and industrial oils, greases; candles, packaging | 1.0% |
| Special naphthas | Solvent and cleaning grades | Naphtha cut | Paint thinners, cleaning solvents | 0.2% |
| Aviation gasoline | Avgas for piston engines | Alkylate and reformate | Light piston aircraft | 0.1% |
| Miscellaneous products | Specialty streams | Various | Various | 0.5% |
| Processing gain | Extra volume as dense feed becomes lighter product | — | — | −5.9% |
Source: yields per U.S. EIA — refinery yield, 2025, as % of crude and other refinery inputs — products therefore sum to about 106%, and EIA books the processing gain as a negative yield that returns the balance to 100% (jet fuel 11.0% + kerosene 0.1%; petrochemical feedstock is naphtha 0.9% + other oils 0.5%); octane grades per EIA; diesel sulphur per US EPA; jet freezing points (ASTM D1655) per ANSI; marine sulphur limits per IMO; base-oil groups per API 1509 Annex E; the 2024 jet-fuel record per U.S. EIA. The “made mainly from” column is the typical unit route, not a plant-specific flow.
Reading the catalogue. Refined products are families of grades cut from overlapping parts of the barrel, and very often the specification — not the molecule — decides the name. Four points do most of the work.
- Gasoline is a blend, not a cut. No single stream from the column meets spec. The pool is assembled from reformate (high-octane naphtha from the reformer), FCC gasoline, alkylate and butane, and US refiners mostly ship it as a blendstock — RBOB or CBOB — that becomes finished gasoline when ethanol is added at the terminal. Octane grade and seasonal vapour-pressure limits are what separate one gasoline from another. Octane is the fuel’s resistance to knock — the unburned mixture igniting on its own under heat and pressure before the spark-lit flame reaches it — which is why a high-compression engine needs premium.
- Diesel, heating oil and jet fuel are cousins. They share the middle of the barrel and differ mainly in cut points and specification: jet fuel is the lighter kerosene end, held to a strict freezing point; diesel and heating oil are heavier, and on-road diesel must meet the 15 ppm sulphur limit. That shared cut is why a refiner can swing output between jet and diesel within limits — 2024’s record jet share was exactly that swing. Diesel’s grade mark works the other way round from octane: a diesel engine has no spark and fires the fuel by compression alone, so cetane measures how quickly it ignites once injected — the shorter the ignition delay, the higher the number.
- “Fuel oil” names very different oils. No. 2 heating oil is a distillate; residual fuel oil is the heavy bottom of the barrel. Marine residual fuel now splits by sulphur (the IMO 2020 cap, Section 2.7) into HSFO (3.5%, usable only by ships with exhaust scrubbers), VLSFO (0.5%) and ULSFO (0.1%, for emission-control areas) — three grades and three prices for what used to be one product.
- Lubricating oils are graded by how clean they are. The API sorts base oils into Group I (under 90% saturates or over 0.03% sulphur), Group II (at least 90% saturates and no more than 0.03% sulphur) and Group III (Group II purity with a viscosity index of 120 or more); synthetic polyalphaolefins are Group IV and everything else Group V. The higher groups need hydroprocessing and sell at a premium, which is why lubricants run as a specialty business inside the refinery (Table 1).
The order of the table is also, roughly, the order of value. Gasoline, diesel and jet sell for more than the crude they are made from — that premium is the crack spread (Section 1.7). Coke, still gas, sulphur and high-sulphur fuel oil are worth less than crude per barrel, so the share of the barrel that ends in them is a direct drag on what the refiner earns (Section 2.6).
Figure 6. US refinery product slate, 2025
Figure data: Table 10, rescaled from % of input to % of total output (each yield ÷ 106.2%); “other products” is LPG, asphalt, residual fuel oil, petrochemical feedstock, lubricants, special naphthas, aviation gasoline and miscellaneous products.
Who burns what, and why. Figure 6 is what US refineries make; the world burns a different mix. Worldwide, gasoil and diesel is the largest product, at about 28% of oil demand in 2025, narrowly ahead of gasoline — the mirror image of a US fleet built around gasoline-making catalytic crackers (Section 1.5). Each product holds its market through one or two physical properties that suit a particular engine or plant, and that pairing ties each product’s crack to a different part of the economy: diesel to freight and industry, gasoline to car travel, jet fuel to air travel. Table 11 sets out the pairings; its substitute column is where the demand peak in Section 5 comes from.
Table 11. Who uses each refined product, and why
| Product | World demand, 2025 | Largest users | Why it fits the job | Demand tracks | Main substitute |
|---|---|---|---|---|---|
| Gasoil and diesel | 29.2 MMbbl/d | Trucks, rail, ships, farm and construction machinery, generators; transport takes 75% of US distillate (2025) | Fires by compression with no spark, so it must ignite quickly (cetane); carries more energy per litre than other liquid fuels, and is less flammable — the reason armies run on it | Freight, industrial output, farming and construction | Trucks running on liquefied natural gas or batteries; renewable diesel |
| Motor gasoline | 27.9 MMbbl/d | Cars, SUVs and light trucks — 91% of US gasoline (2025); small engines, boats and light aircraft | Fires by spark, so it must resist igniting on its own before the flame arrives (octane) | Car travel and the size of the combustion-engine fleet | Electric vehicles (Section 5) |
| LPG and ethane | 15.2 MMbbl/d | Petrochemical crackers; home heating, cooking and crop drying | Propane shrinks 270-fold as a liquid, so it travels by tank and bottle where no gas grid reaches; ethane is the lowest-cost cracker feed where shale gas supplies it | Plastics demand and winter weather | Piped gas and electric heating; naphtha for crackers |
| Other products (asphalt, coke, lubricants and more) | 10.9 MMbbl/d | Roads and roofing; cement kilns, power plants and aluminium anodes; engines and machinery | The heavy and specialty end of the barrel, sold where a solid or viscous product is the point (Table 7) | Construction and industrial activity | Few at scale |
| Jet fuel and kerosene | 7.8 MMbbl/d | Airlines; home heating | Carries at least 42.8 megajoules per kilogram (MJ/kg), because an aircraft flies its fuel’s weight; Jet A-1 must not freeze above −47 °C, for the cold at altitude, and its flash point — the lowest temperature at which its vapour can ignite — is at least 38 °C, for safe handling | Air travel | Sustainable aviation fuel (SAF) |
| Naphtha | 7.2 MMbbl/d | Steam crackers, above all in Asia (Section 3.2); the refinery’s own reformer | A light liquid that ships and stores like any product, so it is the cracker feed wherever cheap ethane is not on hand | Plastics demand and cracker margins | Ethane, where shale gas makes it cheap |
| Residual fuel oil | 6.3 MMbbl/d | Ship bunkers; some power plants and industrial boilers | The cheapest refined fuel — high-sulphur grades sell below crude — and big marine engines are built to burn it, behind a scrubber or as a low-sulphur grade since IMO 2020 (Section 2.7) | Seaborne trade | Liquefied natural gas as marine fuel; gas and renewables in power generation |
Source: world demand per IEA Oil Market Report, May 2026, “Global Demand by Product” (2025 estimate, of a 104.4 MMbbl/d total; “other products” is IEA’s residual category); US end-use shares per U.S. EIA — use of diesel and U.S. EIA — use of gasoline, 2025; octane per U.S. EIA; cetane per DieselNet; jet limits per DEF STAN 91-091 Issue 16, the Jet A-1 specification (US Jet A freezes at −40 °C, Table 10); LPG and ethane per U.S. EIA — uses of hydrocarbon gas liquids; electric vehicles, gas-fuelled trucks and gas and renewables in power per IEA Oil 2025. The “demand tracks” column and the LPG, naphtha, marine and specialty substitutes are the author’s qualitative reading, not measured values.
Figure 7. World oil demand by product, 2025
Figure data: Table 11, as shares of the IEA’s 104.4 MMbbl/d total; residual fuel oil and other products are combined as “fuel oil & other” to hold the six-segment cap, and shares may not sum to 100% because of rounding.
2.3 Product producers & exporters
Downstream trade is the movement of finished products, and it is increasingly long-haul: with Western plants closing and export-scale complexes rising in the East, product travels further from where it is made to where it is burned. Who makes each product and who ships it are two different lists, and Table 12 sets them side by side.
Output follows capacity. Of the countries that report to JODI (the Joint Organisations Data Initiative’s world oil database; Russia does not), the three largest refiners in Table 9 — the United States, China and India — are also the three largest makers of both gasoline and diesel; the US alone refines about 8.1 MMbbl/d of gasoline and 4.9 MMbbl/d of diesel. Naphtha is the exception: China and South Korea lead it, because naphtha is the feed for Asia’s steam crackers (Section 3.2).
Exports follow the surplus. A country exports what is left after its own demand, so the export leaders are the plants built for the sea. The US still leads diesel and gasoline — Gulf Coast refineries ship both to Latin America and Europe — but Russia and Saudi Arabia follow it in diesel, the Singapore and Dutch hubs in gasoline, and Kuwait ships nearly all the diesel it makes — 0.29 of its 0.33 MMbbl/d.
