Show summary Hide summary
- AI growth is rewriting global gas demand forecasts
- Climate concerns: 53 billion tonnes of future emissions
- Economic pressures and political choices
- Building a sustainable AI future: choices on the table
- How much of new gas capacity is linked to AI datacenters?
- Why are utilities choosing gas instead of renewables for AI loads?
- What are the projected emissions from planned gas plants?
- How will local communities near AI-linked gas plants be affected?
- Can AI growth be aligned with climate-friendly energy systems?
By the time a new AI datacenter in Texas powers up, nearby residents may face higher bills, hotter nights and more gas flares on the horizon. That local snapshot captures a global shift: US Spurs a new Global Rise in Gas-Powered Energy, just as climate science urges the opposite.
Behind the glowing server racks and sleek AI demos lies a simple equation: electricity demand is surging faster than the clean grid is expanding. The result is a wave of new gas power plants that could lock in decades of emissions. For communities from El Paso to western Pennsylvania, this is no abstract debate; it is about air quality, water use and the stability of household budgets.
AI growth is rewriting global gas demand forecasts
According to a recent Global Energy Monitor briefing, planned and under‑construction gas projects earmarked for 2026 are set to nearly triple existing global gas capacity. The United States alone accounts for almost a quarter of this pipeline, pushing a Global Rise in gas infrastructure.
Innovative Breakthrough Transforms Exhaust CO2 into Valuable Materials
UK Joins Nine Nations in Ambitious Plan to Develop a 100GW Offshore Wind Power Network in the North Sea
Much of this new capacity is tied directly to AI Growth. GEM estimates that about a third of the 252 gigawatts of gas power in development worldwide could be built on or next to datacenters. By comparison, global datacenter electricity use was roughly 185 terawatt-hours in 2023; US utilities now project up to 900 TWh by 2035. This steep climb is reshaping how grid planners, investors and regulators think about the future of power.
From shale wells to server halls: how the chain works
The physics is straightforward. Training large language models or running AI-driven cloud services requires vast arrays of GPUs. These machines run almost constantly, converting electricity into heat that must then be removed by power-hungry cooling systems. Utilities facing that near‑24/7 demand are turning to gas turbines that can run continuously and ramp quickly.
Upstream, producers are responding. The US is already on track to produce record volumes of natural gas, driven by export demand and domestic power needs, as highlighted in recent US gas production analyses. Every new AI campus becomes another anchor customer for pipelines, liquefied natural gas terminals and gas‑fired plants, hardening the role of Fossil Fuels in an era that is supposed to be about rapid Energy Transition.
Climate concerns: 53 billion tonnes of future emissions
The climate arithmetic is equally stark. GEM’s global tally suggests that if all planned gas power projects proceed, they would emit about 53.2 billion tonnes of CO₂ over their lifetimes. US projects alone could account for 12.1 billion tonnes, roughly twice the country’s current annual emissions from all sources.
For comparison, the Intergovernmental Panel on Climate Change indicates that limiting warming to 1.5°C requires global CO₂ emissions to fall almost in half by 2030 relative to 2019 and reach net zero around mid‑century. New long‑lived gas plants point in the opposite direction. That extra half‑degree of warming is not an abstract number; it is often the difference between stressed coral reefs and widespread coral bleaching, or between rare heatwaves and deadly routine extremes.
Who feels the environmental impact first?
The immediate Environmental Impact falls on communities near both wells and wires. In Texas’s Permian Basin, expanded production to feed LNG exports and future AI‑linked demand means more drilling, more flaring and more methane leaks. In western Pennsylvania’s Homer City, residents now face the prospect of a shuttered coal station reborn as the nation’s largest gas‑fired plant serving a datacenter campus.
Local groups, such as the Clean Air Council, warn about increased nitrogen oxide emissions, ammonia from cooling systems and potential water stress. Those concerns echo global research on how major cities already face rising hydrological pressure, as shown in analyses of water‑stressed urban areas. Adding water‑intensive cooling for AI into already strained watersheds risks compounding those vulnerabilities.
Economic pressures and political choices
There is also a pocketbook dimension. Analysts at the Union of Concerned Scientists estimate that US electricity demand could climb around 60% by 2050 under aggressive datacenter scenarios. In many regions, the first signs are already visible in grid interconnection queues and updated utility plans that forecast rising gas generation through the 2030s.
