US Refinery Capacity in 2026

The United States had 18.2 million barrels per calendar day of operable crude distillation capacity at 130 refineries on January 1, 2026, according to the EIA's annual Refinery Capacity Report. That is 263,000 b/cd, about 1 percent, less than a year earlier, after two closures: LyondellBasell in Houston and Phillips 66 in Los Angeles. A third, Valero's Benicia plant in California, is still inside the January count but stopped refining afterward.

Those three facts, the total, the count, and the closures, are what most headlines repeat. The more useful material sits one layer down: how the capacity is split across the five Petroleum Administration for Defense Districts, what kind of capacity disappeared, and why the national total hides a regional squeeze. This guide walks through the report the way an analyst would read it, with the actual table values, and shows how the annual capacity number connects to the weekly utilization rate that moves fuel markets. All figures below come from EIA publications checked on October 22, 2026; the source list is at the end.

What the EIA actually measures

The headline number is operable atmospheric crude oil distillation capacity. Every refinery starts with an atmospheric distillation unit, the tall column that heats crude and splits it by boiling range. Because every barrel passes through that unit first, its size is the standard way to state how much crude a refinery can take in. The downstream units that convert heavy fractions into gasoline and diesel are counted separately, and they tell a different story, covered further down.

"Operable" has a precise meaning in the EIA glossary. Operable capacity is the sum of capacity that was operating at the start of the period, capacity that was idle but could be placed in operation within 30 days, and capacity that was down for repairs that could be completed within 90 days. Operating capacity is only the first slice. A plant that has announced a permanent shutdown and dismantled its units is not operable; a plant waiting on a pump repair is.

The second distinction is time. A barrel per calendar day(b/cd) figure is the operator's estimate of what the distillation unit can process in 24 hours under usual conditions, allowing for normal downtime such as maintenance and slowdowns. A barrel per stream day(b/sd) figure is what the unit can process running flat out with no allowance for downtime. The EIA says stream-day capacity is typically about 6 percent higher. In the 2026 report the national stream-day total was 19,157,174 against 18,160,493 calendar-day barrels, a gap of 5.5 percent. When two sources disagree on a refinery's size, a calendar-day versus stream-day mismatch is the first thing to check.

Finally, the date. The survey is a snapshot as of January 1, published in June. The 2026 report came out June 26, 2026, and the EIA's summary followed on June 29. Anything that happens after January 1 waits a full year to show up in the annual table, which is the root of the Benicia wrinkle explained below.

The 2026 count in one table

Here is the national picture from the EIA's capacity table, with the prior year beside it so the change is visible.

Source: EIA, Number and Capacity of Operable Petroleum Refineries, annual as of January 1 (checked October 22, 2026). Change column computed by PetroEyes.

Two details in that table are easy to miss. Idle capacity fell even as the idle refinery count rose from one to two: the idle plants on January 1, 2026 are small, together only 34,000 b/cd. And operating capacity fell by less than operable capacity, because part of the 2025 idle capacity either restarted or left the count. Reading only the top line misses both.

Where the barrels are: capacity by PADD

The country is split into five PADDs for petroleum statistics: the East Coast (1), Midwest (2), Gulf Coast (3), Rocky Mountains (4), and West Coast including Alaska and Hawaii (5). If the district map is new, the PADD districts explainercovers the boundaries and why they still matter. Here is the 2026 capacity by district, with refinery counts and each district's share of the national total.

Source: EIA operable atmospheric crude distillation capacity and refinery counts by PAD District and state, annual as of January 1 (checked October 22, 2026). Shares and changes computed by PetroEyes.

The table says three things plainly. First, the Gulf Coast is the refining system: more than half of all capacity and 58 of 130 plants. Texas alone holds 6,134,650 b/cd in 34 refineries and Louisiana another 3,005,373 in 15. Second, every district except the Gulf Coast and the West Coast added capacity in 2025, mostly through small expansions at existing plants; the Midwest gained almost 35,000 b/cd without a new refinery. Third, the two districts that lost capacity lost it for the same reason, a single closure each, but the consequences differ because of scale.

The EIA puts it in percentages: the Los Angeles closure removed about 5 percent of West Coast capacity, while the Houston closure removed about 3 percent of Gulf Coast capacity, and after offsetting gains the Gulf Coast total fell by less than 2 percent. A 3 percent loss in a district with 58 refineries and pipelines to every other region is absorbed. A 5 percent loss in a district that is largely cut off from the rest of the country by mountains and distance is not, as the next sections show.

The closures, plant by plant

Source: EIA Today in Energy, "U.S. refining capacity decreased during 2025," June 29, 2026 (checked October 22, 2026).