Table 12. Largest producers and exporters of each refined product, 2025
| Product | Rank | Producer | Output | Exporter | Exports |
|---|---|---|---|---|---|
| Gasoil and diesel | 1 | United States | 4.92 MMbbl/d | United States | 1.26 MMbbl/d |
| Gasoil and diesel | 2 | China | 3.95 MMbbl/d | Russia 1 | 0.82 MMbbl/d |
| Gasoil and diesel | 3 | India | 2.48 MMbbl/d | Saudi Arabia | 0.68 MMbbl/d |
| Motor gasoline | 1 | United States | 8.13 MMbbl/d | United States | 0.90 MMbbl/d |
| Motor gasoline | 2 | China | 3.60 MMbbl/d | Singapore 2 | 0.68 MMbbl/d |
| Motor gasoline | 3 | India | 1.16 MMbbl/d | Netherlands 2 | 0.44 MMbbl/d |
| Jet fuel and kerosene 3 | 1 | United States | 1.84 MMbbl/d | China, incl. bunkers 4 | 0.46 MMbbl/d |
| Jet fuel and kerosene 3 | 2 | China | 1.25 MMbbl/d | South Korea | 0.27 MMbbl/d |
| Jet fuel and kerosene 3 | 3 | South Korea | 0.53 MMbbl/d | Kuwait | 0.26 MMbbl/d |
| Naphtha | 1 | China | 1.93 MMbbl/d | Netherlands 2 | 0.40 MMbbl/d |
| Naphtha | 2 | South Korea | 0.80 MMbbl/d | Algeria | 0.20 MMbbl/d |
| Naphtha | 3 | India | 0.48 MMbbl/d | Kuwait | 0.17 MMbbl/d |
| Residual fuel oil 5 | 1 | China | 0.79 MMbbl/d | Russia, incl. VGO 1 | 0.88 MMbbl/d |
| Residual fuel oil 5 | 2 | Saudi Arabia | 0.46 MMbbl/d | China, incl. bunkers 4 | 0.39 MMbbl/d |
| Residual fuel oil 5 | 3 | United States | 0.31 MMbbl/d | Singapore 2 | 0.33 MMbbl/d |
Notes to Table 12
- Russia is an IEA estimate from tanker tracking, not a JODI report. Russia, like the UAE, Iran, Iraq, Oman and Qatar, reported no 2025 product data to JODI, so none of them can be ranked as a producer. Its fuel-oil figure is the IEA’s “Resid+VGO” line, which also counts vacuum gasoil, a cracker feed. Its 0.61 MMbbl/d of naphtha, NGLs and LPG is not split out, so Russia is left out of the naphtha ranking.
- Singapore and the Netherlands are trading hubs. They import product as well as refine it, so their gross exports include blended and re-shipped cargoes; Table 13 nets them off. Singapore reports trade to JODI but not refinery output.
- JODI’s “Kerosenes” line, of which kerosene-type jet fuel is the main part. It is used because China and Saudi Arabia report kerosene only in total.
- China’s jet and fuel-oil exports include bunker fuel. Chinese customs books bonded jet fuel sold to international flights and bonded fuel oil sold to ships at its ports as exports (S&P Global, S&P Global); JODI’s definition would count those sales as demand, and JODI marks China’s data as not assessed.
- JODI’s “Fuel oil” line — heavy residual oil, including bunker oil.
Source: JODI Oil World Database, secondary products, 2025 — “Refinery output” and “Exports”, annual averages of the twelve monthly figures in thousand barrels, shown here in MMbbl/d; Russia per the IEA Oil Market Report, April 2026, “Russian Crude and Product Exports”, p.22 (2025, from Kpler and Argus data). Exports exclude transit trade and international marine and aviation bunkers (JODI manual, §4.3); US gasoline output excludes the ethanol blended at the terminal. Data retrieved 25 September 2026; JODI revises recent months.
The diesel trade. Gasoil and diesel is the most traded product among JODI’s reporting countries: about 6.9 MMbbl/d of exports in 2025, over a quarter of their total. The Atlantic Basin is structurally short of it after the European closures and pulls barrels from the US Gulf, the Middle East and India, while Australia is the largest single importer at 0.52 MMbbl/d. Figure 8 ranks the eight largest exporters.
Figure 8. Largest gasoil and diesel exporters, 2025
Chart source: JODI Oil World Database, “Gas/diesel oil” exports, 2025 annual average; Russia per the IEA Oil Market Report, April 2026, “Gasoil”, p.22. The top three repeat Table 12.
Gross exports flatter the hubs. Table 13 ranks countries by total product exports. The US ships about 6.1 MMbbl/d, 23% of world product trade by the Energy Institute’s count — though 2.4 MMbbl/d of that is LPG, and about 90% of US gas liquids come from gas processing plants rather than refineries (see Part 2). The net column is the truer measure of who supplies whom. The Netherlands and Singapore import almost as much as they export, and China, the largest refiner, is a net importer: it imports 1.1 MMbbl/d of LPG and 0.4 MMbbl/d of naphtha, the light feeds of the petrochemical build-out in Section 3.2. The largest net importers are Indonesia (−0.98 MMbbl/d), Australia (−0.83), China (−0.81), Japan (−0.72) and Mexico (−0.57).
Table 13. Largest refined-product exporters, 2025
| Country | Exports | Imports | Net position | Main exports |
|---|---|---|---|---|
| United States | 6.13 MMbbl/d | 1.46 MMbbl/d | +4.68 MMbbl/d | LPG, diesel, other products |
| Russia 1 | 2.57 MMbbl/d | n/d | n/d | Fuel oil and VGO, diesel, naphtha and LPG |
| Netherlands | 2.15 MMbbl/d | 1.64 MMbbl/d | +0.50 MMbbl/d | Diesel, gasoline, naphtha |
| Singapore | 1.63 MMbbl/d | 1.92 MMbbl/d | −0.28 MMbbl/d | Gasoline, diesel, fuel oil |
| Saudi Arabia | 1.49 MMbbl/d | 0.36 MMbbl/d | +1.13 MMbbl/d | Diesel, gasoline, fuel oil |
| India | 1.38 MMbbl/d | 1.27 MMbbl/d | +0.11 MMbbl/d | Diesel, gasoline, jet fuel |
| South Korea | 1.37 MMbbl/d | 1.01 MMbbl/d | +0.35 MMbbl/d | Diesel, gasoline, jet fuel |
| China, incl. bunkers 2 | 1.23 MMbbl/d | 2.04 MMbbl/d | −0.81 MMbbl/d | Jet fuel, fuel oil, gasoline |
| Kuwait | 1.07 MMbbl/d | 0.00 MMbbl/d | +1.07 MMbbl/d | Diesel, jet fuel, LPG |
| Malaysia | 0.79 MMbbl/d | 0.74 MMbbl/d | +0.06 MMbbl/d | Gasoline, diesel, fuel oil |
Notes to Table 13
- Russia is the IEA’s 2025 export estimate (“Total Products”); Russia reports no imports to JODI, so its net position is not shown.
- China’s exports include bonded bunker sales of jet fuel and fuel oil (Table 12, note 4).
Source: JODI Oil World Database, “Total oil products” exports and imports, 2025 annual averages, in MMbbl/d, including LPG and JODI’s “Other oil products” (ethane, petroleum coke, lubricants, bitumen and waxes); net position = exports − imports; Russia per the IEA Oil Market Report, April 2026, p.22. US share of world product trade per the Energy Institute Statistical Review of World Energy 2026; US HGL origin per U.S. EIA (2022). Main exports are each country’s three largest product lines in the same data.
The seaborne product trade rides the same maritime chokepoints as crude — the Strait of Hormuz, the Suez Canal and Bab el-Mandeb, and the Panama Canal — covered in full in the oil guide. But downstream adds a chokepoint of its own: the concentration of export refining in a handful of mega-complexes. When Jamnagar or a Gulf Coast cluster goes down, the product market feels it globally, because there are now fewer, larger sources of export supply than there were when every country refined its own fuel.
2.4 The installed base
The most revealing structural fact about US refining is hidden in a simple count. In 1982 the United States had 301 operable refineries; by 2026 it had just 130 — a fall of more than half. Yet total capacity barely moved, holding around 18 MMbbl/d the whole time. The industry did not shrink; it consolidated. Small, simple plants closed, and the survivors expanded and added conversion units, raising the fleet’s complexity. Fewer, bigger, more complex refineries now do the same work — the refining analogue of midstream absorbing a doubling of gas through the same pipeline mileage.
Figure 9. US operable refineries, 1982–2026
Figure data: U.S. EIA — number of operable refineries; 2000 figure approximate (~160). Peak was 301 in 1982; capacity held roughly flat as survivors expanded.
That consolidation is still running. In 2025 the US lost roughly 400 Mbbl/d of capacity as LyondellBasell closed its 264 Mbbl/d Houston refinery (March 2025) and Phillips 66 shut its 139 Mbbl/d Los Angeles refinery (October 2025); Valero’s 145 Mbbl/d Benicia refinery then ceased refining operations. The West Coast, poorly connected to Gulf Coast supply by pipeline, is now structurally short. Europe is rationalising even faster: INEOS/Petroineos closed Grangemouth — Scotland’s only crude refinery and 13% of UK capacity (Argus) — in April 2025, converting it to an import terminal after losing about $500,000 a day, and Shell and BP closed German plants at Wesseling and Gelsenkirchen. Each closure tightens the market for the refineries that remain, which is a large part of why margins have been so strong.
Table 14. US refining capacity and refinery count, selected years
| Year | Operable refineries | Capacity |
|---|---|---|
| 1982 | 301 plants | ~17.9 MMbbl/d |
| 2000 | ~160 plants | ~16.5 MMbbl/d |
| 2010 | 148 plants | ~17.6 MMbbl/d |
| 2020 | 135 plants | ~19.0 MMbbl/d |
| 2025 | 132 plants | ~18.4 MMbbl/d |
| 2026 | 130 plants | ~18.2 MMbbl/d |
Notes to Table 14
- Capacity is on a calendar-day basis.
Source: U.S. EIA Refinery Capacity Report, the U.S. EIA series for operable refineries and operable capacity as of 1 January, and EIA — US refining capacity decreased during 2025; capacity is atmospheric distillation, barrels per calendar day, and the 2000 count is rounded.
2.5 Capacity vs demand
A commodity market clears through price; a refining market clears through margin. The equivalent of a supply–demand balance is the gap between refining capacity and product demand, and the crack spread is the price of that gap. Tight capacity blows the crack out; loose capacity collapses it until marginal refiners close.
