US propane stocks stood at 109.09 million barrels in the week to 11 September — an all-time record for the date, and 21 per cent above the five-year average. Every market note written this autumn will quote it.

Divide it by the market it supplies and the surplus disappears. On days of cover, US propane sits at 37.0 days against a five-year average of 37.6 — an entirely ordinary position.

Not a glut. Not a shortage. A normal market wearing a record number — which is why, in the same week the record was set, stocks drew 1.4 million barrels and Mont Belvieu rallied 13.5 per cent.

+21%Stocks against the five-year average — the number everyone quotes
−1.7%The same position measured in days of cover
92%Of the surplus sitting on the Gulf Coast, not where winter bites

The short version

  • The record is real. 109.09 MMbbl, an all-time high for the date, +21% on the five-year average.
  • But exports hit a record 2,213 kb/d — 52% above the five-year average. On a four-week basis, cover is 37.0 days against a 37.6 norm. The entire surplus is absorbed by throughput.
  • 92% of it is Gulf Coast. PADD 3 is +40.5% and at an outright record. PADD 2 Midwest — where heating demand lives — is +5.3% and entirely ordinary.
  • Seven winters of Midwest draws have run 14.4 to 20.2 MMbbl against 26.56 on the ground. There is no heating cushion.
  • And the demand side is the risk: NOAA puts a >90% chance of a very strong El Niño this winter, with a 69% chance of an event exceeding anything since 1950. El Niño winters draw the Midwest roughly 9% lighter.
  • And it was inherited, not accumulated: 2026 injected at the slowest rate in eight years. The year simply started 25 MMbbl long.
  • In the record week, stocks drew 1.4 MMbbl and Mont Belvieu rose 13.5% to $0.855/gal, 10.5% above a year ago. The market is not pricing a glut.
EIA Weekly Petroleum Status Report, week ending 11 September 2026 · five-year averages are 2021–2025 · regional split is the 4 September reading, the latest published by district

A Record Measured Against the Wrong Thing

Inventory is only meaningful relative to the throughput it serves. A tank holding thirty days of demand in 2019 and thirty days in 2026 has not changed position, however much the absolute barrels have grown.Interpretation

US propane throughput has grown enormously. Exports in the week to 11 September hit 2,213 thousand barrels a day — a record, and 52 per cent above the five-year average of 1,455. Production was also at a record. Domestic product supplied jumped from 657 to 1,038 as heating load began entering the data, taking total disappearance to 3,251 thousand barrels a day.

Against that, 109.09 million barrels covers 37.0 days of demand on a four-week average basis. The five-year average for the same week is 37.6 days. A record inventory in barrels is an ordinary position in days — 1.7 per cent below normal.Verified

Why a four-week average, and what the single week says Weekly product supplied is volatile: readings for this week of the year have ranged from 607 to 1,478 thousand barrels a day across the last seven years. Taken on the single week to 11 September alone, cover falls to 33.6 days against a 37.2 average — a ten per cent deficit. That reading is real but it rests on one unusually strong week of exports and demand. We use the four-week measure because it is the more defensible of the two, and we report both so the reader can judge. The direction is the same either way; only the magnitude changes. Estimate
Fig 1A Record in Barrels, an Ordinary Position in Days
US propane and propylene position, week to 11 September, against the 2021–25 average
+30% +15% 0 +21% Absolute stocks 109.09 vs 90.04 MMbbl −1.7% Days of cover 37.0 vs 37.6 days four-week average basis ÷ throughput exports +52% vs 5-yr
Source: EIA Weekly Petroleum Status Report, week ending 11 September 2026. Days of cover calculated as ending stocks divided by total disappearance (product supplied plus exports) on a four-week average basis. On the single week the reading is 33.6 against a 37.2 average; both are given in the text. Five-year averages are 2021–2025 for the same reporting week. Verified

Everyone will quote the record. Almost nobody will divide it by throughput.


Ninety-Two Per Cent of It Is in the Wrong Place

National inventory conceals where the barrels sit, and for propane that is the whole question. The molecule that matters in January is in Conway, not Mont Belvieu.Interpretation

On the most recent published district breakdown, for 4 September, of the 22.02 million barrels by which US stocks exceeded their five-year average, 20.25 million sat on the Gulf Coast. PADD 3 holds 70.20 million barrels against a five-year average of 49.95 — 40.5 per cent above normal, and an outright record for the week. It now accounts for 63.5 per cent of all US propane.

