On the surface, the US gas market enters winter in comfortable shape. The Energy Information Administration's storage report released on 1 October put Lower 48 working gas at 3,415 Bcf for the week ending 25 September, 79 Bcf above the five-year average. The November futures contract settled at $3.03 per MMBtu on 30 September. Europe and Asia, meanwhile, are paying roughly eight times as much for their gas. It is tempting to read the gap as a sign that America has gas to spare.
We read it differently. Henry Hub is low because export capacity caps how much of the global shortage can reach US prices, not because the domestic balance is loose. Underneath the national total, the storage picture is tightening, the official end-October forecast looks hard to reach, and the deficit is concentrated in the one place that matters most in a cold snap: salt caverns on the Gulf Coast.
The divergence, in numbers
The American Gas Association's weekly indicators, drawing on futures data, set out how far the markets have separated since the Middle East war began. Between the week ending 22 February and the week ending 27 September, the weekly average of Dutch TTF futures rose about 127 per cent, from $10.78 to $24.49 per MMBtu, and Asian JKM futures rose about 145 per cent, from $10.63 to $26.08. Henry Hub futures rose 7 per cent, from $2.89 to $3.11. The spot market tells the same story: the EIA's daily Henry Hub series, published through FRED, showed $3.13 on 28 September and $3.18 on 29 September.
Europe's position explains why the overseas bid is so persistent. Canada LNG Group, citing AGSI+ data, put EU underground storage at 70.6 per cent full on 25 September, 18.3 per cent below the five-year average. A continent that short of gas heading into winter will keep bidding for every flexible US cargo it can find.
The physical link between those markets is liquefaction, and it was running close to full. LNG feedgas averaged 18.0 Bcf per day in September according to Rystad Energy data cited by the AGA, up about 0.8 Bcf per day on August as plants returned from maintenance. Energy Edge Consulting's note on the 1 October report put feedgas at 18.2 Bcf per day, with Sabine Pass flows lower because four of six engines at Train 3 appeared to be offline for planned maintenance, and Cove Point maintenance expected to last into the second half of October. Pipeline exports to Mexico were averaging 7.6 Bcf per day for the week.
When every available liquefaction train is already running, a higher price in Rotterdam or Tokyo cannot pull more molecules out of Louisiana. The spread simply widens, and the rent accrues to whoever holds the export capacity and the cargo. That is why Henry Hub has barely moved while overseas prices more than doubled. New capacity is arriving, with Corpus Christi Stage 3 completed on 28 August and lifting the terminal's peak capacity to 3.9 Bcf per day according to the AGA, but it raises the ceiling in steps rather than removing it.
The storage surplus is shrinking
The national number hides the trend. The 64 Bcf injection in the latest week was in line with expectations, but it was the seventh consecutive week of builds below the five-year average, according to the AGA. The previous week's build was 53 Bcf. Stocks are now 138 Bcf below last year, and the surplus to the five-year average, 2.4 per cent, has been narrowing as the season closes.
The EIA's September Short-Term Energy Outlook forecasts 3,969 Bcf in storage on 31 October, 5 per cent above the five-year average and 1 per cent above October 2025. The arithmetic on that forecast is demanding. Getting from 3,415 Bcf on 25 September to 3,969 Bcf on 31 October requires 554 Bcf in 36 days, an average of about 15 Bcf per day or roughly 108 Bcf a week. The last two weekly builds were 53 and 64 Bcf. Unless injections nearly double for five straight weeks, end-October stocks will land below the STEO path. A shortfall of 100 Bcf or more against that forecast would still leave the country close to its five-year norm, but it would remove the cushion the forecast implies.
Supply explains part of the slower pace. Energy Edge put production at 109.3 Bcf per day on 30 September, with a force majeure on the TCO system between Braxton and Stonewall weighing on Appalachian flows by about 0.5 Bcf per day. The STEO expects marketed production to grow by 4.5 Bcf per day in 2026, with the Permian and Haynesville accounting for more than 70 per cent of growth. Over a full year that growth is real. Over the next five weeks, it is not arriving fast enough to restore the summer surplus.
Where the gas is, and where it is not
The regional detail in the EIA table, for stocks on 25 September, is the core of our concern.
- East: 840 Bcf, 4.6 per cent above the five-year average
- Midwest: 984 Bcf, 2.9 per cent above
- Mountain: 248 Bcf, 6.4 per cent above
- Pacific: 295 Bcf, 8.9 per cent above
- South Central: 1,048 Bcf, 2.4 per cent below
- South Central salt: 213 Bcf, 15.5 per cent below
- South Central non-salt: 834 Bcf, 1.5 per cent above
South Central is the only region below its five-year average, and within it, salt storage is 15.5 per cent below the average and 27.3 per cent below a year ago. Salt was the one sub-region to draw down in the latest week, by 4 Bcf, while every other region injected.
That matters out of proportion to its size. Salt caverns can cycle gas in and out quickly, and they sit next to the Gulf Coast's export terminals, petrochemical plants and gas-fired power fleet. They are the market's shock absorber during a cold snap or a production freeze. The STEO expected South Central to enter the withdrawal season 4 per cent above its five-year average. As of 25 September it is 2.4 per cent below, with salt far lower. The forecast and the data are moving in opposite directions in precisely the region that sets marginal prices on cold days.
What this means for buyers and generators
The national averages argue for a quiet winter, and the EIA's own price outlook agrees: the September STEO, as summarised by the AGA, expects Henry Hub to average $3.43 per MMBtu in 2026 and $3.28 in 2027. We do not dispute the annual averages. Our concern is the distribution around them.
A market with a liquefaction ceiling and thin Gulf Coast salt behaves asymmetrically. In mild weather, the ceiling holds prices down because domestic demand falls while exports are already maxed out. In a severe cold event, the ceiling stops helping. LNG plants will try to keep running because overseas netbacks are so far above Henry Hub, power and heating demand spikes, and the fast-cycling storage that normally meets that spike is below normal. Price spikes in that situation would be local and short, but they would land on gas-fired generators and industrial buyers along the Gulf Coast first.
For utilities and large consumers, that argues for paying more attention to winter-strip and daily-delivery hedges in South Central than the national storage number alone would suggest. For power market operators, it argues for treating fuel assurance in Texas and the Southeast as a live question this winter, not a settled one.
What to watch
Three releases will test this view in October. The weekly storage reports on 8, 15, 22 and 29 October will show whether builds accelerate towards the roughly 108 Bcf a week the STEO path needs, and whether salt turns back to injection. The October Short-Term Energy Outlook will show whether the EIA trims its end-October figure and its South Central forecast. And the return of Sabine Pass Train 3 and Cove Point from maintenance will lift feedgas, adding demand just as heating load begins to appear.
Three-dollar gas is genuine, and for most of the country this winter it will probably hold. But it is a price set by the size of the export pipe, not by a deep domestic cushion. The cushion is thinner than it looks, and thinnest exactly where a cold week would test it.