The demand side is observable and, for the first time in the sector’s history, close to its peak. US gasoline demand runs about 8.9 MMbbl/d and distillate about 3.9 MMbbl/d in 2025, both still below their pre-pandemic highs (U.S. EIA, product supplied). Globally, product demand is within a few years of its projected peak — the numbers, and what they mean for the fleet, are in Section 5.
Table 15. Refining margins and utilisation across the cycle
| Metric | Reading | Period | Note |
|---|---|---|---|
| US Gulf Coast 3-2-1 crack spread | ~$17/bbl average | 2010–2021 | The long-run “normal” |
| New York Harbor diesel crack (ULSD less Brent) | ~$89/bbl monthly average | Oct 2022 | Peak of settled history |
| US refining margin | Losses / negative | 2020 | Demand collapse in COVID |
| Atlantic Basin refining margin | All-time highs | Jul 2026 (running) | Supply shortfalls + low stocks |
| US refinery utilisation | ~90% typical | 2024–25 | Gross inputs ÷ operable capacity |
Source: U.S. EIA spot prices, author’s arithmetic — the diesel crack is New York Harbor ULSD ($4.345/gal × 42) less Brent ($93.33), October 2022 monthly averages, the highest month in the series through 2025; the 3-2-1 average is two parts Gulf Coast conventional gasoline and one part Gulf Coast ULSD less WTI, annual averages 2010–2021; crack-spread convention per U.S. EIA; utilisation per U.S. EIA (90.5% in 2024, 92.0% in 2025); the 2026 row per the IEA Oil Market Report, August 2026 — an episode still running at the data date, shown as live context, not as a settled reference. The crack spread is a gross-margin proxy, not a refiner’s realised margin.
Where the shortage shows up. A refining bottleneck does not raise the price of crude — it raises the price of products relative to crude, i.e. the crack. Two recent episodes make the mechanism concrete, and both are settled history. In 2020, demand collapsed in the pandemic and refiners ran at a loss, forcing a wave of closures. Barely two years later the opposite happened: post-pandemic demand recovery met a fleet that had been shrunk by those very closures, and the 2022 crack spread reached records — New York Harbor diesel cracks averaged about $89/bbl in October 2022 — handing refiners the best year they had ever posted — a super-margin, this guide’s term for a window in which cracks run far above their long-run average. Atlantic Basin margins hit all-time highs again in July 2026 (IEA) on further closures and Middle East supply disruption, but that run was still in progress at this guide’s data date, so it is not used as a settled reference here. The lesson is that in refining, scarcity is created by the closures that a downturn forces, which is why the cycle is so violent and so self-correcting.
2.6 Refining economics
The crack spread (Section 1.7) is the market’s shorthand for refining profitability. This section follows a real barrel from that shorthand to profit, using the reported per-barrel results of Valero — one of the two largest US independent refiners, running about 3 MMbbl/d — and then answers the question investors ask most often: does any of it depend on the oil price?
Revenue is mostly crude passing through. A refiner’s top line flatters its size and hides its profitability, because most of the revenue is simply the cost of crude bought and resold in another form. Valero booked $122.7 bn of revenue in 2025 and kept $2.35 bn of net income — 1.9%; even in the record year of 2022 the figure was only 6.5%. A percentage-of-revenue margin therefore says little, and the industry measures itself per barrel of throughput instead.
From crack to profit. Four links separate the headline number from the money. The benchmark crack is a market indicator for a standard product slate. The realised refining margin is what the company actually captures — product revenue less crude and other feedstock cost — and it drifts away from the benchmark for concrete reasons: a heavy-sour discount adds to it (Section 1.4), byproducts that sell below crude subtract from it (Section 1.6), and compliance costs such as RINs and the location of the plant move it either way. From that margin come cash operating expenses — natural gas and power, labour, maintenance, catalysts and chemicals — and depreciation, much of which reflects the heavy, recurring turnaround spend. What survives is operating income per barrel.
Figure 10. From refining margin to operating income, per barrel — Valero, 2025
margin
expenses
income
Figure data: Table 16, 2025 row.
Figure 10 shows the 2025 result on each of the segment’s 2.99 MMbbl/d — about $5.3 bn of adjusted operating income for the year. Some 60% of the margin is gone before any profit appears, and that cost base barely moves with the market: it is paid whether the crack is $5 or $40. That is the operating leverage behind the sector’s violence. Across the six years in Table 16, the margin moved between $5.32 and $21.82 a barrel — roughly fourfold — while operating income swung from a loss of $1.18 to a profit of $14.62.
Table 16. Valero refining segment per barrel, 2020–2025
| Year | Brent | US Gulf Coast diesel crack | Refining margin | Operating expenses | Depreciation & amortisation | Operating income | Net income ÷ revenue |
|---|---|---|---|---|---|---|---|
| 2020 | $43.15/bbl | $7.11/bbl | $5.32/bbl | $4.22/bbl | $2.28/bbl | −$1.18/bbl | −2.2% |
| 2021 | $70.79/bbl | $13.75/bbl | $9.04/bbl | $5.00/bbl | $2.13/bbl | $1.91/bbl | 0.8% |
| 2022 | $98.86/bbl | $38.73/bbl | $21.82/bbl | $5.11/bbl | $2.09/bbl | $14.62/bbl | 6.5% |
| 2023 | $82.27/bbl | $25.06/bbl | $17.55/bbl | $4.79/bbl | $2.16/bbl | $10.60/bbl | 6.1% |
| 2024 | $79.79/bbl | $15.76/bbl | $10.62/bbl | $4.64/bbl | $2.24/bbl | $3.74/bbl | 2.1% |
| 2025 | $68.18/bbl | $19.10/bbl | $12.29/bbl | $4.93/bbl | $2.53/bbl | $4.83/bbl | 1.9% |
Notes to Table 16
- Operating income is the refining segment’s adjusted figure.
Source: Valero fourth-quarter and full-year earnings releases — 2021 (2020–21 figures), 2023 (2022–23) and 2025 (2024–25); refining-segment figures per barrel of throughput, with Brent and the crack as each release’s market reference prices. The diesel crack is US Gulf Coast ultra-low-sulphur diesel less Brent as printed; the 2025 release shows it adjusted for renewable-fuel compliance cost, so the column is not strictly like-for-like across the window. Net income ÷ revenue is Metal Pilot arithmetic on the consolidated figures. Company results are concluded history, used to illustrate the mechanics — not a valuation or a view on the company.
Does it depend on the oil price? Directly, much less than intuition suggests: the margin is a spread, so in principle the level of crude cancels out. Table 16 shows it. From 2024 to 2025 Brent fell by more than $11 a barrel, yet Valero’s margin rose, because diesel cracks widened. The years 2020–2022, when oil and margins soared together, are not a counter-example — both were lifted by the same demand recovery meeting a fleet shrunk by closures (Section 2.5), a common cause rather than one driving the other. The oil price still reaches a refiner’s results, though, through five side doors.
Table 17. How the oil price reaches a refiner’s earnings
| Channel | Mechanism | When crude rises | Most exposed |
|---|---|---|---|
| Inventory valuation | Crude and product stocks are revalued as prices move | FIFO books a holding gain that reverses when prices fall; LIFO keeps most of it out of profit | IFRS reporters, which cannot use LIFO |
| Low-value byproducts | Coke, sulphur and high-sulphur fuel oil sell below crude and lag its moves | Realised margin slips further below the crack | Deep-conversion plants with large coke yields |
| Energy cost | Refineries burn fuel gas plus purchased natural gas and power | Operating cost per barrel rises when gas rises with oil | Refiners buying gas at high regional prices |
| Crude differentials | Heavy and sour discounts widen or narrow with the supply of those grades | Either direction; not tied to the price level | Complex refiners whose edge is the discount |
| Demand | Very high pump prices curb consumption; a price crash usually signals collapsing demand | Cracks compress once demand destruction sets in | Gasoline-heavy slates |
Source: inventory cost formulas per IFRS — IAS 2 Inventories and replacement-cost reporting per TotalEnergies; energy cost and differentials per the Valero 2025 release and the TotalEnergies European refining margin marker. The “most exposed” column is the author’s qualitative assessment, not a measured ranking.
Three of those doors are worth a closer look. Inventory is an accounting effect, but a large one: a refiner holds millions of barrels in tanks, so a price move revalues them. US GAAP allows LIFO (last-in, first-out) costing, which keeps most of that holding gain out of reported profit; IAS 2 permits only FIFO or weighted-average cost, so IFRS reporters book the gain and then publish replacement-cost or current-cost-of-supply (CCS) results that strip it back out. Energy is a real cost: Valero’s reference natural-gas price rose from $1.88 to $3.04 per million British thermal units (MMBtu) between 2024 and 2025, and TotalEnergies’ European refining margin marker deducts gas and EU carbon costs precisely because they bite. And differentials can erode the complex refiner’s edge while the headline oil price does something else entirely: in 2025 the Brent–Maya discount narrowed from $11.43 to $8.46 a barrel and Brent–WCS at Houston from $10.36 to $7.21, even as Brent fell.
The practical order for reading a refiner’s results is therefore cracks first, then differentials, then energy and byproduct drag, then inventory effects — and only then the oil price. What that means for owning refiners alongside producers is developed in Section 4.3.
2.7 Regulation & fuel standards
Downstream has no OPEC and no rate regulator setting its margin. What shapes it instead is product regulation — the specifications fuels must meet and the mandates for blending renewables — and these are genuinely consequential for refinery economics.