The Midwest, where heating demand actually lives, holds 26.56 million barrels against a five-year average of 25.23. That is 5.3 per cent above normal and sits inside its own 2018-to-2025 range of 22.80 to 26.97 million barrels. The East Coast is 6.4 per cent above on a small base.Verified

Fig 2Where the Surplus Actually Sits
Stocks above or below the five-year average by region, million barrels · 4 September, latest district data
0 +10 +20 MMbbl PADD 3 Gulf Coast +20.25 — 92% of the surplus +40.5%, record for the week PADD 2 Midwest +1.33 +5.3%, inside its normal range PADD 1 East Coast +0.52 The glut is export-facing length, not a winter cushion.
Source: EIA weekly ending stocks by PADD, week ending 4 September 2026, against 2021–2025 averages for the same week. Verified

That distinction decides what the surplus means. Gulf Coast length is export-facing: it clears through terminals, freight and the Asian arbitrage, and it is the same length that has been struggling to reach Asia through a rationed Panama Canal. Midwest length is heating cover. They are different commodities in everything but chemistry.Interpretation


There Is No Cushion Where It Would Matter

The reason the regional split matters is that a normal Midwest winter consumes most of what the Midwest holds.Verified

Across the last seven heating seasons, PADD 2 has drawn between 14.43 and 20.19 million barrels from peak to trough. The region currently holds 26.56 million. A draw at the top of that historical range would leave roughly six million barrels — a level the Midwest has reached before, and not comfortably.

Fig 3Seven Winters of Midwest Draws, Against What Is on the Ground
PADD 2 peak-to-trough propane draw, million barrels
28 14 0 26.56 MMbbl on the ground today 16.8319/20 14.4320/21 16.2021/22 15.3222/23 16.5323/24 20.1924/25 16.3525/26 A 24/25-style winter would leave the Midwest with about six million barrels.
Source: EIA weekly PADD 2 propane and propylene ending stocks. Draw measured as the peak reading between 1 October and the trough by 31 March of the following year. Verified

So the national record buys the Midwest almost nothing. The 20 million surplus barrels are on the Gulf Coast, and moving propane north against the grain of the pipeline system in January is exactly the manoeuvre that has failed in previous cold snaps.Interpretation

Unless the winter does not arrive

That arithmetic assumes a normal heating season, and the forecast does not. NOAA's Climate Prediction Center has carried an El Niño Advisory since June and, as of 8 September, puts the odds of a very strong event through winter 2026–27 above 90 per cent — up from 63 per cent in June and 81 per cent in July. It assigns a 69 per cent chance that the October-to-December reading exceeds +2.5°C, which would surpass every El Niño in the record back to 1950.Verified

El Niño winters are typically milder across the northern tier of the United States and Canada, and the propane data reflects it. The two strong El Niño winters in the weekly series drew 14.91 and 16.47 million barrels from PADD 2, averaging 15.69 against 17.18 for La Niña winters — roughly nine per cent lighter.

Two observations are not a forecast, and the ranges overlap — the 2022–23 La Niña drew just 14.66 million barrels. But the direction is clear enough: a normal draw leaves the Midwest around nine or ten million barrels, the worst in the series would leave 6.4, and a mild El Niño winter leaves it comfortable.Estimate

The barrels say glut. The days of cover say ordinary. The weather forecast is the tiebreaker, and it currently favours the barrels.


Inherited, Not Accumulated

There is a natural assumption that a record inventory implies an unusually strong build. The opposite happened.Verified

Between the end of March and mid-September, US stocks rose at roughly 1.4 million barrels a week. That is the slowest injection season in eight years — against 2.33 in 2025, 2.11 in 2024 and a 1.9-to-2.0 norm across the period. And the season has now turned: the week to 11 September brought the first draw, of 1.4 million barrels.

The record exists because of where the year started. Stocks stood at 77.60 million barrels at the end of March 2026, against a range of 34 to 59 million in every prior year of the series. The market entered the injection season roughly 25 million barrels long and has simply carried that forward at a below-average pace.

Fig 4The Slowest Build in Eight Years
Starting stocks at end-March and weekly injection rate to early September, million barrels
END-MARCH STARTING STOCKS 202233.8 202357.0 202457.6 202544.7 202677.6 WEEKLY BUILD RATE 20222.03 20231.96 20242.11 20252.33 20261.49 Started 25 MMbbl long. Built slower than any year since 2019. An overhang carried forward will not be cleared by one strong export month. Build measured from the last reporting week of March to the first of September.
Source: EIA weekly US propane and propylene ending stocks. Build rate is the change between the last reporting week of March and the first of September, divided by weeks elapsed. Verified

The Market Is Not Pricing a Glut

If the record surplus were as large as the headline implies, the price would say so. It does not.Verified

In the week the record was set, Mont Belvieu propane rose 13.5 per cent to $0.855 a gallon — its level a year earlier was $0.774, and two years earlier $0.676. Propane rallied more than a tenth in seven days, into an all-time record inventory, while that inventory drew.

