U.S. data centers are on track to become a major natural gas demand engine by 2035

U.S. data centers are projected to consume more natural gas than Germany and Japan combined by 2035, underscoring how the AI buildout is reshaping energy demand. A new BloombergNEF forecast says the facilities could burn about 18 billion cubic feet of gas per day by then, nearly twice the firm’s estimate from just nine months earlier.

The projection matters because it suggests the artificial intelligence boom is no longer only a story about chips, cloud contracts and software breakthroughs. It is increasingly a story about turbines, pipelines, power plants and gas prices, with implications for utilities, ratepayers and climate emissions.

BloombergNEF’s revised outlook reflects a more cautious view of what will actually get built, even as announcements from Big Tech continue to pile up. The report still points to data centers becoming the second-largest source of growth in U.S. natural gas demand over the next decade, trailing only LNG exports.

Why this forecast matters now

The new estimate arrives as the AI industry accelerates its infrastructure spending across the country. Companies racing to train and run larger models need massive amounts of electricity, and many of the fastest-moving operators are opting for gas-fired generation rather than waiting for grid upgrades that can take years.

That choice could redraw the country’s energy map. If the forecast holds, data centers would become one of the most important industrial drivers of gas demand growth in the United States, with consequences that reach well beyond the tech sector.

How big is the projected increase?

BloombergNEF says data centers could use roughly 18 billion cubic feet of natural gas per day by 2035. To put that in perspective, that is about as much as all data centers use today when natural gas used for grid electricity is included, and it is nearly double what the firm expected only months ago.

The revised figure also implies that the AI buildout is advancing quickly enough to matter even after accounting for project delays, cancellations and other forms of attrition that often affect large infrastructure plans.

What is driving the natural gas surge?

The surge is being driven by a mix of hyperscale demand, tight power timelines and a limited willingness to wait for new transmission or utility buildouts. In practice, that means more data centers are choosing to generate electricity on-site or to secure grid power from systems that still depend heavily on natural gas.

BloombergNEF’s analysis separates two broad pathways: facilities that build their own gas-fired power and those that lean on the grid. Both matter, but the grid-connected segment appears likely to shape overall demand far more than the much-discussed on-site plants.

Onsite power has grabbed headlines

Companies including Meta, Microsoft, Google and Amazon have all announced plans for natural gas power plants tied to new data center developments. Those projects have become a symbol of the industry’s willingness to bypass slow-moving infrastructure constraints in order to secure reliable power for AI workloads.

According to BloombergNEF, those onsite systems could account for 2.9 billion to 3.4 billion cubic feet of gas per day by 2035.

That is a significant amount of fuel, but the report suggests it is only part of the broader picture. In other words, the eye-catching headlines about tech companies building their own plants may be less important than the less visible, but larger, increase coming from ordinary utility-connected facilities.

Grid-connected data centers may have the bigger impact

By the middle of the next decade, BloombergNEF expects grid-connected data centers to add another 15 billion cubic feet per day of gas demand through the power sector. The firm says that amount of growth would be five times greater than the increase expected from all other grid-connected sectors combined.

That distinction is critical. It means the biggest climate and price effects may not come from a handful of branded corporate power projects, but from the cumulative load of thousands of AI-related servers drawing electricity from regional grids already under strain.

How could this affect gas prices and utilities?

If the forecast is realized, natural gas prices could rise, particularly as data center demand intersects with growing LNG export volumes. Analysts cited by BloombergNEF and Noreva argue that the current era of relatively stable gas prices may prove temporary if both pressures intensify at once.

For large tech firms, a price increase may be manageable. Their balance sheets can absorb bigger energy bills more easily than most industries can. The greater risk may fall on utilities and ordinary customers, especially in regions where new data centers are being connected to local grids without matching investments in generation or transmission.

Analysts say the assumption of permanently stable gas prices may be too optimistic, warning that the combined pull of AI infrastructure and export growth could push costs higher for everyone else.