Some forecasts, such as those compiled by RMI and summarized in recent power demand assessments, suggest an 18% increase in US gas‑fired generation between 2024 and 2035. Consumers may shoulder the costs of new pipelines, plants and grid reinforcements, even when these assets mainly serve private AI loads. That is why grassroots backlash has emerged in places where power prices and local pollution are rising while promised jobs remain uncertain.
AI boom versus energy transition targets
The tension is obvious: governments have signed climate pledges, while industrial policy pushes for AI leadership. As one energy economist put it, the US is “building yesterday’s infrastructure for tomorrow’s technology.” Locking in gas plants with 30‑ to 40‑year lifetimes makes every future clean‑energy decision harder. Retrofitting with carbon capture or hydrogen blends remains expensive and unproven at the necessary scale.
Yet there are alternatives. Studies from institutions including the US National Renewable Energy Laboratory and the International Energy Agency show that high shares of wind, solar, storage and demand‑side flexibility can reliably meet rising loads. The question is not whether AI can exist in a low‑carbon system, but whether policy will align deployment timelines with climate limits.
Building a sustainable AI future: choices on the table
Consider “NorthRiver Cloud Campus,” a hypothetical AI complex outside Atlanta. Its initial plan leans on a new gas plant and minimal efficiency standards. Reworked, it could instead sign long‑term contracts with regional solar and wind farms, add on‑site batteries, use advanced liquid cooling to cut energy use, and shift non‑urgent computing to times when renewable output is highest.
Scaled across hundreds of datacenters, similar approaches could turn a wave of new demand into a driver of Sustainability rather than a brake. Some European facilities already match their consumption hourly with renewable generation, rather than relying on annual offsets. That standard is technologically feasible in the US as well, especially if regulators and investors demand it.
- Efficiency first: Set strict energy‑use limits per unit of AI workload and require best‑available cooling technology.
- Clean power commitments: Mandate that new large datacenters procure near‑100% carbon‑free electricity on an hourly basis within a defined timeframe.
- Grid‑friendly design: Encourage flexible workloads that can ramp down during stress events and ramp up when renewable output is high.
- Local safeguards: Tie permits to community benefits, transparent water‑use plans and air‑quality monitoring.
- Planning reform: Integrate AI demand scenarios into long‑term resource plans, giving renewables, storage and efficiency a fair comparison with gas.
These measures do not eliminate the need for new infrastructure, but they change its composition. In the same way that earlier digital booms drove fibre‑optic expansion, the AI surge can be directed to strengthen a cleaner grid rather than entrenching Fossil Fuels.
How much of new gas capacity is linked to AI datacenters?
Global Energy Monitor estimates that about one third of the 252 GW of gas power now in development worldwide may be built on or near datacenters. These facilities are being designed to serve the continuous, power‑hungry demand of Artificial Intelligence and cloud computing, especially in the United States, China, Vietnam, Iraq and Brazil.
Why are utilities choosing gas instead of renewables for AI loads?
Gas turbines provide dispatchable, around‑the‑clock power and can ramp quickly, which appeals to utilities facing rapid demand growth. Many grid operators argue they cannot yet rely on variable wind and solar plus storage for all new demand. However, recent studies show that combining renewables, batteries, demand response and grid upgrades can meet much of this load without locking in long‑lived gas plants.
What are the projected emissions from planned gas plants?
If every planned gas‑fired project moves forward, Global Energy Monitor projects about 53.2 billion tonnes of CO₂ over their operating lifetimes. US projects alone would represent around 12.1 billion tonnes. These future emissions would make global climate goals significantly harder to reach, particularly the objective of limiting warming to around 1.5°C above pre‑industrial levels.
How will local communities near AI-linked gas plants be affected?
Communities close to new gas plants and datacenters may experience increased air pollution, higher water consumption for cooling, noise, industrial traffic and higher electricity prices. In places such as Texas and western Pennsylvania, residents and advocacy groups have already raised concerns about health impacts and whether the benefits of AI‑driven investment truly outweigh these local costs.
Can AI growth be aligned with climate-friendly energy systems?
Researchers Manipulate Minuscule Crystals to Harness Electrical Properties
Powergen 2026’s second keynote highlights potential challenges for the grid
Yes. AI can be powered mainly by renewables if datacenters commit to strong efficiency standards, procure carbon‑free electricity on an hourly basis, use storage, and design flexible workloads. Policy tools, from clean‑energy mandates to improved grid planning, can steer investment away from gas expansion and support a faster, more resilient Energy Transition compatible with climate targets.
