Powering the AI Boom: How Data Centers will be Powered & Cooled
The artificial intelligence buildout has turned data centers into one of the fastest growing sources of electricity demand in the United States. After roughly fifteen years of flat consumption, data center electricity use has climbed sharply. The U.S. Department of Energy’s Lawrence Berkeley National Laboratory found that data centers consumed about 4.4 percent of total U.S. electricity in 2023 and projected that share could reach between 6.7 and 12 percent by 2028. To put it bluntly, taking directly from The International Energy Agency – there is no AI without energy; at the same time, AI has the potential to transform the energy sector. The IEA estimates global data center electricity demand could more than double by 2030 to roughly 945 terawatt hours, an amount comparable to the entire electricity consumption of Japan today.
That demand must be met by something, and increasingly the answer is natural gas. The grid cannot add renewable capacity and transmission fast enough to satisfy hyperscale operators who want firm, around the clock power. As a result, developers are turning to on site gas turbines and new combined cycle plants located next to the campuses themselves.
Where the Gas Will Come From
The most likely supply basins are the ones already producing the cheapest associated and dry gas. The Appalachian Basin, primarily the Marcellus and Utica shales across Pennsylvania, West Virginia, and Ohio, remains the largest gas producing region in the country, accounting for roughly a third of U.S. dry gas production. The Permian Basin in West Texas and southeastern New Mexico produces enormous volumes of associated gas alongside its oil, and the Haynesville shale in Louisiana and East Texas sits close to both Gulf Coast demand and existing pipeline infrastructure.
This proximity is shaping where the data centers themselves are landing. Northern Virginia’s “Data Center Alley” in Loudoun County remains the largest concentration in the world, but new growth is moving toward gas. Developers have announced massive campuses in West Texas, including projects tied directly to Permian gas, and in 2025 a wave of announcements targeted Louisiana, Texas, and the Appalachian region precisely because gas and land are abundant there. Meta’s roughly 4 gigawatt Louisiana campus and various Texas “behind the meter” gas projects illustrate the pattern of siting compute next to the fuel.
The H2S Problem in the Gas Stream
Raw natural gas is rarely pipeline ready or turbine ready as it comes out of the ground. Many producing zones, particularly portions of the Permian and certain carbonate formations, yield “sour” gas containing hydrogen sulfide (H2S). Hydrogen sulfide is highly toxic, corrosive to steel and turbine components, and regulated. Pipeline tariffs typically limit H2S to about roughly 4 parts per million across the board.
Before that gas can fuel a turbine, it must be sweetened. There are a variety of methods to treat sour natural gas, be it amine, triazine, non-triazine chemicals, or other catalysts. For a data center relying on local gas, this treatment is not optional. It is a prerequisite, and it adds cost, footprint, and an emissions and sulfur handling consideration to any “gas first” siting decision.
Cooling: The Water Question
Power is only half the resource story. Data centers also consume water, both directly for cooling and indirectly through thermoelectric power generation. Lawrence Berkeley’s 2024 report estimated that U.S. data centers directly consumed on the order of 17 billion gallons of water in 2023, a figure that could roughly double by 2028, before counting the water embedded in the electricity they buy. A single large facility using evaporative cooling can consume hundreds of thousands to millions of gallons per day.
Where does that water come from? Traditionally from municipal supplies and groundwater, which has sparked friction in drought stressed regions of Arizona, Texas, and elsewhere. To reduce that strain, operators are increasingly pursuing alternative sources, including recycled municipal wastewater and, notably in oil and gas country, treated produced water.
Produced Water and a Second H2S Concern
Produced water is the brine that comes up alongside oil and gas. The industry generates enormous volumes of it. Estimates put U.S. production well above 20 billion barrels per year, with the Permian Basin alone generating several times more water than oil by volume. Using this otherwise waste stream to cool data centers is attractive because it avoids drawing on freshwater, and Texas in particular has moved to encourage produced water reuse through legislation and a dedicated research consortium.
But produced water carries the same villain that appears in the gas stream. Sour formations produce sour water, and dissolved hydrogen sulfide is common in produced water from H2S bearing reservoirs. That presents corrosion risks to cooling equipment, worker safety hazards, and odor and air emission concerns. As a result, produced water destined for cooling must be treated, and H2S removal is part of that treatment, typically through aeration or air stripping, chemical oxidation, or biological and chemical scavenging, on top of the desalination and solids removal needed to make hypersaline brine usable. In other words, H2S appears twice in this story: once in the fuel and once in the cooling water, and both must be addressed before a data center can run.
The Takeaway
The future of AI will be determined as much by energy and water as by chips and software. The regions attracting the next wave of data center investment – the Permian, Appalachia, and the Gulf Coast – offer abundant natural gas, available land, and growing opportunities for water reuse. Those advantages are powerful, but they are not free.
Hydrogen sulfide sits at the center of both resource streams. It must be removed from the gas before it can fuel turbines and from the water before it can cool servers. The same molecule appears twice in the data center supply chain, creating treatment, infrastructure, and operating costs that cannot be ignored.
The AI boom may be digital, but its foundation is physical. Behind every data center are molecules, pipelines, treatment systems, and power plants. As AI scales, the winners will not simply be those who build the most computing capacity, it will be those who most effectively secure and manage the energy and water that make it possible.
For more on these thoughts with meaningful solutions, contact Q2 Technologies today.
Works Cited
International Energy Agency. “Energy and AI.” IEA, April 2025. https://www.iea.org/reports/energy-and-ai
U.S. Energy Information Administration. “Natural Gas Explained” and “Drilling Productivity Report.” EIA. https://www.eia.gov/petroleum/drilling/
S&P Global Commodity Insights. “US data center power demand and the natural gas response.” S&P Global, 2025. https://www.spglobal.com/commodityinsights/en/market-insights/latest-news/natural-gas
Kohl, A. and Nielsen, R. “Gas Purification,” 5th ed., Gulf Publishing; and U.S. EIA, “Natural Gas Processing.” See also pipeline quality H2S specifications summarized at https://www.eia.gov/energyexplained/natural-gas/
Texas Produced Water Consortium / Texas Water Development Board. “Produced Water in Texas” reports; and Scanlon, B.R., et al. “Will water issues constrain oil and gas production?” Environmental Science & Technology. https://www.depts.ttu.edu/research/tx-water-consortium/
“Energy, water use and pollution of AI and data centers rival most countries.” AP, 2026. https://www.yahoo.com/news/science/articles/un-calculates-nation-sized-environmental-140011529.html