That possibility has broad consequences. Utilities could face higher fuel procurement costs, pressure to add capacity faster, and renewed scrutiny over who pays for the infrastructure needed to support the next wave of digital demand.

What does this mean for the climate?

The climate stakes are substantial. The International Energy Agency says that every cubic foot of natural gas burned produces the equivalent of about 60 grams of carbon dioxide when extraction, processing and distribution are included. On BloombergNEF’s forecast, the added data center demand would generate about 1 million metric tons of greenhouse gas pollution each day.

That daily increase is roughly 12% of current total U.S. greenhouse gas emissions, according to the figures cited in the report. Framed another way, the AI sector’s appetite for reliable power could add emissions on a scale that rivals entire categories of national pollution.

This does not mean every data center will run on gas-fired generation directly. Some facilities may use a mix of sources, purchase renewable electricity or pair operations with storage. But the speed and scale of the buildout make gas the most readily available option in many places, particularly where developers want firm power now rather than after a long wait for grid expansion.

Why renewables alone are not solving the problem yet

Renewable energy is growing quickly, but utility-scale solar, wind, batteries and transmission lines still require land, permits, interconnection studies and construction time. AI infrastructure, by contrast, is often being planned on compressed schedules measured in months rather than years.

That mismatch is one reason gas remains attractive. It offers dispatchable power, established supply chains and a relatively fast path to operation. For developers under pressure to bring capacity online quickly, those features can outweigh longer-term climate concerns.

Which companies are shaping the trend?

Several of the world’s biggest technology companies are helping to define the new power equation. Meta, Microsoft, Google and Amazon have each announced plans tied to natural gas generation for data center projects, signaling that the AI race is now influencing energy strategy at the corporate level.

The common thread is not just computational ambition. It is the need for dependable electricity at industrial scale, often in places where the grid cannot immediately deliver enough capacity.

• Meta has signaled interest in onsite power for data center growth.
• Microsoft has also moved toward gas-linked generation for future capacity needs.
• Google is among the firms pursuing dedicated power solutions for AI infrastructure.
• Amazon has joined the group exploring gas-based supply for expansion.

These announcements do not guarantee every proposed project will be completed. BloombergNEF explicitly says its new forecast accounts for the fact that not all announced data centers or power plants will make it to operation.

How BloombergNEF revised its outlook

The new estimate is lower than a simple tally of announcements might suggest, because the analysts applied completion rates and other filters to the project pipeline. That caution is important: data center planning often outpaces what is eventually built, especially when financing, permitting or transmission issues intervene.

Even with those adjustments, the scale remains striking. The report implies that the AI boom is already large enough to alter national fuel demand trajectories, and it could still grow substantially if more projects move from announcement to completion.

What happens next?

The next phase will depend on whether AI growth continues at its current pace, whether utility systems can be expanded quickly enough, and whether policymakers or regulators intervene to change who bears the costs of new infrastructure. If the sector keeps scaling at this speed, the pressure on gas markets may intensify even before 2035 arrives.

There is also a broader strategic question for both technology companies and energy planners: whether the industry will continue relying on natural gas as the default bridge fuel or move faster toward lower-carbon alternatives such as nuclear, long-duration storage, advanced geothermal and accelerated grid upgrades.

For now, BloombergNEF’s forecast offers a stark reminder that the AI boom is not confined to the digital realm. Its physical footprint is expanding into power plants, fuel markets and the atmosphere itself.

Why this report stands out

What makes the forecast notable is not simply the size of the number, but the speed at which expectations are changing. In less than a year, BloombergNEF nearly doubled its estimate for data center natural gas use in 2035, a sign that the energy implications of AI are still being underestimated by many observers.

As the industry scales, every new server hall, training cluster and inference deployment adds another layer of demand. The question is no longer whether data centers will consume more energy. It is how much gas the U.S. will be willing, or forced, to burn to keep the AI race moving.

In plain terms: the AI buildout could turn U.S. data centers into one of the biggest new buyers of natural gas in the world, with major implications for prices, utilities and emissions.