The impact of data center emissions has gained considerable attention, but a recent study from former Microsoft sustainability professionals highlights a potentially more concerning issue: the role of artificial intelligence (AI) in the oil and gas sector. This research, published in the journal *npj Climate Action*, suggests that while AI can improve productivity in fossil fuel extraction, it may also lead to a significant rise in global energy-related emissions.
According to the study, the increased use of AI could result in additional emissions equating to those of Mexico at a minimum, and potentially reaching levels akin to Russia, which is the fourth-largest greenhouse gas emitter in the world. The paper asserts that the emissions from AI's impact on fossil fuel production outweigh the benefits derived from its applications in clean energy technologies such as solar and wind power. This increase is also projected to surpass the expected emissions from the expansion of global data centers.
Will Alpine, one of the researchers involved, notes the intertwined relationship between technology firms and fossil fuel producers, indicating that these two sectors should not be considered in isolation. "There’s a self-reinforcing dynamic between their supply and demand," he emphasizes, marking one of the key points of the research.
Together with his wife and co-author, Holly Alpine, both former Microsoft sustainability experts, Will left the tech giant in early 2024 due to concerns over its collaboration with the oil and gas industry. They have since become advocates for raising awareness about how AI interacts with fossil fuel industries.
Historically, oil and gas companies have harnessed various AI technologies for enhanced resource exploration and development, actions that the Alpines argue contribute to increased fossil fuel reliance and hinder progress toward climate goals. While tech companies often evaluate their emissions and those from their supply chains, they frequently overlook the emissions facilitated by their technologies in boosting fossil fuel production—a phenomenon the Alpines term "enabled emissions."
Holly explains that tech firms focus predominantly on their operational emissions rather than these enabled emissions, creating a gap in understanding their full environmental impact. The new paper highlights the critical need to account for these emissions, which can pose a significant threat to the climate.
To analyze the potential impact of AI on fossil fuel emissions, the Alpines employed an intricate economic model designed to simulate various influences across the economy. Their model assessed AI's effect on productivity throughout the fossil fuels sector, encompassing extraction, refining, and energy generation, ultimately estimating how this would impact global emissions. The findings indicated that AI might increase energy-related emissions by 1.2 to 4.8 percent, a considerably greater figure than what previous data center emissions projections suggested.
Will Alpine expressed surprise at the scale of the predicted emissions increase. The collaboration between fossil fuel companies and data centers is intensifying; for instance, Chevron and Microsoft have announced plans for a substantial gas plant in Texas to provide power for Microsoft’s data centers.
During a recent analyst call, Jeff Gustavson, head of Chevron’s New Energies division, hinted that this partnership could also extend to enhancing Chevron's AI capabilities, stating they would utilize some of the computational power generated by the new plant for their internal AI initiatives.
"This partnership exemplifies the link between AI and fossil fuels," Will remarks about the collaboration.
Jon Koomey, an energy researcher who commented on the implications of this relationship, noted that while some proponents of AI claim it could help resolve climate issues, they often overlook its broader impacts across industries, including fossil fuels. Koomey acknowledges the study's findings as credible and supports the extensive research approach taken by the Alpines.
He points out that while AI can optimize data center cooling by 30-40 percent, it may simultaneously streamline fossil fuel extraction, a trade-off whose overall effects remain uncertain. The Alpines' research endeavors to shed light on these complex dynamics using macroeconomic modeling.
