The explosive growth of artificial intelligence is setting U.S. data centers on a path to become one of the world's largest consumers of natural gas, with a new forecast predicting they will use more of the fuel than Germany and Japan combined by 2035. According to a report from BloombergNEF, covered by TechCrunch, these facilities are projected to consume about 18 billion cubic feet of natural gas per day within the next decade.
This staggering figure represents a near-doubling of BloombergNEF's own forecast from just nine months ago, underscoring how rapidly projections are escalating alongside the AI boom. The data center sector is now expected to be the second-strongest driver of natural gas demand growth in the United States, trailing only the expansion of liquefied natural gas (LNG) exports.
The forecast delineates two major sources of this demand surge. The largest share, approximately 15 billion cubic feet per day, is expected to come from grid-connected data centers. This consumption represents a 117% increase from a previous forecast of 6.9 billion cubic feet and is five times the growth projected from all other grid-connected sectors combined. The remaining 2.9 to 3.4 billion cubic feet per day will come from a newer trend: data centers generating their own power on-site with natural gas plants.
This move toward on-site generation has been led by tech giants including Meta, Microsoft, Google, and Amazon, all of which have announced plans for new natural gas power plants designed to bypass the traditional electrical grid. The consumption from these proprietary plants alone by 2035 is projected to match the total amount of natural gas consumed by all U.S. data centers today, including both on-site and grid-sourced power.
The collective demand would position U.S. data centers as the fifth-largest natural gas consumer in the world by 2035, according to Tom's Hardware. Only the entire nations of China, Russia, Iran, and the United States itself would consume more. This level of consumption aligns with broader forecasts suggesting data centers will account for 20% of total U.S. power demand by 2035, equivalent to roughly 194 gigawatts.
The primary driver of this unprecedented energy draw is the insatiable demand for compute power required to train and run increasingly large and complex AI models. As the AI race intensifies, companies are building ever-larger data center campuses, each filled with power-hungry servers and requiring immense cooling infrastructure. The geographical concentration of this new AI load in specific power-constrained regions is expected to create localized strains on energy infrastructure and affect natural gas pricing dynamics beyond simple increases at the Henry Hub benchmark.
The environmental implications are significant. The incremental emissions from this surge in natural gas consumption are estimated at 1 million metric tons of greenhouse gases per day. This would represent an increase of roughly 12% over current total U.S. emissions, presenting a major challenge to corporate sustainability goals and national climate targets.
Economically, the analysis suggests the AI-driven demand, coupled with expanding LNG exports, could push natural gas prices "materially higher" and raise utility costs for consumers. The impact is not just on price levels but on the differentials between regions, with producers in areas like Appalachia with access to key transportation corridors likely to benefit.
This forecast highlights a critical tension at the heart of the modern tech industry. The pursuit of more powerful AI, which promises advancements in numerous fields, is directly at odds with efforts to decarbonize the economy and transition to renewable energy. The trend toward on-site gas generation, while ensuring reliability and bypassing grid constraints, represents a doubling down on fossil fuel dependency for the sector. As projections continue to be revised upward, the energy appetite of the AI revolution is emerging as a dominant theme in energy policy, corporate strategy, and climate negotiations for the coming decade.
