Data centers rely on natural gas to power AI expansion

Tech companies are increasingly using natural gas turbines and engines to generate on-site electricity for data centers amid surging AI demand. This trend is leading to a boom in fossil fuel projects, particularly in the United States. Experts warn it could lock in higher emissions and hinder renewable energy adoption.

The artificial intelligence boom has spurred a rapid expansion of data centers, creating intense pressure on electricity supplies. Developers are turning to behind-the-meter power generation, often powered by natural gas, to meet the needs of these facilities. According to Cleanview, at least 46 data centers with a combined capacity of 56 gigawatts are employing this approach.

One notable deal involves Boom Supersonic, which signed a $1.25 billion agreement with developer Crusoe to supply 29 jet-engine gas turbines for OpenAI's data centers across the U.S. In Texas, Crusoe's Stargate campus in Abilene requires 1.2 gigawatts and uses aeroderivative turbines based on airplane models. Meta's data center in El Paso will draw power from more than 800 mobile mini-turbines, while a West Virginia facility has received gas engines from Caterpillar.

Global Energy Monitor reports that natural gas-fired power projects totaling over 1,000 gigawatts are in development worldwide, a 31 percent increase from the previous year. The U.S. accounts for about a quarter of this pipeline, with more than a third dedicated to data centers. In Texas, nearly 58 gigawatts of natural gas power are in planning or construction, with almost half serving data centers exclusively, without grid connections.

This shift raises climate concerns. Researchers at Cornell University estimate the data center build-out could add 44 million metric tons of carbon dioxide emissions by 2030, equivalent to the annual output of 10 million passenger cars. Cara Fogler of the Sierra Club described it as "a huge proposed build-out," noting that existing coal plants and new gas projects may "box out clean energy."

Alternatives like reciprocating engines, used by Titus Low Carbon Ventures, offer flexibility for fluctuating loads but are less efficient than combined-cycle turbines. Jenny Martos of Global Energy Monitor explained that such engines "produce more emissions than the others." In New Mexico, the proposed $165 billion Project Jupiter campus would rely on simple-cycle gas turbines, prompting opposition from the Center for Biological Diversity over its potential emissions.

Labaran da ke da alaƙa

Permits for 11 natural gas-powered data centers across the United States project annual greenhouse gas emissions exceeding 129 million tons—more than Morocco released in 2024. Linked to AI companies OpenAI, Meta, Microsoft, and xAI, these facilities underscore the fossil fuel reliance in the data center boom fueling AI expansion, according to a WIRED review of air permit documents.

An Ruwaito ta hanyar AI

President Trump met with executives from major tech companies last month, where they signed a voluntary pledge to cover energy costs for their data centers. The agreement, dubbed the Ratepayer Protection Pledge, includes securing their own power supplies and funding necessary infrastructure. Critics have dismissed it as lacking enforcement mechanisms.

Residents in rural areas across the United States are opposing new data center projects over concerns about water supplies, energy costs, and farmland loss. In Tazewell County, Illinois, locals successfully blocked a proposed facility after packed council meetings and petitions. The backlash highlights tensions between AI infrastructure growth and agricultural needs.

An Ruwaito ta hanyar AI

Data centres powering artificial intelligence are causing land surface temperatures to rise by an average of 2°C near their locations, with extremes reaching 9.1°C, according to researchers at the University of Cambridge. The heat effects extend up to 10 kilometres away, potentially impacting more than 340 million people worldwide. The findings come from a study analysing satellite data over the past two decades.

Wannan shafin yana amfani da cookies

Muna amfani da cookies don nazari don inganta shafin mu. Karanta manufar sirri mu don ƙarin bayani.
Ƙi