AI Data Centers: Natural Gas Demand Forecast Nearly Doubles in Nine Months

According to a BloombergNEF report relayed by TechCrunch on September 15, U.S. data centers could consume up to 18 billion cubic feet of natural gas per day (Bcf/d) by 2035 — an upward revision that is nearly double the forecast made just nine months earlier. That volume would exceed the combined current natural gas consumption of Germany and Japan.

The breakdown is telling: data centers with their own onsite power generation (Meta, Microsoft, Google and Amazon are all building onsite capacity, much of it gas-fired) would account for 2.9 to 3.4 Bcf/d, while grid-connected data centers would pull an additional ~15 Bcf/d of demand through the power sector — a demand increase five times larger than all other sectors of the U.S. economy combined through 2035.

The climate impact follows mechanically: applying the International Energy Agency's emissions factor (roughly 60 grams of CO2-equivalent per cubic foot, including extraction and distribution), this additional demand would generate approximately 1 million metric tons of greenhouse gas pollution per day — about 12% of total current U.S. emissions. The report also flags a price risk: the simultaneous race by multiple hyperscalers and LNG exporters for the same gas capacity could push prices higher.

Cardan-AI angle: this near-doubling of the forecast in nine months is not just an energy statistic — it is a warning signal for any industrial company exposed to natural gas as an input (electricity, process heat, petrochemical feedstock). Gas infrastructure investment decisions being made today (pipelines, turbines, LNG terminals) rest on a forecast that just proved extremely unstable barely a year after being set — a risk parameter that finance functions across energy and O&G should fold into their own cost scenarios without delay.