In the United States, the Renewable Fuel Standard (RFS) requires transport fuel to contain a rising volume of renewable fuel, tracked through tradable credits called RINs (Renewable Identification Numbers). A refiner that does not blend enough biofuel must buy RINs to comply, which is a direct cost that moves with credit prices — large enough that Valero’s 2025 release prints its reference diesel crack net of it (Table 16), so check the compliance line before comparing one refiner’s margin with another’s. The 2025 obligation ran to billions of gallons, and renewable-diesel capacity grew from about 1.75 bn gallons a year at the start of 2022 to 4.7 bn at the start of 2025 (U.S. EIA), much of it on the West Coast where federal RINs stack with state low-carbon-fuel-standard (LCFS) credits. Fuel-quality specifications — Tier 3 gasoline (low sulphur) and ultra-low-sulphur diesel — force refiners to run hydrotreating capacity that simpler foreign plants can skip. The Jones Act raises the cost of moving product by ship between US ports, which is why the East and West Coasts often import foreign product rather than buy from the Gulf.
The most consequential product rule of the last decade came from the sea. IMO 2020 cut the sulphur limit in marine fuel from 3.5% to 0.5% worldwide (0.1% in emission-control areas), forcing shippers onto very-low-sulphur fuel oil (VLSFO) and rewarding complex refineries that could make clean distillate instead of dirty residual fuel. It cut shipping’s sulphur emissions by about 70% and permanently widened the value gap between simple and complex refineries.
Table 18. Who and what regulates downstream
| Regime | Body | What it governs | Practical consequence |
|---|---|---|---|
| Renewable Fuel Standard (RFS) | US EPA | Biofuel blending, RINs | Non-blenders buy RIN credits; a direct cost |
| Fuel-quality specs (Tier 3, ULSD) | US EPA | Sulphur, octane, additives | Requires hydrotreating; favours complexity |
| Low-carbon fuel standards | California et al. | Carbon intensity of fuels | Stacks with RINs; drives renewable diesel |
| Jones Act | US (maritime) | US-flag coastwise shipping | Raises inter-US product freight; favours imports |
| IMO 2020 | IMO (global) | Marine fuel sulphur (0.5%) | Killed high-sulphur fuel oil; rewards distillate |
| Emissions / carbon policy | EU Emissions Trading System (ETS), others | Refinery CO₂ | Adds cost; pressures European margins |
| Product export / import rules | Various | Trade in fuels | Shapes Atlantic-basin and Pacific flows |
Source: US EPA Renewable Fuel Standard, IMO 2020 sulphur cap and standard fuel-specification references; the 70% cut in shipping’s sulphur oxide emissions per the IMO press briefing.
2.8 Storage, inventories & resilience
Downstream holds the system’s product buffer. Where crude has strategic reserves (covered in the oil guide), the product system’s resilience is mostly commercial: refineries, terminals and pipelines hold weeks of gasoline, diesel and jet fuel inventory, and the level of those stocks relative to their seasonal range is one of the market’s clearest tightness signals. Low distillate inventories going into winter, for example, are what turn a firm diesel crack into a spike. The US numbers arrive every week in EIA’s Weekly Petroleum Status Report, which also expresses stocks as days of supply — the figure to read against its five-year seasonal range.
Product storage also gives downstream a small counter-cyclical hedge. When the forward curve is in contango — future product prices above today’s — a refiner or marketer with tankage can store product and sell it forward at a profit, earning when flat physical margins are weak. It is a modest offset to the sector’s dominant exposure, which is the crack, but it is real and it rewards scale in logistics.
3. The companies & the value chain
3.1 The largest downstream companies
Downstream is less concentrated than midstream but is dominated at the top by a mix of independent refiners, integrated majors and national oil companies. Scale is best understood asset by asset, because a company’s refining capacity, complexity and integration all matter more than any single financial number.
Two facts in Table 19 matter more than the ranking: one of the largest refiners in the world, Sinopec, is a state company processing some 250 Mt of crude a year, and the largest single refinery, Reliance’s Jamnagar, is a private Indian complex integrated with a giant petrochemical site — the template the whole industry is moving toward. In the United States the merchant refiners lead, with Marathon Petroleum and Valero the two largest.
Table 19. Major downstream companies and their signature assets
| Company | Country | Focus | Signature asset or scale | Listing |
|---|---|---|---|---|
| Sinopec | China | Integrated refining + petchem | ~250 Mt crude processed (2025) | State-controlled (NYSE/HKEX/SSE: SNP) |
| ExxonMobil | US | Integrated major | ~4.1 MMbbl/d global refining (its share) | Public (NYSE:XOM) |
| Saudi Aramco | Saudi Arabia | Integrated NOC | Gross ~7.8 MMbbl/d refining interests (net ~4.2) | State-controlled (Tadawul: 2222) |
| Marathon Petroleum | US | Independent refiner | Galveston Bay (~631 Mbbl/d); Speedway legacy | Public (NYSE:MPC) |
| Valero Energy | US | Independent refiner | 14-refinery system; large renewable diesel | Public (NYSE:VLO) |
| Phillips 66 | US | Refiner + midstream + chemicals | CPChem JV (petrochemicals) | Public (NYSE:PSX) |
| Reliance Industries | India | Integrated refining + petchem | Jamnagar (~1.4 MMbbl/d), world’s largest | Public (NSE/BSE: RELIANCE) |
| TotalEnergies | France | Integrated major | European + SATORP (Saudi) refining | Public (NYSE/Euronext: TTE) |
| Shell | UK | Integrated major | Pernis, Pulau Bukom energy & chemicals parks | Public (NYSE:SHEL, LSE:SHEL) |
| Chevron | US | Integrated major | Pascagoula, Richmond refineries | Public (NYSE:CVX) |
| PBF Energy | US | Independent refiner | Six US refineries | Public (NYSE:PBF) |
| HF Sinclair | US | Independent refiner | ~678 Mbbl/d, 7 refineries; Sinclair brand | Public (NYSE:DINO) |
| SK Innovation | South Korea | Refiner + batteries | Ulsan complex | Public (KRX:096770) |
| Formosa Petrochemical | Taiwan | Refiner + petchem | Mailiao complex | Public (TWSE: 6505) |
| PetroChina | China | Integrated NOC | National refining system | State-controlled (NYSE/HKEX/SSE: PTR) |
Sources: company disclosures, 2025–26 — Sinopec 2025 annual results (refinery throughput 250.33 Mt), ExxonMobil 10-K 2025, Item 2 (its share of refining capacity at year-end 2025, 4,113 thousand barrels daily, stream-day basis), Aramco Annual Report 2025 (gross 7.8 and net 4.2 MMbbl/d at 31 December 2025), Valero (14 refineries), Reliance (Jamnagar) and HF Sinclair (~678,000 b/d, annual average); Galveston Bay per the U.S. EIA Refinery Capacity Report, Table 3 (1 January 2026). Listing gives the primary exchange and ticker; NOC entries are majority state-owned. Signature asset or scale mixes group capacity with single plants, so it indicates scale rather than ranking it — Figure 11 ranks single refineries. No market-capitalisation figures are shown by design.
Figure 11. Five of the world’s largest refineries by capacity
Source: Reliance (Jamnagar, 1.4 MMbbl/d), U.S. EIA Country Analysis Brief: Venezuela (Paraguaná, nameplate, 2024), SK (Ulsan, “up to 840,000 barrels” a day) and the U.S. EIA Refinery Capacity Report, Table 3 (US plants, b/cd, 1 January 2026); figures approximate, nameplate/complex basis, and a selection rather than a complete ranking. Motiva Port Arthur is the largest US refinery on both bases (656 Mbbl/cd and 678 Mbbl/sd against Galveston Bay’s 631 and 665).
3.2 Marketing, retail & petrochemicals
Refining is the volatile heart of downstream, but the other two legs are what make an integrated business more than a bet on the crack spread.
Marketing and retail is where the sector meets the consumer, and in the US it is strikingly fragmented: there are about 152,000 convenience stores, and even the largest operator — 7-Eleven, with about 12,369 stores — is under a tenth of the total. Circle K (about 5,851 sites) and Casey’s (about 2,899) follow. Refiners increasingly sell their brands (Marathon, Shell, ExxonMobil, 76, Sinclair) to independent operators rather than owning stations directly, capturing a steady wholesale margin and licensing fee while the volatile retail and merchandise business sits with franchisees. The merchandise margin inside a convenience store — coffee, snacks, tobacco — is far larger and steadier than the margin on the fuel itself, which is why the two are sold together.
Figure 12. US fuel-retail leaders by store count, 2025
Source: NACS / Convenience Store News Top 100, 2025; US convenience-store total ~152,000. The market is highly fragmented — the three largest chains together are only about 14% of all US convenience stores.
Where it shows up at the pump. EIA’s breakdown of the 2025 average US retail gasoline price — $3.10 a gallon — assigns 51.4% to crude oil, 14.3% to refining costs and profits, 17.8% to distribution and marketing and 16.6% to taxes. The refining slice, about 44¢ a gallon or $18.6 a barrel, has to cover the whole cost of running the plant as well as its profit (Section 2.6), and over 2016–2025 it averaged 15.0% of the pump price. The marketing slice is larger, but it is shared across terminals, wholesalers and some 152,000 stations.
Petrochemicals is the growth engine and, increasingly, the reason a refinery is built at all. Steam crackers turn cheap feedstock — ethane in the US, naphtha in Asia — into ethylene and propylene, the building blocks of the plastics, fibres and packaging in nearly every manufactured good. The IEA expects petrochemicals to surpass road transport as the main source of oil-demand growth, consuming roughly one in every six barrels by 2030. But the near-term picture is oversupplied: China became the world’s largest ethylene producer in 2022 (Industrial Info) and kept building, pushing the global ethylene operating rate down to about 84% in 2025 (from over 90% before 2019). Wood Mackenzie counts some 27% of global ethylene capacity at some risk of closure — and in Europe, announced closures (12% by 2028) and plants up for sale (13%) together cover roughly a quarter of capacity. The same consolidation reshaping refining is now hitting chemicals: high-cost Western crackers close while integrated Middle Eastern and Chinese complexes expand.
