EIA Injection Slowdown Shifts ERCOT Gas Signal Toward Neutral
A 30 Bcf weekly injection, running below the recent pace, has tightened the U.S. storage surplus to roughly 160 Bcf above the five-year average, easing bearish pressure on Texas power costs.
NYMEX Henry Hub front-month touched $2.937 intraday on Friday (2026-09-04) before recovering, after the EIA's weekly storage report showed an injection of just 30 billion cubic feet — smaller than the pace seen in recent reports, though broadly in line with near-term expectations, FXEmpire reported. The contract ended Friday's (2026-09-04) session at $2.98/MMBtu, just short of the $3 resistance that has capped the market for weeks.5
The significance for ERCOT is direct. Gas-fired generation sets the marginal clearing price across a large share of Texas dispatch hours. A storage picture that is becoming less bearish translates into less downward pressure on real-time ERCOT power costs, even before any change in the physical supply of gas.5
The storage surplus tells the story. U.S. working gas stocks sit roughly 160 billion cubic feet above the five-year seasonal average, down approximately 50 billion cubic feet compared with year-ago levels, per EIA data reported by FXEmpire on Friday (2026-09-04). The overhang is narrowing, not widening. That has removed the clearest argument for a sustained push lower in gas.5
Rising LNG export demand has contributed to the compression. FXEmpire cited increased LNG offtake as one factor behind the smaller injection, without providing a specific throughput figure.5 If LNG export volumes hold at elevated levels through September, the surplus will continue to erode faster than seasonal norms imply.
In May 2026, NYMEX Henry Hub front-month was trading near $2.58 on Thursday (2026-05-14), under pressure from storage injections running above analyst forecasts and the five-year average. Total U.S. gas production held at 101.5 billion cubic feet per day during that period, while domestic consumption fell 4.3% week-over-week and power-sector demand declined 5.7%.1 Production was stable, demand was weak, and injections were large. Three months later, only the first of those three conditions still holds.5
The ERCOT demand backdrop adds pressure from the other side. The grid hit an unofficial all-time peak on July 22 (2026-07-22). Ascend Analytics projected in August 2026 that peak demand in the territory could reach 120 gigawatts by 2030, more than 30% above that July record, but below ERCOT's own projections, with supply constraints expected to cap growth short of the grid operator's own forecasts.4 A grid pushing toward successive summer records changes the gas dispatch risk profile, independent of where storage sits in any given week.
Data center load is already reshaping ERCOT's demand profile. The grid operator approved two sets of rules on Tuesday (2026-06-02) governing how large electricity users connect, including a batch-review process for major new loads.2 Enel Green Power operates more than 4,500 megawatts of solar, battery storage, and wind within ERCOT, a renewable fleet that displaces gas-fired output during high-generation periods, with no equivalent offset during evening ramp periods or weather-driven peaks where the gas price floor still governs.3
The current market balance is genuinely neutral. The storage surplus is still present but shrinking. LNG demand is pulling supply away from storage. ERCOT's summer peak set an unofficial record. Gas prices have sat within one percent of $3 for most of the summer, with the $3 mark capping the market since late August. None of these signals points clearly in one direction.5
What resolves the standoff is the September injection cadence. If the next two or three EIA reports confirm a trend toward smaller injections, the surplus could shed another 30 to 50 billion cubic feet before October, tightening the winter cushion and shifting ERCOT's forward curve higher. A return to larger injections would extend the soft-to-neutral environment that has prevailed since spring. The next weekly EIA report is the first data point of consequence.5