The air conditioners are about to lose the argument
By August 20 the summer tape is tired. NGI's desk notes from the 19th already had traders looking past the heat toward a shoulder season that could leave balances even looser. Production is hovering near record. LNG demand is still below what you would need to absorb the extra supply. That is not a forecast for October. That is the setup for the next four weeks.
You can feel why people wanted a different story. Texas was hot. September futures tagged $2.80 on a hotter overnight model. Power burn in the south did the work. Then the national storage path refused to cooperate. The surplus versus the five-year average widened through the stretch that was supposed to shrink it. EIA's end-October call is still 3,985 Bcf. Sellers do not need that print to arrive. They are already trading it.
What shoulder season actually does
In a tight year, August heat draws the surplus and September is a pause before heating. In a long year, August heat is a speed bump and September is when the injections accelerate because the only bid left is storage. We are in the second year.
AGA's mid-August weekly already showed the split. Texas and the Southeast power burn were up. The West, Northeast, and Rockies were down hard. National electric demand fell. Once school starts and nights cool off, even the southern bid gets quieter. NOAA's late-August outlook still leans warm across a lot of the Lower 48. Warm is not the same as a heat dome that clears 200 Bcf.
Freeport is the other shoulder-season variable. Maintenance has kept about 2 Bcf per day off the water since July 10. The official line is late August. Train 3 tripped on the 19th after a power-feed interruption. If that plant is late into September, shoulder season inherits the same extra gas summer inherited. If it comes back on time, you get a feedgas bid into a market that is otherwise going to sleep.
The curve is not as asleep as the prompt
Winter contracts still sit at a premium to September. That is normal seasonality plus the chance a cold 2026-27 actually draws the pile. XInvest and others have been writing Henry Hub near $2.70 against that fat storage call, with the back end still offering a bid. The 12-month strip around $3.14 that AGA cited on August 20 is the market saying "not now, maybe later."
Later only works if something changes. A historic El Niño, which NOAA is now putting real odds on, is a change in the wrong direction for bulls. A Freeport return and a European restock are changes in the right one. Shoulder season is when you find out which story the injections believe.
Alberta feels this earlier
Western Canada does not have a Texas air-conditioning bid. AECO already lives in shoulder season most of the summer. C$1.46 per gigajoule in mid-August is what a basin looks like when the only new door is Kitimat and Kitimat is not 6 Bcf a day. If U.S. injections pick up in September, continental gas stays heavy and the southbound pipe stays optional. Canadian producers need the Pacific cargo, not a hope that Houston gets another week of 100 degrees.
Write the next month as a test, not as a crisis. The surplus survived the heat. Now it has to survive the quiet.