3.3 Company archetypes & structures
Downstream companies differ along two axes: what they own across the refining–marketing–petrochemical chain, and how integrated they are with upstream production.
The second axis is the one that sets margin volatility. A merchant refiner that owns no production books every swing in the crack straight to its bottom line; an integrated major nets a refining loss against an upstream gain — high crude hurts one segment and helps the other — so its downstream is a hedge, not the whole company. The first axis decides which cycle a company rides: a petrochemical-led name rides the polymer cycle rather than the fuel cycle, and a marketing- and retail-led convenience chain barely touches refining, earning a merchandise-and-fuel margin that behaves more like retail than energy. National oil companies sit off the volatility scale, because refining there is state infrastructure with strategic and employment goals alongside profit. Table 20 names the examples.
Table 20. Downstream archetypes and their economics
| Archetype | What they own | Margin driver | Margin volatility | Examples |
|---|---|---|---|---|
| Independent refiner | Refining + some marketing | Crack spread | Very high | Valero, PBF, HF Sinclair |
| Integrated major | Upstream + downstream + chemicals | Blended; downstream hedges upstream | Low–medium | ExxonMobil, Shell, Chevron |
| Petrochemical-led | Crackers + polymers (+ refining) | Integration / polymer margin | High | Formosa, Reliance |
| Marketing & retail-led | Stations + merchandise | Fuel + merchandise margin | Low | Casey’s, Couche-Tard |
| National oil company | State refining + petchem | Policy + margin | Managed | Sinopec, Aramco, PetroChina |
Source: archetype taxonomy from company filings and the Metal Pilot project model; examples are illustrative, not a recommendation.
Figure 13. Downstream archetypes by exposure to the crack and margin volatility
Low
High
Source: data in Table 20; positions are the author’s qualitative assessment, not measured values. Exposure to the crack is how much of an archetype’s earnings moves with refinery throughput and the crack spread. The shaded zone marks the high-exposure, high-volatility quadrant where merchant refiners sit.
3.4 Assets on the balance sheet
A downstream balance sheet looks like heavy industry, not like an oil producer’s. There are no reserves, no depletion and no reserve-life ratio. What sits there instead is property, plant and equipment — refineries, crackers, terminals and stations at historical cost less depreciation — and for a long-lived, much-modified asset, book value can bear little relation to replacement cost.
Four line items carry most of the information. Refining capacity and complexity determine earning power far more than book value does — a high-complexity plant on a coast with crude-supply optionality is worth a multiple of a simple inland one of the same nominal capacity. Turnaround and maintenance capital is large, lumpy and non-discretionary: refineries shut every few years for major overhauls, and the timing swings reported earnings. Working capital is unusually large and volatile because the company holds billions of dollars of crude and product inventory whose value moves with prices — an inventory gain or loss can dominate a quarter. And environmental and decommissioning liabilities are a growing tail item as plants age and close. The practical instruction: when reading a downstream filing, start with capacity, complexity and configuration, then turnaround schedule and capex, then inventory and working-capital exposure — in that order.
4. Investing in downstream
4.1 Ways to own it
Like midstream, downstream has no commodity to buy — there is no futures contract on a refinery’s margin that a retail investor can hold for long. Every route in is an equity or a fund. But unlike midstream’s fee stream, what you are buying is a cyclical manufacturing margin, so the structural trap is different, as the caveat below explains.
The vehicle spectrum in Table 21 runs from purest to most diluted exposure to the crack — independent refiners at one end; integrated majors, petrochemical names and convenience retailers diluting it in different directions; and broad energy ETFs at the other. There is no widely held refiner ETF analogous to midstream’s MLP funds, so most exposure is taken through individual equities or broad energy funds.
The cyclicality caveat — the downstream equivalent of midstream’s tax drag. In midstream the classic trap is a fund bleeding value to tax drag (Part 2, Section 5.1); in downstream it is mistaking a super-margin for a permanent one. Refiner shares look cheapest on trailing earnings exactly when margins have peaked and are about to revert — a low price-to-earnings multiple on record earnings is a warning, not a bargain. The disciplined approach is to value refiners on mid-cycle margins — the average across a full cycle — not the current print, and to treat a period of record cracks as a windfall to be returned to shareholders rather than a new baseline. Investors who bought refiners on peak 2022 earnings and sold on the reversion learned this the expensive way.
Table 21. Ways to take downstream exposure
| Vehicle | What you own | Margin exposure | Key risk | Best suited for |
|---|---|---|---|---|
| Independent-refiner shares | A pure refining business | Direct crack spread | Full margin cyclicality | High-conviction cycle investors |
| Integrated-major shares | Upstream + downstream + chemicals | Diluted, hedged | Downstream is a small slice | Lower-volatility exposure |
| Petrochemical-led shares | Crackers + polymers | Polymer cycle | Global overcapacity | Chemicals-cycle investors |
| Convenience-retail shares | Stations + merchandise | Marketing margin | Fuel-demand peak | Lowest-volatility, consumer-like |
| Renewable-fuels names | Renewable diesel / SAF | Spread + policy credits | RIN/policy risk | Policy-driven growth |
| Energy-sector ETF | A basket incl. refiners | Broad, indirect | Not a pure play | Diversified exposure, no single-name risk |
| Physical product / cracks | Not available to retail | — | Requires a trading desk | Industrial buyers only |
Source: vehicle structures from issuer documentation and company filings; structural attributes only — no fees, yields or prices are quoted.
4.2 How to value & screen downstream
Producers are screened on reserves and cost; refiners have neither, so the screen is built from capacity, complexity, configuration and balance-sheet resilience. Six measures do most of the work, with integration as a seventh cross-check (Table 22), and they are all in the filings.
Complexity and configuration come first — the Nelson index and the mix of conversion units tell you whether a refinery can profit from cheap heavy crude or is stuck as a simple margin-taker. Location and crude-supply advantage decide the feedstock cost: a plant with access to discounted local crude or an integrated pipeline earns more than one buying at the waterborne benchmark. Mid-cycle margin capture — not the current crack — is the number to value on (the trap in Section 4.1). Utilisation and turnaround discipline show whether the asset runs reliably, since an unplanned outage in a strong margin environment is enormously costly. Balance-sheet strength matters more here than almost anywhere because the model must survive years of thin or negative margins; the refiners that entered 2020 with low leverage are the ones that lived to enjoy 2022. And capital return — buybacks and dividends funded by mid-cycle, not peak, cash flow — tells you whether management understands its own cyclicality.
Valuation itself is done on normalised (mid-cycle) cash-flow multiples and on replacement-cost or per-complexity-barrel metrics for the assets. Those ratios are the screening machinery rather than facts to print in an evergreen guide, which is why this report names them without quoting levels — you can screen the downstream universe on Metal Pilot and, because a refiner’s margin is only as good as the crude it can source and the products it can move, cross-check the logistics that feed and evacuate it at the same time.
Table 22. Screening metrics for downstream companies
| Metric | What it tells you | Healthy vs concerning | Where to find it | Taught in |
|---|---|---|---|---|
| Nelson complexity | Ability to profit from cheap heavy crude | High and rising vs simple hydroskimming | 10-K asset descriptions | 1.3, 1.5 |
| Crude-supply advantage | Feedstock cost vs benchmark | Discounted/integrated vs waterborne-priced | Sourcing disclosure | 1.4, 2.1 |
| Mid-cycle margin capture | Earnings power through the cycle | Above peers on normalised basis vs only in peaks | Segment history | 2.5, 4.1 |
| Utilisation & turnaround record | Operating reliability | Consistently high vs frequent outages | Operating statistics | 2.4, 3.4 |
| Net debt / EBITDA (mid-cycle) | Survival through a trough | Low, laddered vs stretched | Financial statements | 3.4 |
| Capital return discipline | Payout funded by mid-cycle cash | Buybacks in strength, held in weakness | Capital-allocation guidance | 4.1 |
| Integration (petchem/marketing) | Cushion against the crack | Diversified vs pure-crack | Segment mix | 3.2, 3.3 |
Source: metric taxonomy from downstream company filings; the Metal Pilot project model captures the asset-side inputs — capacity, complexity, configuration — behind several of these.
A hypothetical illustration of mid-cycle valuation. Take a 300 Mbbl/d refiner that keeps $5 of operating income per barrel at a mid-cycle margin and $15 at a peak — roughly the range in Table 16. It earns about $0.55 bn a year mid-cycle and $1.64 bn at the peak. Valued at 6× mid-cycle operating income — a round multiple chosen for the arithmetic, not a benchmark — it is worth about $3.3 bn, and at that same price the peak year prints a multiple of just 2×. That is how a refiner comes to look cheapest exactly when its margin is about to revert. The figures illustrate the method and value no real company.
4.3 Macro regimes, rates & correlations
Downstream’s macro profile is the near-inverse of an oil producer’s, and the difference is the reason to own it as a distinct exposure.
The oil price is not the driver — the margin is, and the two often move apart. Section 2.6 showed why: the crack is a spread, so the level of crude cancels out except through five side doors. What that means for a portfolio is the sign. Producers are long the barrel; refiners are long the gap, which widens most reliably when crude falls faster than products (Table 23, second row) and can narrow when crude spikes ahead of demand (third row). Refiners are therefore a diversifier alongside a producer position, not a hedge against a high oil price.
Demand strength is the dominant variable. Refining is a bet on people driving, flying and shipping. A strong economy with tight refining capacity is the sector’s ideal regime; a recession that destroys fuel demand is the worst. Interest rates matter less than for midstream because downstream is less leveraged and less bond-like, though refiners’ large capital-return programmes make them sensitive to the cost of capital at the margin.