The state rows let you check the closures against the totals. California's capacity fell from 1,637,871 to 1,499,171 b/cd, a drop of exactly 138,700, the Los Angeles plant. Texas fell from 6,343,466 to 6,134,650, a drop of 208,816; the Houston plant was 263,776, so other Texas refineries added about 55,000 b/cd during the year. That is the "marginal capacity increases at other, existing refineries" the EIA refers to, and it is a pattern worth knowing: US refining capacity changes mostly through closures on one side and debottlenecking at surviving plants on the other, not through new refineries.

The West Coast, five years of shrinkage

Taken one year at a time, the West Coast changes look modest. Lined up across the six reports from 2021 to 2026, they are not.

Source: EIA operable crude distillation capacity by state, 2021 to 2026 (checked October 22, 2026). Alaska and Nevada are part of PADD 5 but are omitted from the columns; they are included in the total.

Over five reports California lost 249,000 b/cd, 14.2 percent of its 2021 capacity, and the West Coast district lost 293,700, or 10.8 percent. Now apply the Benicia adjustment the annual table cannot yet show. Subtracting its 145,000 b/d from the 2026 California figure gives roughly 1,354,000 b/cd of working capacity in the state, about 22.5 percent below the 2021 level. That is a PetroEyes calculation that mixes a calendar-day total with a plant capacity the EIA quotes in b/d, so treat it as an approximation, but the direction is not in doubt.

Why does this matter more than a similar loss in Texas? In the EIA's words (May 5, 2025), California is geographically isolated from other US refining centers, no pipelines supply it from across the Rocky Mountains, and the state mandates a special blend of gasoline. It already imports gasoline from countries such as India and South Korea. When in-state refining shrinks, the gap has to be filled by tankers, which adds transit time and freight cost and makes the market more sensitive to an unplanned outage at the plants that remain. Those are structural reasons, not a forecast of any particular price; for how the refining layer shows up in the pump price, see the gasoline price breakdown.

What vanished downstream: the conversion units

Distillation capacity says how much crude a refinery can swallow. Conversion capacity says how much of that crude it can turn into gasoline, diesel, and jet fuel instead of heavy residue. The EIA publishes downstream charge capacity in barrels per stream day, so the fair comparison is against stream-day crude capacity.

Source: EIA, Operable Crude Oil and Downstream Charge Capacity of Petroleum Refineries, annual as of January 1 (checked October 22, 2026). Percent changes and ratios computed by PetroEyes.

Coking capacity fell 4.9 percent, almost three times the 1.7 percent drop in stream-day crude capacity: 144,070 barrels per stream day of delayed coking left the system in one year. Cokers are the units that let a refinery run heavy, high-residue crude, such as Canadian heavy or Latin American grades, and still make light products. When coking disappears faster than crude capacity, the remaining system loses some of its appetite for heavy barrels in particular. That links this report to the price gap between heavy and light crude covered in the WCS-WTI differential explainer: fewer cokers means fewer natural buyers for discounted heavy crude.

Catalytic cracking, the main gasoline-making unit, fell 2.8 percent. Hydrocracking, which leans toward diesel and jet fuel, barely moved. One-year changes are too small to call a trend in product mix, but they are the right place to look when someone claims that a closure will hit one fuel harder than another.

From the annual table to the weekly utilization rate

The annual capacity figure is a denominator. The number markets watch each Wednesday is the refinery utilization rate in the EIA's Weekly Petroleum Status Report, which the glossary defines as gross input to atmospheric distillation units divided by operable capacity. If the weekly report is unfamiliar, the guide to reading the Weekly Petroleum Status Report shows where the line sits. Because the weekly series uses the monthly capacity estimate rather than waiting for the next annual survey, a closure changes the weekly denominator within weeks.

Worked example: what the denominator does to the headline

Inputs, all from the EIA weekly series, checked October 22, 2026: for the week ending October 16, 2026, gross inputs to US refineries were 16,860 thousand b/d and utilization was 93.4 percent.

1. Implied operable capacity: 16,860 / 0.934 = about 18,050 thousand b/d. Because utilization is rounded to one decimal, the true value lies between roughly 18,040 and 18,060.
2. Compare with the annual figure minus Benicia: 18,160 - 145 = 18,015 thousand b/cd. The implied weekly denominator is within about 35,000 b/d of that, consistent with small capacity additions since January and rounding.
3. Counterfactual: had Benicia's 145,000 b/d stayed in the denominator, the same 16,860 of inputs would print as 16,860 / 18,160 = 92.8 percent utilization rather than 93.4 to 93.6.

The lesson: a refinery closure raises reported utilization by itself, even if not one extra barrel is processed. When utilization looks high after a year of closures, part of that is a smaller denominator, not a busier industry. Inputs in barrels are the cleaner measure of activity.

