This story was originally published by Wired and is reproduced here as part of the Climate Desk collaboration.
The emissions from powering data centers rightly get a lot of attention. But new research from former Microsoft sustainability workers cautions that the way artificial intelligence enhances productivity in the oil and gas industry could be much more damaging to the planet.
Published last week in the journal npj Climate Action, the research finds that AI use could significantly increase global energy emissions simply by enabling the fossil fuel industry to produce more oil and gas. At the low end, the research finds that the additional yearly emissions could be equal to Mexico’s; at the high end, AI boosting the fossil fuel industry could add as much greenhouse gas emissions as Russia, the world’s fourth-largest emitter.
That increase in emissions, the paper finds, outweighs the benefits AI provides to developing solar, wind, and other clean technologies. It also significantly outpaces projections of emissions from the global data center buildout.
The relationship between technology companies and fossil fuels is “a self-reinforcing effect between supply and demand,” says Will Alpine, one of the authors of the research. “One of the key insights of our paper is that you cannot treat them independently. They are two sides of the same coin.”
Alpine and his wife and coauthor, Holly, are former Microsoft sustainability workers with years of experience between them at the company. They quit their jobs at the start of 2024 because of the company’s continued work with the oil and gas industry and began publicly campaigning to draw attention to the relationship between AI and fossil fuels.
Oil and gas companies have been using various types of AI for decades to help find and develop underground resources more efficiently. That increases global dependence on fossil fuels and makes it harder for the world to reach its climate targets, the Alpines argue.
While tech companies measure their own emissions and that of their supply chain, they don’t tend to measure just how much their tools help to increase fossil fuel production. The Alpines refer to AI-supported greenhouse gas pollution as “enabled emissions.”
“Sustainability measures [within tech companies] are very much focused on operational emissions” rather than enabled emissions, Holly says. The new paper argues that while accounting for that pollution is important, it ignores emissions that do much more damage to the climate.
To get an idea of how AI can affect fossil fuel companies’ emissions, the Alpines used a complex economic model that allows researchers to introduce various factors to test out how they might play out over the broader economy. Using reports from oil and gas companies about demonstrated gains from AI tools, the Alpines modeled AI as a productivity enhancer across various sectors of the fossil fuel industry, from extraction to refining to electricity generation. They also estimated how that in turn can increase global emissions.
The modeling found that AI as a productivity enhancer for the fossil fuel industry could help increase global energy-related emissions between 1.2 to 4.8 percent. That’s significantly higher than multiple projections around emissions from data center energy use.
“The scale of this was staggering,” says Will.
Fossil fuel companies are increasingly providing power for data centers. Chevron and Microsoft recently confirmed that the oil giant would be building a large behind-the-meter gas plant in Texas to power data centers for the tech company.
In a call with analysts in June, Jeff Gustavson, the president of Chevron’s New Energies division, hinted that the deal would benefit Chevron’s AI capacities as well. Chevron, according to Gustavson, will “use some of that compute” generated by the power from the power plant serving Microsoft “to actually power AI inside of our company.” (“Chevron and Microsoft have worked together for years to accelerate digital transformation, leveraging the capabilities of a trusted cloud to generate insights, scale innovation, and unlock value across the organization,” Chevron spokesperson Paula Beasley told WIRED in an email, in response to several questions about the specific relationship between the oil giant and the tech company.)
“It’s perfectly illustrative of the relationship between AI and fossil fuels,” Will says of the Chevron deal with Microsoft.
“There are many AI boosters who blithely claim that AI will solve the climate problem so we should go ahead and develop it as quickly as possible,” Jon Koomey, an energy researcher, tells WIRED in an email. “Such hand-waving arguments ignore the effects that AI will have on ALL industries, not just renewable energy and efficiency.”
Koomey, who was not involved in the new analysis, says that the conclusions the researchers draw seem solid.
“Machine learning can make data center cooling 30-40 percent more efficient, but [could] also make fossil fuel extraction much cheaper and faster,” he says. “How that nets out nobody yet knows for sure, but this new research is a credible attempt to answer that question using a macroeconomic model.”