The worked episode — refiners amplify the margin, not the barrel. The margin swing of 2020–22 (Section 2.5) shows up in the equities as amplification. In the 2022 super-margin window the same companies that had posted heavy losses in 2020 produced the best results in their history: Valero and Marathon each generated about $26 of free cash flow per share (operating cash flow less capital spending, per diluted share). The equities followed violently — from the end of 2020 to the end of 2022, Valero returned about 145% and HF Sinclair about 107%, dividends included, while Brent rose about 60%. That is the signature of the sector: it does not track the oil price, it amplifies the refining margin, in both directions.
Table 23. Downstream across macro regimes
| Regime | Typical downstream response | Why | Historical reference |
|---|---|---|---|
| Strong demand, tight capacity | Very favourable | Cracks blow out; refiners mint cash | 2022 super-margin |
| Weak crude, resilient product demand | Favourable | Feedstock cheapens faster than products | Periods of falling oil, firm fuel |
| High crude, demand-led | Mixed | Margins can compress if products lag | Oil spikes with soft demand |
| Rising real rates | Mildly challenged | Higher cost of capital on buybacks/capex | 2022–23 hiking cycle |
| Recession / demand destruction | Severely negative | Fuel demand falls; margins collapse | 2020: refiners posted losses |
| Structural overcapacity | Negative | New Asian/Mid-East plants flood the market | Ethylene downcycle, 2024–25 |
Source: 2022 free cash flow per the Valero and Marathon Petroleum full-year 2022 results releases, Metal Pilot arithmetic (Valero: $12.574 bn operating cash flow less $2.259 bn capital investments attributable to Valero, ÷ 396m diluted shares; Marathon: $16.4 bn less $2.788 bn capital expenditures, ÷ 516m); share returns are year-end 2020 to year-end 2022 on dividend-adjusted prices per Macrotrends (Valero, HF Sinclair, HollyFrontier before 2022), and Brent is the U.S. EIA weekly spot price for the last week of each year ($50.82 to $81.83); regime responses are substantiated by the episodes cited. Historical, not predictive; correlations are regime-dependent. Cross-commodity regime behaviour is covered in the macro regime guide.
4.4 Value drivers & cycles
Seven drivers move downstream economics, and none of them is the daily oil price. Five are developed earlier in the guide — product demand (split by fuel, because each answers to a different part of the economy), refining capacity, crude differentials, product specifications and the demand peak — and Table 24 points to each one’s home section and the metric to watch. Two are new here: capital discipline, because a payout funded by mid-cycle rather than peak cash flow is the one lever management controls, and the petrochemical cycle, which cushions or drags a refiner depending on how integrated it is. The risk that hangs over all seven is that this time the demand peak arrives before the next capacity shortage, breaking the cycle that has always rescued the sector (Section 5).
Table 24. Downstream value drivers
| Driver | Direction of effect | Why | What to watch | Section |
|---|---|---|---|---|
| Diesel demand | Positive | The world’s largest product; tracks freight and industry, so it moves with the economy | Freight volumes, industrial output | 2.2, 2.4 |
| Gasoline demand | Positive | The largest US product; tracks car travel | Miles driven, electric share of new-car sales | 2.2, 5 |
| Jet demand | Positive | The transport fuel still growing | Air-passenger traffic | 2.2, 5 |
| Refining capacity | Two-way | Closures tighten; new plants loosen | Closure announcements, Asian final investment decisions (FIDs) | 2.3 |
| Crude differentials | Positive for complex plants | Cheap heavy/sour crude widens margin | WCS, sour-sweet spreads | 1.4, 2.5 |
| Product specifications | Positive for complex plants | Raises the bar simpler plants can’t meet | IMO, EPA, LCFS rules | 2.6 |
| Demand peak | Negative long-term | Caps terminal value of assets | EV penetration, jet recovery | 5 |
| Petrochemical cycle | Two-way | Integration cushions or drags | Ethylene operating rates | 3.2 |
| Capital discipline | Positive | Mid-cycle payout avoids peak traps | Buybacks vs leverage | 4.1, 4.2 |
Source: drivers derived from the concluded episodes cited in Sections 2.5 and 4.3 and from company capital-allocation disclosures, 2020–2026; the three product-demand rows follow the product ranking and end uses in Table 11.
4.5 Risks, controversies & ESG
Downstream’s risks are real, and most of them are not oil-price risks.
Margin cyclicality is the defining risk — the spread is uncontracted and mean-reverting (Section 1.7), and the trap it sets for a buyer is described in Section 4.1.
Overcapacity is the structural threat. The ethylene downcycle in Section 3.2 — a global operating rate near 84% and more than a quarter of capacity at some risk of closure — is the live case, and the same glut can hit fuels if Asian refining outruns demand.
The demand peak is the long-horizon controversy, and it is genuinely two-sided. The bear case is that a 40-year refinery faces a fuel-demand peak inside its life (the IEA numbers are in Section 5), so today’s cash flow overstates terminal value. The bull case is that the peak is slow and uneven: jet fuel and petrochemical feedstock keep growing, diesel declines only slowly, the developing world keeps motorising, and Western closures could keep the surviving refineries profitable long after total demand rolls over. Both positions rest on assumptions about 2040s demand no one can verify.
Safety, emissions and transition capex are the operational trio. Refineries are heavy industrial sites with fire, explosion and toxic-release risk; a single incident can take a plant offline for months in a strong margin market. Refining is carbon- and energy-intensive, making it a target of emissions policy and a large decarbonisation-capex burden. And the pivot to renewable diesel and SAF requires capital whose returns depend on policy credits (RINs, LCFS) that can change with an election.
Table 25. Downstream risk register
| Risk | Likelihood | Impact | Who is most exposed | Mitigant |
|---|---|---|---|---|
| Margin collapse / cyclicality | High | High | Independent refiners | Mid-cycle balance sheet, integration |
| Structural overcapacity | High | Medium | Simple / high-cost refiners & crackers | Complexity, low-cost feedstock |
| Fuel-demand peak by 2030s–40s | Medium–high | High | Gasoline-heavy refiners | Diesel/jet/petchem slate, rationalisation |
| Regulatory / carbon cost | High | Low–medium | European refiners | Efficiency, renewable-fuel pivot |
| RIN / policy-credit swings | Medium | Medium | Renewable-fuel investors | Diversified compliance strategy |
| Safety / operational incident | Low per asset | High | All refiners | Turnaround discipline, integrity spend |
| Crude-differential compression | Medium | Low–medium | Complex refiners | Feedstock flexibility |
Source: risks derived from the concluded episodes cited above — the 2020 losses, the 2022 super-margin, the 2025 closures and the ethylene downcycle. Likelihood and impact are the author’s qualitative assessment, not a modelled probability.
Figure 14. Downstream risk map
Rare
Likely
Source: data in Table 25. Each point prints its likelihood × impact score; the shaded region is the high-likelihood, high-impact quadrant. Ratings are the author’s assessment, not disclosed figures.
5. Future outlook & catalysts
Three forces will decide downstream’s next decade, and for the first time they point in genuinely different directions.
The demand plateau is the defining backdrop. The IEA expects refined-product demand to peak around 2027 at roughly 86 MMbbl/d, barely above 2024, before gasoline and diesel decline faster than jet and petrochemical feedstock grow. Global gasoline demand is plateauing — about 27.2 MMbbl/d in 2024 and an estimated 27.4 MMbbl/d in 2025 — and the IEA projects it easing to about 26.2 MMbbl/d by 2030 as electric vehicles displace an expected 5.4 MMbbl/d of oil demand. Diesel faces the same squeeze as trucks switch to liquefied natural gas and batteries — the IEA singles out China’s gas-fuelled truck fleet — while jet fuel keeps growing with air travel and naphtha with plastics (Table 11). This does not mean refining ends — it means the growth is gone, and value shifts to the survivors and to the products that keep growing.
The geography shift is the largest active change. The map in Section 2.1 keeps tilting east: out to 2030 the IEA projects about 4.2 MMbbl/d of new refining capacity, partly offset by 1.6 MMbbl/d of closures — net growth of some 2.6 MMbbl/d, led by China and India while Europe and the United States shut plants. The result is a bifurcated industry: high-cost Western capacity rationalising into fewer, more profitable survivors, and low-cost Eastern capacity expanding into a market that is barely growing overall.
Petrochemicals are the only real growth engine (Section 3.2), which is why every new mega-project — Jamnagar, Ras Laffan, Shaheen — is refining integrated with chemicals, and why a standalone simple refinery is the least attractive asset in the sector.
Table 26. What to watch
| Catalyst / theme | Timing | Why it matters | Watch this metric |
|---|---|---|---|
| Western refinery closures | Ongoing | Tightens the market for survivors | Announced shutdowns, utilisation |
| Refined-product demand peak | ~2027 | Caps sector-wide growth | IEA/EIA demand outlook |
| Gasoline demand decline | From ~2025 plateau | Erodes the largest product | EV penetration, miles driven |
| Diesel demand shift | Ongoing | The world’s largest product loses trucking | Gas-fuelled and electric truck sales |
| Asian & Mid-East capacity | 2026–2030 | Most of net additions, led by China and India | New-refinery start-ups |
| Ethylene / petchem cycle | 2026–2028 | Overcapacity vs new demand | Global ethylene operating rate |
| Renewable diesel & SAF | Ongoing | Policy-driven growth niche | RIN/LCFS values, mandates |
| Crack-spread reversion | Rolling | Cracks revert toward mid-cycle | 3-2-1 crack vs mid-cycle |
Source: demand and capacity outlook, including gas-fuelled trucks, per IEA Oil 2025 and its full report (gasoline 27.2 MMbbl/d in 2024, 27.4 in 2025 and 26.2 in 2030; net additions are author’s arithmetic on the IEA’s 4.2 and 1.6 MMbbl/d); product ranking per IEA Oil Market Report, May 2026; petrochemical share per IEA; margin history per U.S. EIA spot prices (New York Harbor ULSD, Table 15). Timings are agency projections and slip routinely.