For context, annual average utilization was 92.0 percent in 2025, the highest since 2018's 93.1 percent, against 90.5 percent in 2024 and 78.9 percent in the pandemic year 2020. In late summer 2026 the weekly rate reached 98.0 percent (week ending August 28) before falling to the 92 to 94 percent range as autumn maintenance began. A system that runs at 98 percent of capacity in peak season has very little room to cover an unplanned outage, and that is the practical meaning of losing capacity: not lower average output, but less slack when something breaks.

The biggest plants and the biggest refiners

The EIA's 2026 summary names the largest refinery two ways. Motiva's Port Arthur refinery is the largest on a calendar-day basis at 656,000 b/cd. Marathon's Galveston Bay refinery is the largest on a stream-day basis at 678,000 barrels per stream day. Both sit on the upper Texas coast. Port Arthur alone is about 3.6 percent of national calendar-day capacity, more than the entire Rocky Mountain district's 653,089 b/cd.

At the company level, the EIA reports that the three largest refiners, Marathon, Valero, and ExxonMobil, each held calendar-day capacity within 1 percent of their 2025 level, that fourth-ranked Phillips 66 shrank with the Los Angeles closure, and that Chevron overtook PBF Energy as the fifth-largest refiner. Company rankings matter to equity readers, but for fuel supply the district table is more informative: who owns a plant matters less than where it is and what pipelines it feeds.

Where readers go wrong with capacity numbers

• Treating capacity as output. 18.2 million b/cd is a ceiling. Actual crude runs in the week of October 16, 2026 were 16.86 million b/d of gross inputs. Product supply questions are answered by inputs, stocks, and trade, not by capacity alone.
• Mixing calendar-day and stream-day figures. A 5 to 6 percent gap can make a refinery look like it expanded when only the unit changed.
• Reading the annual report as current. The January 1 snapshot was published in June and already excludes nothing that happened after New Year. Benicia is the 2026 example.
• Assuming a national 1 percent drop is a national problem. The Gulf Coast lost less than 2 percent and has spare connections. The West Coast lost about 5 percent in one year and is isolated. The same number means different things in different districts.
• Ignoring conversion units. A plant closure removes a particular mix of cokers, crackers, and treaters. The 2026 data show coking falling faster than crude capacity, which matters for heavy crude buyers more than for the headline.
• Over-reading utilization after closures. As the worked example shows, a smaller denominator lifts the percentage by itself.

How capacity connects to refining margins

Refiners earn the difference between what products sell for and what crude costs, the crack spread. Capacity is one of the slow variables behind that spread. When a region loses capacity and demand does not fall by the same amount, the remaining plants run harder, imports fill the gap, and the regional product price tends to carry a larger premium over crude, especially during outages. When capacity is added, the reverse pressure applies. The refinery crack spreads guide explains the 3-2-1 calculation and how to read it; capacity data tell you which regions are likely to show the widest and most volatile spreads, not what the spread will be next month.

It is also worth separating capacity from inventories. Product stocks are the shock absorber for a refinery outage, and the product supplied explainer shows how to turn stocks into days of cover. A region with less refining capacity can still be comfortable if stocks are high and import lanes are open; it becomes fragile when all three, capacity, stocks, and imports, are tight at once.

A five-step check for the next capacity headline

1. Which measure is quoted: operable or operating, calendar day or stream day?
2. As of what date? The annual survey is January 1; monthly estimates can differ.
3. Which PADD, and what share of that district does the change represent?
4. What kind of capacity: distillation only, or conversion units too?
5. Does the change show up in weekly gross inputs, or only in the utilization percentage?

Run any closure or expansion story through those five questions and most of the confusion disappears. The rest of the answer is in the weekly data: inputs, regional stocks, and imports into the affected district.

Questions readers ask about refining capacity

Next step: capacity is the ceiling, and the weekly report shows how close the industry runs to it. The Weekly Petroleum Status Report guide shows exactly where gross inputs and utilization appear each Wednesday, so the arithmetic above can be repeated on next week's numbers.

Sources (all checked October 22, 2026): EIA Today in Energy, U.S. refining capacity decreased during 2025 (June 29, 2026); EIA Refinery Capacity Report 2026 (released June 26, 2026); EIA number and capacity of operable refineries; EIA downstream charge capacity; EIA glossary (operable capacity, utilization); EIA Today in Energy, Why California usually pays more at the pump (May 5, 2025); EIA annual refinery utilization; EIA weekly refinery utilization; EIA weekly gross inputs. Shares, changes, ratios, and the worked example are PetroEyes calculations from those published values.