Gas in the backyard
Behind-the-meter turbines are the nearest fuel for the AI buildout. Appalachian producers are counting on a price floor. Last quarter, realized prices went down.
If the grid will not give you megawatts until the end of the decade, you buy a turbine or a genset and park it next to the hall. That machine almost always burns gas. About 101 gigawatts of behind-the-meter gas plant has been announced. The molecule that feeds it is the un-glamorous end of the same power bottleneck: not Caterpillar’s backlog, not a nuclear PPA, but an Appalachian wellhead.
Data-center gas demand is sized at about 6.1 billion cubic feet a day by 2030. Halls go from 4.4 percent of US power in 2023 to somewhere between 6.7 and 12 percent by 2028. More than thirty-five campuses are already in ten-year-plus take-or-pay supply deals. EQT’s chief executive is the one who said the quiet part on a call.
The next wave of natural gas demand growth is emerging in our backyard.
EQT, second-quarter 2026
A floor on the forwards, not yet in the print
The Energy Information Administration’s own balance has demand growing faster than supply — plus 3.2 billion cubic feet a day against plus 1.4. Storage slipped below the five-year average. Henry Hub was written up from $3.10 in 2025 toward $4.00 in 2026. A July Today-in-Energy note then put 2026 just under $3.50 and 2027 just under $4.60, on an LNG ramp. The September 9 Short-Term Energy Outlook is a different print: $3.43 in 2026 and $3.28 in 2027 — a decline into 2027. Keep both named. The $4.60 lift is not the current STEO. Neither number is a spike. Matthew Smith’s version is that this is a timing and flow problem, not a reserves problem. The squeeze, if it arrives, shows up as a price floor, not as a field running dry.
Appalachia’s pipes are getting less tight. Mountain Valley Pipeline eases downstream bottlenecks in 2027–28. M-2 basis futures for 2029 and 2030 have tightened more than 20 cents. Tetco M-2 forwards sit at $2.940 for 2026 and $3.017 for 2027. Winter strips on TETCO M2 and Dominion South were at all-time highs as of July. EQT’s CFO says basin basis should “structurally tighten through the end of the decade.” That is the confirmed half of the story: the floor, if Henry Hub holds it, can reach the wellhead instead of dying in a trapped-basis discount.
The other half has not printed. EQT realized $2.65 per Mcfe in the second quarter, with basis a minus 67 cents that was “favorable to prior expectations” and still a decline. Adjusted earnings missed. Net income fell on lower derivative gains and lower realized prices. Demand deals are stacking. Spot prices went down. Argus has said the price impact may not hit until 2027. AEGIS has a colder read: some of the record basis is mechanical. When Henry Hub softens, less gas is pushed out of the basin, so local basis rises — a symptom of a weak national price, not proof of an AI bid.
Who still owns the upside
Among the four large Appalachian names, hedge books decide who would actually keep a floor. EQT is about 25 percent hedged for 2026, with 2027 calls struck around $4.51 — it sold the spike, not the lift. Net debt is down to $5.5 billion from $7.7 billion. The marquee offtake is a ten-year CPV Power deal at 325,000 dekatherms a day, priced to PJM power, not to a gas index. That is power-price beta with a gas costume. A five-year LNG SPA, 0.5 million tonnes from 2028, is the cleaner molecule contract.
Expand Energy is the scale name — about 7.5 billion cubic feet a day, Haynesville plus Marcellus, a “Hedge to Wedge” book that is supposed to keep upside. It markets itself as powering US data centers. Its own second quarter realized $2.90 per Mcf with hedges, and $2.15 in Northeast Appalachia against $2.62 in the Haynesville. The release had no 2027 hedge ladder and no data-center-versus-LNG split. The CEO talked about a $1.25 billion marketing acquisition. Antero is 60 percent-plus hedged for 2026; it has already sold most of a bare-gas floor, and lives instead on NGLs and firm transport to the Gulf. Range sits in the middle, with a roughly thirty-year inventory and a hedge percent the filings reviewed did not disclose.
2026 gas hedge, two ends of the book
Range’s 2026 hedge percent is not on its entity page. Expand’s program is described as retaining upside; a single 2026 percentage was not filed there.