6. Summary
Downstream oil and gas is the conversion-and-sale end of the hydrocarbon chain — refining, fuel marketing and petrochemicals — and it earns a processing margin, not a fee or a commodity price. Because that margin is a crack spread and not a contracted toll, downstream is the most violently cyclical part of the energy complex: refiners lost money in 2020 and posted the best results in their history in 2022, with no change in the assets — only in the margin. Per barrel it is thin, and the oil price reaches earnings only through side doors — inventory, energy, byproducts, differentials and demand.
The asset base is shrinking in the West and expanding in the East: China has overtaken the United States as the world’s largest refiner, and the US now runs just 130 refineries, down from 301 in 1982, at roughly unchanged capacity. Complexity, not size, is the durable moat — a plant’s configuration decides which crude it can profitably run and which byproducts it must dispose of.
Five archetypes cover the listed universe: the merchant refiner is the pure crack-spread play, the integrated major hedges it with upstream, the petrochemical-led name rides the polymer cycle, the convenience chain earns a margin that behaves like retail, and the national oil company runs refining as infrastructure. Their balance sheets read like heavy industry — capacity and complexity first, then turnarounds, then working capital. Product travels further, from fewer export complexes, with inventories as the tightness signal; and in a portfolio downstream is the near-inverse of a producer — long the gap, not the barrel — with equities that amplify the margin in both directions.
Shortage shows up as a margin spike, not a crude shortage, and it is self-created: the closures a downturn forces set up the next super-margin. Regulation shapes the product, not the price — the RFS, Tier 3 and IMO 2020 all reward complex refineries. The forward case is a demand peak around 2027, capacity migrating to Asia, and petrochemicals as the only real growth left. The disciplined way to own it is on mid-cycle margins, never on the peak-earnings print that makes a refiner look cheapest exactly when it is most dangerous.
To screen the refiners, marketers and petrochemical producers behind this map by capacity, complexity and configuration, explore Metal Pilot’s downstream oil & gas stock screener.
Next in the series → Part 4: Oilfield Services — the drillers, pressure pumpers and equipment makers that producers pay to get the barrel out of the ground in the first place.
Terms in this post (66)
- Alkylation
- Joins light gases into high-octane gasoline using an acid catalyst. The refinery’s main consumer of sulphuric acid, returned as spent acid.
- API gravity
- A density scale for crude oil that runs backwards: a higher reading means lighter crude. Light crude yields more high-value fuels and usually sells for more.
- Asphaltenes
- The heaviest, tar-like molecules in crude, which drop out when incompatible crudes are mixed. Fouling that limits which grades can be blended.
- Atmospheric / vacuum distillation
- The primary column that splits crude by boiling point, and the second column that splits its heavy residue under reduced pressure. The base measure of capacity, and the feed to the conversion units.
- Base oil groups
- The API’s Group I–V classes of lubricant base stock. Higher groups need hydroprocessing and earn a premium.
- Brent
- The North Sea crude benchmark that prices most of the world’s internationally traded oil. In depth: Oil market guide — Pricing & benchmarks
- Bunkers / bonded fuel
- Fuel sold to ships (marine bunkers) or aircraft on international voyages; bonded fuel is supplied from customs-bonded storage, free of domestic duty. Counted as demand, not exports, in JODI data but as exports in China’s customs data.
- Calendar-day / stream-day capacity
- Capacity averaged over a year with downtime (b/cd, the figure EIA reports) vs the flat-out maximum (b/sd, about 6% higher). Confusing them overstates capacity.
- Coking / petroleum coke
- The unit that extracts the last liquids from the barrel bottom, and the solid carbon it leaves, sold as fuel, anode or needle grade. Lets a plant run the heaviest crude; the coke drags realised margin.
- Contango
- A futures curve on which later-dated prices sit above spot. A rolled long position loses money in contango, which often signals ample supply.
- Conversion
- Cracking, coking and reforming that upgrade heavy molecules to light ones. Where refining margin is created.
- Crack spread / 3-2-1 crack
- The margin between crude and the products made from it; the 3-2-1 version turns 3 barrels of crude into 2 of gasoline and 1 of distillate. The refiner’s real revenue driver and its standard benchmark.
- Cracking (FCC, hydrocracking)
- Units that break heavy molecules into lighter ones — with a catalyst (fluid catalytic cracking, mainly gasoline) or with hydrogen (hydrocracking, mainly diesel and jet). The core conversion units.
- Crude assay / straight-run yield
- The lab profile of a crude — cuts, sulphur, metals, acidity — and what distillation alone recovers of each cut. Defines a refinery’s crude envelope and how much conversion a crude needs.
- Crude differential
- The price gap between crude grades (e.g. WCS–WTI). The discount a complex refiner captures as margin.
- Crude quality (API gravity; sweet / sour)
- Density — high API is light, low API heavy — and sulphur content, low (sweet) or high (sour). Heavy and sour crudes are discounted and need a complex refinery.
- Cuts: naphtha, VGO, vacuum residue
- Boiling-range slices of crude: naphtha (up to about 150 °C), vacuum gasoil ( about 370–550 °C) and the residue above about 550 °C. Gasoline and petrochemical feed; the cracker’s main feed; coker feed, asphalt and fuel oil.
- Distillate / ULSD / cetane
- Diesel and heating oil; ultra-low-sulphur diesel holds 15 ppm sulphur or less; cetane rates how quickly diesel ignites under compression. The second-largest US product and the largest worldwide; ULSD requires deep hydrotreating.
- Downstream / refining
- The refining, marketing and petrochemical stage after production and transport; refining is the conversion of crude into finished products. Defines a business that sells a manufactured product, not a commodity.
- EU ETS
- The EU Emissions Trading System, which prices refinery CO₂. Adds cost and pressures European margins.
- Final investment decision
- A company’s formal commitment to fund and build a project. Capital is committed from that point, so a project is far more certain after it.
- Frac spread
- The value of natural gas liquids less the cost of the gas used to extract them — the margin of a gas processor.
- Gasoline pool (reformate, alkylate, RBOB / CBOB, octane)
- High-octane components blended into blendstocks that are finished with ethanol at the terminal; octane is the resistance to auto-ignition (knock) that separates grades. The form in which US refiners make and ship gasoline.
- Hydrotreating
- Using hydrogen to strip sulphur, nitrogen and metals from refinery streams so fuels meet their specifications.
- IMO 2020
- The 0.5% marine-fuel sulphur cap. Rewards complex refineries.
- Integration margin
- The spread across refining + petrochemicals. Cushions the pure crack spread.
- Isomerisation / isomerate
- Rearranges straight-chain light naphtha and butane into branched, higher-octane molecules; the product is isomerate. A gasoline blendstock, and a source of isobutane for alkylation.
- Jet A / Jet A-1
- Kerosene-type jet fuels freezing at −40 °C / −47 °C or below; Jet A-1 also holds a flash point of at least 38 °C and at least 42.8 MJ/kg of energy. The US and international aviation grades.
- Jones Act
- US coastwise-shipping law. Raises inter-US product freight cost.
- LIFO / FIFO; replacement-cost (CCS) earnings
- Last-in-first-out vs first-in-first-out inventory costing; profit restated at the current cost of supply. Decide how much of an oil-price move hits reported profit.
- Linear-programming (LP) model
- The optimiser a refinery uses to pick the most profitable crude mix and product slate its units allow. How a plant re-optimises its crude basket as differentials move.
- LPG / NGLs
- Liquefied petroleum gas (propane, butane), and natural gas liquids — the ethane, propane and butane separated from natural gas. Heating, alkylation and petrochemical feed.
- Marketing margin
- The wholesale/retail fuel margin (¢/gal). The stable end of downstream.
- Mid-cycle
- A level typical of a whole commodity cycle rather than of today’s price; where the base deck sits far from its five-year average, the target multiples are held at mid-cycle. In depth: The Oil & Gas Value Chain: Downstream — How to value & screen downstream
- Mid-cycle margin
- The normalised margin across a full cycle. The correct basis to value a refiner on.
- MMbbl, MMbbl/d
- Million barrels; million barrels a day.
- MMBtu
- One million British thermal units, the energy unit North American gas is priced in; roughly a thousand cubic feet of gas.
- Nelson Complexity Index (NCI)
- A measure of a refinery’s conversion capacity relative to distillation. Higher = can profit from cheap heavy crude.
- Net debt / EBITDA
- Net debt divided by a year’s EBITDA: how many years of cash earnings would repay the debt. Below 1× is low; net cash means there is none. Lower is stronger. In depth: The Oil & Gas Value Chain: Upstream — How to read & screen a producer
- Net exporter / trading hub
- A country whose product exports exceed its imports; a hub, such as Singapore or the Netherlands, imports, blends and re-ships product, so its gross exports overstate its own refining. Net, not gross, trade shows who actually supplies whom.
- NGLs (natural gas liquids)
- Ethane, propane, butanes and heavier liquids separated from produced gas, sold at a fraction of the oil price. In depth: The Oil & Gas Value Chain: Upstream — The product slate & crude quality
- Non-controlling interest
- The share of a consolidated subsidiary that others own; deducted so the value counted is the parent’s shareholders’ alone.
- Off-ladder plants
- Condensate splitters built for ultra-light feed, and crude-to-chemicals complexes that send much of the barrel to a steam cracker. Plants built for a different job than fuels refining.
