Two of the biggest stories in American energy this year draw on the same resource. Between March and June, three large LNG export projects reached final investment decision: Phase 2 of Venture Global's CP2 LNG on March 13, Caturus's 9.5 million metric ton per year Commonwealth LNG on May 15, and Delfin Midstream's first floating liquefaction vessel on June 3. At the same time, electricity consumption is reaching record levels, driven largely by data centers, and natural gas generates about 40% of the country's power. Both export terminals and data centers ultimately run on Permian and Haynesville gas. The question is whether there is enough to go around, and at what price.
On the evidence of the Energy Information Administration's September Short-Term Energy Outlook, the answer for the next two years is yes. The harder question concerns the end of the decade, when the new export capacity starts up just as data center load approaches its forecast peak.
The demand side
EIA forecasts US electricity sales of 4,135 billion kilowatthours in 2026, an increase of almost 2% over 2025, with a further increase of nearly 2% in 2027. It attributes the growth to data center development and increased manufacturing in the commercial and industrial sectors. Natural gas generation rises 2% in 2026 and 1% in 2027, while coal generation falls 8% and 6% and solar grows 21% and 18%. EIA expects PJM to account for nearly 45% of the growth in total US generation, with gas driving the increase in 2026.
On the export side, LNG exports keep rising through 2027 as recently completed terminals ramp up, including Golden Pass, which shipped its first cargo in April. EIA also notes that commercial sector sales, where most data center consumption is recorded, make up a large part of the growth: the agency expects electricity consumption to total 4,211 billion kilowatthours in 2027.
The supply side
Production is the reason the squeeze has not arrived. EIA forecasts marketed natural gas production rising by 4.5 Bcf/d in 2026 and 4.6 Bcf/d in 2027. The Permian and Haynesville together account for more than 70% of that growth. Permian output grows 1.7 Bcf/d in 2026 and 2.2 Bcf/d in 2027, mostly as associated gas from oil wells; EIA notes that the Permian gas-to-oil ratio averaged nearly 4,200 cubic feet per barrel in 2025, 15% higher than the year before. Haynesville production rises 1.4 Bcf/d and 1.3 Bcf/d, supported by stable Henry Hub prices and proximity to Gulf Coast export terminals.
Inventories reflect that growth. EIA expects working gas in storage to total 3,969 billion cubic feet on October 31, 5% above the previous five-year average. Henry Hub is forecast to average about $3.43 per million British thermal units in 2026 and $3.28 in 2027. Those are not the prices of a market that is short of gas.
Why the timing matters
The three LNG projects that reached FID this year will not export a cargo until around 2030. Commonwealth's operations are scheduled to begin by 2030, and Delfin expects its first vessel to start production in 2030. CP2, with a peak capacity of 29 million metric tons a year across both phases, has contracted nearly all of its nameplate output on long-term terms with buyers mainly in Europe and Asia. These are commitments made in 2026 for gas that will be needed in the 2030s.
Data center demand follows a different clock. PJM's 2026 long-term forecast has summer peak rising by more than 65,000 MW over the next ten years, and ERCOT is tracking hundreds of gigawatts of large load requests. Much of that load, if it materializes, will arrive between 2027 and 2032, overlapping with the start-up of the new export trains. If both arrive at once, the Permian and Haynesville will need to grow faster than they have this year, or prices will rise to ration demand.
How much gas could data centers use
The International Energy Agency's Energy and AI report offers one benchmark. It projects that natural gas generation expands by 175 TWh to meet growing data center demand worldwide by 2035, particularly in the United States. As a rough conversion of our own, assuming a gas plant heat rate of 7,000 Btu per kWh, 175 TWh of gas generation would require about 1.2 trillion cubic feet a year, or roughly 3.4 Bcf/d. Even if most of that is American, it is smaller than the roughly 4.5 Bcf/d of production growth EIA expects in a single year, 2026. The scale of data center gas demand is meaningful but not overwhelming at the national level.
The pressure is regional. Data centers cluster in PJM, Texas and the Southeast, and gas has to reach them through pipelines. Appalachian production, which serves PJM directly, grows only about 0.6 Bcf/d in 2026 and 0.3 Bcf/d in 2027 in EIA's forecast, far less than the Permian. EIA notes that inventories in the East began the injection season 11% below the five-year average after cold-weather withdrawals. A national surplus can coexist with regional tightness when pipelines are constrained, and a cold winter in the Mid-Atlantic is when data center load, heating load and gas supply constraints collide.
Who wins a bidding war
If the market tightens, LNG buyers and power generators will compete for the same gas at Henry Hub and regional hubs. LNG offtakers have long-term contracts, typically priced at a markup to Henry Hub, which commit them to lift cargoes regardless of US prices. Power generators buy gas to run plants when power prices justify it. Data centers, as end users, are largely insulated from fuel costs in the short run, but those costs flow through wholesale power prices and eventually through utility rates and contracts.
That points to the real tension. It is not that LNG exports will starve data centers of fuel. It is that rising gas demand from both, without matching growth in production and pipelines, will raise the price of gas-fired electricity for every customer, including households. The current forecasts suggest that point is still a few years away. The FIDs taken this year make it more likely that it will arrive around 2030.