U.S. natural gas futures settled down 1.5% at $2.871 in a late July 2026 session, extending a pullback from the $3 level tested earlier in the month and capping a bruising July for the prompt contract. The settlement on July 23 came the same day the EIA released its Weekly Natural Gas Storage Report, which showed working gas in storage at 3,056 Bcf as of Friday, July 17, 2026, with a net injection of 32 Bcf for that week (WSJ; EIA).

The pullback was not a single-session story. Futures shed more than 10% in July as of around July 10, with the slide accelerating after a Texas LNG terminal closed for maintenance, denting export flows (WSJ Live Coverage, July 10). On July 9, the contract posted its biggest single-day drop in more than three months (WSJ). By July 16, Reuters reported futures had slid about 2% to a two-month low, citing rising domestic output and lower LNG export flows as the twin drivers (Reuters, July 16).

The supply picture is straightforward: working gas storage at 3,056 Bcf sits comfortably above seasonal norms, and a 32 Bcf net injection during a week when the market was already leaning bearish reinforces the inventory overhang. When storage is ample, traders need either extreme weather demand or a supply disruption to bid prices higher. Neither materialized with sufficient force. The Texas LNG terminal maintenance removed a key export outlet, meaning more gas stayed trapped in the domestic market precisely as production was ramping. That is a classic bearish combination — rising supply meeting constrained offtake.

Weather, the other perennial swing factor for summer gas, failed to provide a sustained bid. The WSJ's July 23 story flagged the weather outlook as the trigger for intraday gains evaporating into a fractional close lower. Summer cooling demand can spike prompt prices when heat domes lock in over major demand centers, but forecasts apparently did not hold enough risk premium to keep the contract above the $3 mark it had tested earlier in the month.

For market participants, the $3 level is the near-term pivot. The contract probed above it, found no follow-through, and settled back to $2.871. That round number carries weight both technically — as a psychological magnet for algorithmic flows and stop orders — and fundamentally, since it roughly demarcates the zone where associated-gas economics and rig-count decisions begin to influence supply response.

Looking at the structural backdrop, the July selloff compresses what had been a more balanced narrative. Lower LNG export flows from the Texas terminal outage directly reduce feedgas demand, and with U.S. LNG export capacity now a marginal price-setting force on the Henry Hub curve, any maintenance event at a major terminal has outsized impact. Rising output, cited by Reuters, compounds the effect — Appalachian and associated production gains have outpaced demand growth, and storage injections are absorbing the surplus. The 32 Bcf injection for the week ended July 17 landed within the range the market had been positioned for, but coming alongside already-weak price action, it offered no catalyst for a reversal.

The broader context here is a market searching for equilibrium between growing productive capacity and intermittent export demand. The prompt contract's inability to hold $3 amid ample storage, rising output, and a temporary export disruption tells you the balance of risks leans bearish in the near term. The offsetting factor — that LNG terminal maintenance is by definition temporary — means part of the downward pressure is transient. When the Texas facility returns, feedgas demand should resume, tightening the domestic balance. Whether that coincides with continued production growth or a seasonal demand shift will determine whether the July pullback holds or reverses into August.

For now, the market has voted with its feet. Storage is sufficient, output is rising, export flows are constrained, and weather is not cooperating. That configuration settled the contract at $2.871 on July 23, and the path of least resistance remains lower unless one of those four variables flips.