US refinery closures by year since 2020: the complete record

The dominant narrative around U.S. gasoline prices holds that the country has been steadily “losing refineries.” That is true as a count, and this page compiles every one of them, with capacity, date and primary source. But the capacity arithmetic is more specific than the headline: the gross figure removed by closures is roughly double the net decline, because the two facts are usually reported separately. This record puts both in one place.

The record: every U.S. refinery closed or converted since 2020

Each row is a refinery removed from U.S. operable crude capacity since the January 2020 peak, with its published capacity, the year the change took effect, and whether it was a hard closure or a conversion (to renewable fuels or a terminal). Bars are scaled to the largest entry.

US refinery closures by year, 2019–2026

Year by year, the record reads: one site in 2019 (Philadelphia), five in 2020, one in 2021, none in 2022 or 2023, one in 2024, two in 2025 and one so far in 2026. By capacity, 2020 is the heaviest year, with 466,146 b/d removed across five sites. 2025 comes next, with 402,476 b/d from two hard closures in Houston and Los Angeles.

Each refinery is dated to the year it stopped processing crude. For a conversion, crude runs stop before the renewable unit starts, so a list dated by the end of the conversion files Martinez, Cheyenne and Dickinson later than this one does.

The refinery count moves less than the record. The EIA’s count of operable refineries, read here on January 1 of the following year, includes idle plants (two of the 130 on January 1, 2026) and plants entering or returning to the series: it rose from 129 to 132 during 2023, a year with no removal in this record. Capacity, further down, is the better gauge of what left. The same register taken back three decades, ninety permanent shutdowns in all, is in the site-by-site record of shutdowns since 1990.

Why the net loss is smaller than the closures suggest

A page that showed only the closures would be incomplete. Capacity did not only leave: it was also added. The single largest addition was ExxonMobil’s Beaumont, Texas expansion, which brought roughly 240,000 b/d online in 2023; incremental de-bottlenecking at surviving refineries added the rest. Netting the two sides against the EIA’s annual capacity series gives the honest figure. Reading supply in barrels of capacity rather than in price is the discipline shared across physical commodity markets, from oil and gas to critical minerals.

From the 18,976,085 b/cd peak, capacity stood at 18,160,493 as of January 1, 2026: a net decline of 815,592 b/cd (−4.3%). Valero’s Benicia refinery (145,000 b/d) was still counted on January 1, 2026. The EIA’s monthly capacity series dropped it in March, and by July 2026 operable distillation capacity stood at 18.04 million b/cd, about 932,000 b/cd (−4.9%) below the peak. Set against roughly 1.58 million b/d of gross removals already reflected by January 2026, that implies about 763,000 b/d of offsetting expansions and de-bottlenecking. Gross removals are near 9% of the peak; the net decline is near 5%.

Three waves

What the record does, and does not, show

The count is unambiguous: eleven refineries left U.S. crude processing since the 2020 peak. What it does not establish is a proportional supply shortfall. Net operable capacity fell far less than the gross closures, surviving refineries run at high utilisation and greater complexity, and the United States has remained a net exporter of refined products throughout the period. The long-run pattern the EIA has documented for decades (fewer, larger, more complex refineries) continued rather than broke. A legitimate reading of the same data is that capacity concentrated, not simply disappeared. Any national shortfall would register against how much oil the US consumes each day, not against a refinery count. Where closures matter most is regional: the Phillips 66 Los Angeles closure alone removed about 5% of West Coast (PADD 5) capacity, and California accounts for the largest state share of the total removed. Compare with the refined-barrel margin replayed across forty years.

Methodology & scope

What counts. A facility is included when it permanently stopped processing crude oil since the January 1, 2020 operable-capacity peak: whether by outright closure or by conversion to renewable fuels or a terminal. Temporary idling that later reversed is excluded.

Capacity basis. Figures are operable atmospheric crude distillation capacity (barrels per calendar day) from the EIA Refinery Capacity Report, except Alliance and the San Francisco (Rodeo + Santa Maria) system, which use the operator’s published net crude throughput. The difference between the two bases is small and noted per row.

The 2020 bound. The starting point is the EIA operable-capacity peak, January 1, 2020 (18,976,085 b/cd). PES is dated to its 2019 physical shutdown but was carried as idle in the January 2020 figure and removed by the EIA in the 2021 report, which is why the capacity peak is 2020 despite PES stopping in 2019.

Excluded edge cases. Limetree Bay / St. Croix, U.S. Virgin Islands (~210,000 b/d) is excluded: it had been idle since 2012, ran only briefly in early 2021, and sits outside the 50-state operable series. Calcasieu Refining (Lake Charles, LA) is excluded because it was not treated as a permanent removal in the EIA series.

Data: U.S. Energy Information Administration, Refinery Capacity Report; company SEC filings; primary trade press. Eco3min text and analysis: CC BY 4.0. Data, files and charts remain subject to their producer’s terms, stated on each page; Eco3min grants no licence over third-party data.