US Data Centers Could Soon Burn More Gas Than Germany and Japan Combined

U.S. data centers could consume about 18 billion cubic feet per day of natural gas by 2035, more than Germany and Japan combined.

That figure comes from a BloombergNEF report published on Sept. 15, 2026, as described by TechCrunch. It covers both gas burned in power plants serving grid-connected facilities and gas burned onsite at data center campuses.

Data centers would be the second-strongest driver of U.S. natural gas demand growth over the next decade, after LNG exports. The latest forecast is nearly double the volume BloombergNEF predicted nine months earlier, a revision that reflects faster build schedules and larger load interconnection queues.

The 18 Bcf/d total splits into two distinct procurement models. Onsite-powered projects, where turbines or fuel cells sit inside the fence, are expected to consume 2.9 billion to 3.4 billion cubic feet per day by 2035. Grid-connected data centers are predicted to drive an additional 15 billion cubic feet per day of gas consumption by the power sector by the middle of the next decade.

That grid-connected increment is striking in isolation. Projected gas demand growth through 2035 from grid-connected data centers is five times more than from all other grid-connected sectors combined.

The electricity numbers behind the gas forecast are equally sharp. U.S. data centers will account for about 20% of the nation's electricity consumption in 2035, up from 5.9% today, according to Bloomberg reporting in July, cited via Bloomberg. The U.S. Energy Information Administration forecasts the strongest four-year growth in U.S. electricity demand since 2000, fueled by data centers, as noted by the EIA. Through 2050, the agency projects consumption will continue growing at 0.9% to 1.6% annually, with server energy use a major factor.

EIA analysts also expect the load to reshape the built environment. By 2050, as much as 7% of all U.S. commercial floorspace requires additional energy to meet data center demand across most building types.

Global figures put the U.S. concentration in perspective. Global data center energy consumption could approach 1,050 TWh by 2026, according to an estimate cited by Brookings. Data centres' share of global electricity demand is projected to remain less than 2% in 2035, according to the IEA. Natural gas and coal together are expected to meet over 40% of the additional electricity demand from data centres until 2030, with natural gas expanding by 175 TWh to meet growing demand.

Onsite gas is already part of that global picture. Around 15-27 GW of onsite natural gas may power data centres by 2030, mostly in the United States. The U.S. trajectory to 2.9 to 3.4 Bcf/d for onsite projects by 2035 extends that trend and points to sustained orders for medium-frame turbines, reciprocating engines and associated interconnection equipment.

Supply-side capital is moving in parallel. Japanese trading house Mitsui is looking to invest in LNG projects across the Middle East, the U.S. and Australia to meet rising power demand from data centers, according to Reuters. On Aug. 28, 2026, Cheniere Energy completed its Corpus Christi Liquefaction Stage 3 Project in Texas and took custody and control of the seventh and last LNG train in that project.

The emissions arithmetic is direct. Additional natural gas demand from data centers will generate 1 million metric tons more greenhouse gas pollution daily.

The broader context here is speed of revision rather than direction. Load forecasters have been raising data center estimates for three years, but a near doubling in nine months suggests interconnection filings and behind-the-fence generation plans are outrunning utility integrated resource plans. For operators, gas offers firm dispatch, known permitting paths and delivery timelines that fit 2027 to 2029 energization targets.

In this author's view, the tension worth watching is not gas versus renewables in the abstract. It is firm capacity versus queue time. Grid-connected growth at five times all other sectors combined will test transformer supply, turbine backlog and pipeline lateral capacity at the same time. Onsite systems ease the interconnection bottleneck but shift emissions accounting and air-permit scrutiny to the campus.

Worth flagging for enterprise architects: power procurement is becoming a primary site-selection constraint, alongside fiber, water and tax treatment. Contracts that lock in electrons and molecules for 10 to 15 years will shape where inference capacity lands, and at what latency and cost. I covered similar constraint-driven siting during the cloud buildout, when tax and fiber maps decided regions, and power maps are now exerting the same pull.

The long arc still points to more useful compute per unit of energy, and to grids that eventually absorb this load with cleaner mixes. The near term, however, will run substantially on gas.