US Data Centers Could Consume More Natural Gas Than Germany and Japan Combined by 2035

US data centers are projected to consume about 18 billion cubic feet of natural gas per day by 2035, placing their demand above that of many national markets. The estimate comes from a new BloombergNEF outlook and reflects the rapid expansion of artificial intelligence computing in the United States.

The forecast represents a sharp increase from BloombergNEF’s previous estimate, which was issued roughly 9 months earlier. It also accounts for projects that have been announced but are not expected to reach completion. TechCrunch reported that data centers could consume more gas than Germany and Japan combined, while Bloomberg and Yahoo Finance described the facilities as a larger gas market than most countries.

Data centers are expected to become the second-largest source of growth in US natural-gas demand over the next decade, behind liquefied natural gas exports. That ranking places AI infrastructure alongside one of the energy industry’s biggest established growth engines.

The shift reflects the difficulty of supplying large computing campuses with enough electricity through existing grids. Some operators and power developers are turning to on-site gas generation, allowing facilities to add capacity without waiting for transmission upgrades. Industry forecasts also point to substantial demand growth tied directly to data-center construction, with East Daley Analytics estimating an additional 6 billion cubic feet per day.

The projected gas consumption would require new investment across production, pipelines, storage and power generation. Data centers need continuous electricity for servers and cooling, which makes fuel availability and delivery capacity central to project planning.

A buildout of this scale would also affect regional gas markets. Concentrated clusters of data centers can increase pipeline congestion and raise competition for power during periods of extreme weather or peak demand. Industry publications including OilPrice, Woodway Energy and East Daley have linked the data-center boom to a broader expansion of gas infrastructure, though their available excerpts provide limited detail on individual projects or regional forecasts.

Natural gas can provide dependable power for data centers, but higher consumption would increase scrutiny of emissions, air quality and methane leakage. On-site generation can reduce reliance on constrained grids while tying AI growth more directly to fossil-fuel infrastructure.

The energy choice also creates a policy conflict. The United States is seeking to expand domestic AI capacity while managing power-sector emissions and supporting cleaner electricity. A gas-heavy buildout could supply computing demand quickly, but it would shape infrastructure decisions for decades and complicate efforts to decarbonize electricity. A policy paper titled Powering AI With American Energy: Natural Gas frames gas as a tool for meeting that demand, while the BloombergNEF forecast quantifies the scale of the resulting market.

The forecast raises the energy cost of expanding AI services, even when the direct fuel bill is paid by cloud providers or data-center operators. Access to power, gas pipelines and generation capacity will increasingly influence where computing campuses are built and how quickly they can come online.

For technology companies, the issue reaches beyond electricity prices. Long-term power contracts, permitting, emissions rules and local opposition can affect training and inference capacity. Gas suppliers and infrastructure developers gain a large new customer base, while utilities face pressure to connect facilities without weakening reliability for households and existing businesses. The limited reporting from Epmag and Threads mirrors the headline-level nature of much of the discussion, leaving BloombergNEF’s quantified forecast as the central benchmark.

The central question for the next decade is how much of the announced data-center pipeline reaches operation and which fuels ultimately serve it. BloombergNEF’s estimate already discounts projects that are unlikely to be completed, but the gap between its new and previous forecasts shows how quickly assumptions about AI infrastructure are changing.

Gas demand will depend on data-center construction, server efficiency, grid expansion, renewable generation, battery deployment and the availability of alternative firm power. The projected 18 billion cubic feet per day establishes a scale that energy planners, regulators and technology companies must account for now. If the buildout proceeds, AI infrastructure will become a major force in US natural-gas markets rather than a small industrial load.