- Operating costs
- What it costs to run the business before depreciation — for a producer, the cash cost of operating its mines, wells and plants, before royalties, transport and overhead; as an income-statement total, every operating cost line, overhead included — in total or per unit, as the cell gives it. Lower is better. In depth: The Oil & Gas Value Chain: Upstream — What a barrel costs
- Operating income
- Revenue less the costs of running the business — the cost of sales or operating costs, depreciation and depletion, selling, research, exploration and administration — before interest and tax.
- Processing gain
- The extra product volume from making lighter, hydrogen-enriched products. Lets a refinery sell more barrels than it buys.
- Realised margin / operating income per barrel
- Product revenue less crude and feedstock cost per barrel; what is left after operating expenses and depreciation. What a company actually captures and keeps, versus the benchmark crack.
- Refinery configuration
- The set of units a plant is built with — topping (distillation only), hydroskimming (little conversion), cracking, deep conversion. Decides which crude grades it can profitably run; simple configurations close first.
- Refinery yield
- The mix of products made per barrel. Determines which demand a refiner is exposed to.
- Renewable diesel / SAF
- Diesel and sustainable aviation fuel made from fats, oils and other renewable feedstock. Policy-driven growth whose returns depend on credits.
- Residual and marine fuel oil (HSFO, VLSFO, ULSFO)
- The heavy bottom product, now split into 3.5%, 0.5% and 0.1% sulphur marine grades. Devalued by IMO 2020; HSFO needs a scrubber.
- RFS / RIN / LCFS
- The US Renewable Fuel Standard; its tradable Renewable Identification Number credits; and state low-carbon fuel standards whose credits stack on top. A compliance cost for non-blenders, and the economics of renewable diesel.
- Sour water
- Process water carrying H₂S and ammonia. Stripped before treatment; its H₂S goes to the sulphur plant.
- Steam cracker / ethylene & propylene
- The core petrochemical unit, and the building-block olefins it makes for plastics. The petrochemical demand engine.
- Steam methane reformer (SMR)
- Makes hydrogen from natural gas, emitting CO₂. Ties hydrogen cost to gas prices; a carbon-capture target.
- Still gas
- Light gas from refinery units, burned as fuel. Cuts purchased energy; about 4% of US output.
- Stream
- An upfront payment for the right to buy a share of a mine’s future metal at a fixed, low price: the operator gets capital now, the stream holder metal later. In depth: Silver market guide — Company archetypes along the value chain
- Sulphur recovery (amine unit, Claus, SRU, WSA)
- The chain that scrubs hydrogen sulphide (H₂S) out of gas streams and converts it to elemental sulphur (Claus process, sulphur recovery unit) or directly to sulphuric acid (wet sulphuric acid unit). Often the real cap on how sour a crude a plant can run.
- Super-margin
- This guide’s term for a window in which cracks run far above their long-run average, as in 2022. The earnings a buyer must not mistake for a new baseline.
- Tier 3 gasoline
- The US gasoline standard averaging 10 ppm sulphur. A specification that needs hydrotreating simpler plants can skip.
- Total acid number (TAN)
- A measure of a crude’s naphthenic acid content. High-TAN crude corrodes steel; caps which grades a plant can buy.
- Turnaround
- A planned major maintenance shutdown of a refinery or plant, lasting weeks.
- Upgrader / synthetic crude
- A plant that cokes or hydroconverts oil-sands bitumen into a light, sweet synthetic crude oil. Takes barrels out of the discounted heavy pool that complex refiners compete for.
- Utilisation
- Throughput ÷ operable capacity. The refining tightness signal.
- Visbreaking
- Mild thermal cracking that lowers the viscosity of heavy residue without converting much of it. The low-cost alternative to a coker: it makes fuel oil saleable rather than turning residue into light products.
- WTI
- West Texas Intermediate, the US crude benchmark, delivered at Cushing, Oklahoma. In depth: Oil market guide — Pricing & benchmarks
7. Sources, methodology & disclaimer
7.1 Sources, methodology & data vintage
Agencies and regulators. U.S. EIA Refinery Capacity Report and number of operable refineries; EIA — US refining capacity decreased during 2025; EIA — refinery yield; EIA — an introduction to crack spreads and U.S. EIA spot prices for New York Harbor ULSD, Brent, Gulf Coast gasoline, Gulf Coast ULSD and WTI, with refinery utilisation and operable capacity; EIA — refineries vary by level of complexity; U.S. EIA Country Analysis Brief: Venezuela for Paraguaná; U.S. EIA — refining crude oil for the refinery process sequence; IEA Oil 2025 and its full report for the demand, capacity and petrochemical outlook; IEA Oil Market Report, August 2026 for the 2026 margin run; IEA Oil Market Report, May 2026 for world demand by product; U.S. EIA — use of diesel, use of gasoline and uses of hydrocarbon gas liquids for end uses; US EPA Renewable Fuel Standard; IMO 2020 sulphur cap and IMO press briefing on its outcome; EIA Weekly Petroleum Status Report; EIA — refinery yield by product, octane and factors affecting gasoline prices; USGS Mineral Commodity Summaries 2026 — sulfur and 2024 — vanadium; US EPA AP-42 §8.13 sulfur recovery, diesel fuel standards and Tier 3 fuel standards; OSHA Technical Manual — Petroleum Refining Process for refinery unit operations and operating conditions; US DOE on Port Arthur CO₂ capture; CRS — Petroleum Coke (R43263); IFRS — IAS 2 Inventories; JODI Oil World Database and its manual for refinery output and trade by product and country, with the IEA Oil Market Report, April 2026 for Russian product exports and U.S. EIA — where hydrocarbon gas liquids come from for the origin of US LPG.
Industry bodies and market data. Synthetic crude quality per CrudeMonitor and Alberta’s upgraded bitumen share per the Alberta Energy Regulator (ST98); Energy Institute Statistical Review of World Energy 2026 for refining capacity by country and the US share of product trade, with the capacity values confirmed through Energy Factbook; NACS / Convenience Store News Top 100, 2025; Wood Mackenzie on the steam cracker closure threat; Macrotrends dividend-adjusted share prices for Valero and HF Sinclair; ExxonMobil crude assays for straight-run yields; S&P Global Platts Americas crude specifications and S&P Global on the WTI grades for the WTI Midland specification and WTI’s Permian origin; Topsoe on WSA; Veolia on sulphuric acid regeneration; Bryan Research & Engineering on sour water; API 1509 Annex E for base-oil groups; ANSI on ASTM D1655 and DEF STAN 91-091 Issue 16 for jet fuel; DieselNet on ignition quality for cetane; TotalEnergies main indicators for the European refining margin marker; S&P Global on China’s bonded jet fuel and bonded fuel-oil exports.
Company sources. Sinopec 2025 annual results, ExxonMobil 10-K 2025, Aramco Annual Report 2025, Valero, Reliance, SK and HF Sinclair for refining capacity and signature assets, and TotalEnergies, Shell, Chevron, PBF Energy and Phillips 66 disclosures, 2025–26, for their assets; the Valero and Marathon Petroleum 2022 results for free cash flow; Valero’s full-year earnings releases for 2021, 2023 and 2025 for the per-barrel refining economics.
Methodology. Refining capacity is atmospheric distillation capacity from EIA operator reports (US) and the Energy Institute (global) at the latest published year; because these publish on different cycles, some figures carry different vintages and are labelled where material. Country and company capacity figures use differing definitions (nameplate, calendar-day, gross vs net interest), so they indicate scale rather than a like-for-like ranking, and Russia’s nameplate overstates sanctioned utilisation. Product yields are EIA shares of refinery input for 2025, so products sum to about 106% before the processing gain; the product-slate figure rescales them to shares of output. World demand by product is the IEA’s 2025 estimate from its May 2026 Oil Market Report, whose total includes LPG, ethane and other products outside the refined-product count of about 86 MMbbl/d; the end-use and substitute columns of Table 11 are the author’s qualitative reading where no source is named. Product output and trade by country are JODI annual averages of the twelve 2025 monthly figures, retrieved on 25 September 2026, for the countries that report — Russia, the UAE, Iran and Iraq do not — with Russia’s exports taken from the IEA’s tanker-tracking estimate; exports are gross, so a hub’s re-exports are included, and China’s jet and fuel-oil exports include bonded bunker sales. Straight-run yields are single ExxonMobil assays grouped into five cuts by volume, with butane and lighter taken by difference. Per-barrel refining economics are Valero’s reported refining-segment figures (operating income as adjusted), each year taken from the release that prints it; the net-income-to-revenue ratio is Metal Pilot arithmetic. The sulphur-per-barrel example and the mid-cycle valuation example are hypothetical illustrations. The crack spread is a gross-margin proxy, not a refiner’s realised margin; the 3-2-1 average and the diesel crack are author’s arithmetic on EIA spot prices. Refiner free-cash-flow and equity-return figures are for concluded windows and are settled history; the 2026 margin run was still in progress at the data date and is treated as live, not settled. Demand forecasts — the refined-product peak around 2027, the 2030 gasoline, capacity and petrochemical figures — are IEA projections and are labelled as such. No market capitalisations, share prices or enterprise values appear anywhere in this report, by design.
Data as of 18 August 2026. Intended update cadence: annually, after the EIA Refinery Capacity Report and the IEA Oil report.
7.2 Disclaimer & disclosure
This report is for informational purposes only and is not investment advice, a recommendation, or an offer to buy or sell any security. Downstream companies carry margin, cyclical, regulatory, operational and commodity risks, and capacity, complexity and margin figures are estimates as of the stated date that change with each reporting cycle. Historical margin and equity behaviour is presented as settled history; past performance is not indicative of future results, and the correlations and regime patterns described here are sample-dependent and can break down, particularly in a crisis. Readers should do their own research and consider consulting a licensed financial adviser before acting. This report was prepared with the assistance of AI, consistent with the ai = true flag in its front matter; figures were sourced from the references above and reviewed, but readers should verify any figure they intend to rely on. The author holds no position in any company named in this report.