Markets do not do that when they are long. They do it when the physical position is ordinary rather than loose — which is exactly what 37 days of cover describes. The price is agreeing with the denominator, not the numerator.Interpretation

One further series settles the question that usually dogs this data. The EIA reports propane and propylene together, which invites the objection that the build is petrochemical rather than fuel. It also publishes propane, fractionated and ready for use: 74.05 million barrels, or 67.9 per cent of the headline. The equivalent share was 62.4 per cent a year ago and 70.2 per cent in 2024. The fuel-ready proportion is entirely normal, so the record is not a propylene artefact.Verified

The honest objection

Three arguments cut against this reading. Days of cover is flattered by a low domestic demand reading in early September, when heating load is at its seasonal minimum — the ratio will look tighter once winter demand enters the denominator, and the comparison is like-for-like across years but the level should not be read as a winter measure. The regional split is a 4 September reading, the most recent published by district, so the geography is a week behind the national figures. And the El Niño draw comparison rests on two observations, which is not enough to forecast from — it is offered as a direction of risk, not a number to trade on. And a record is still a record: if exports stall for any reason — freight, canal access, an Asian demand shock — the cushion that looks reasonable in days becomes very large in barrels very quickly.


What Follows

1 · Stop quoting the national number. It is true, it is a record, and it describes a Gulf Coast export position rather than a US supply position. The two behave differently and respond to different variables. Anyone sizing winter risk from the national figure is reading the wrong series.

2 · Watch days of cover, not barrels, and watch PADD 2. Cover is already below its five-year average and heating demand has only just begun entering the denominator. The Midwest holds roughly a normal book against a seven-winter draw range of 14.4 to 20.2 million barrels. Both series are published free every Wednesday.

3 · Treat the Gulf length as a freight problem. Twenty million surplus barrels on the Gulf Coast clear through export terminals and the Asian arbitrage. Their fate depends on canal access, VLGC availability and the delivered arb far more than on US weather.

4 · Use days of cover, not barrels, whenever the market is growing underneath the tank. Exports have risen more than 20 per cent in a year. Any inventory metric that ignores that will keep generating records that mean less than they appear to.Interpretation

Our call, stated so it can be checked We expect the gap between the barrel reading and the cover reading to widen as heating demand enters the denominator through October, and we expect four-week days of cover to remain within roughly five per cent of its five-year average rather than moving to any genuine surplus. We expect PADD 2 to enter winter within roughly 10 per cent of its five-year average rather than at any meaningful surplus. And we expect Mont Belvieu to hold above its 2024 and 2025 levels for equivalent weeks, because the market is pricing cover rather than barrels — with the clear caveat that a very strong El Niño, which NOAA now rates at better than 90 per cent, is the single most likely way that view is wrong. And we expect the Gulf Coast share of US stocks to fall only when export routing eases, not when US demand rises. We will revisit this against the same public series in twelve months. Estimate

What would change our mind

  • A confirmed very strong El Niño with a mild December. Heating demand is the largest single input to the denominator. If it fails to arrive, cover rises without a barrel being added and the surplus becomes real. This is the most probable route to our being wrong.
  • Four-week cover rising materially above the five-year average. If exports fall back toward 1,455 kb/d while stocks hold, cover jumps and the headline surplus reading becomes the correct one.
  • The single-week and four-week measures diverging further. We use the four-week figure because weekly disappearance is noisy. If the two persistently disagree, the metric itself needs rethinking before either is quoted.
  • PADD 2 building sharply into October. If Midwest stocks push above their 2018–2025 range, a genuine heating cushion exists and the regional argument weakens.
  • Mont Belvieu breaking below 2025 levels. Price is the cleanest cross-check available. If it falls back through $0.774 while stocks stay at a record, the market disagrees with the cover reading and so should we.
  • An export stall. Exports at 20 per cent above the five-year average are doing the work in the denominator. If they drop back toward the average, days of cover rises sharply without a single extra barrel being produced.

Record Barrels, Ordinary Days

US propane stocks have never been higher. That will be written many times this autumn, and it will be accurate every time.

It will also describe a market that has outgrown its own inventory. Exports at a record and 52 per cent above their five-year average, production at a record, and a surplus that vanishes entirely the moment it is divided by the throughput it serves — while sitting, almost entirely, in the one region where winter never bites.

In the week the record was set, stocks drew and the price rose more than a tenth. That is what an ordinary market looks like wearing a remarkable number.

What decides which reading is right is not the inventory at all. It is whether a winter arrives to consume it — and NOAA currently gives better than nine-to-one odds that this one is very strong El Niño. The barrels and the days disagree. The weather will settle it.

The Midwest goes into heating season with an ordinary book and a seven-year history of drawing most of it down. The Gulf Coast goes into it with twenty million surplus barrels and a routing problem. Those are two different stories, and the national number tells neither.

IS

Insight Stream — LNG/LPG & Waterborne Trade Analysis

LNG contract economics, LPG benchmarks and Pacific trade flows, VLGC freight, China PDH and NGL export infrastructure. Independent market intelligence, built only on sources you can check.

Read More →