The AI climate paradox: The green technology that could increase emissions
Integrating AI into the fossil fuel industry could increase carbon emissions by up to 5%, outpacing the environmental benefits of AI in renewable energy, according to a study published in Nature.
Efficiency gains and increased productivity resulting from the integration of AI technologies into the fossil fuel industry are expected to lead to an increase of up to 5% in carbon emissions, a growth rate greater than the reduction in emissions that will be made possible by integrating these technologies into renewable energy industries — this is according to a study published last week in the scientific journal Nature.
One of the most prominent promises of AI companies is that their models will help address a host of humanity's critical problems, including the climate crisis. This is to be achieved by finding solutions that humans alone could not develop or implement. The promise regarding the climate is particularly significant, as the massive demand for computing power created by the AI revolution has led to an increase in carbon emissions due to the high energy consumption of AI data centers.
But the new study, conducted by independent researchers who are former Microsoft employees, argues that the integration of AI technologies is actually set to increase the volume of carbon emissions. In their work, the researchers used a model to examine the effects of AI adoption on output and efficiency in two industries: fossil fuels and renewable energy.
"For example, AI adoption in oil extraction can reduce marginal costs and increase extraction efficiency, including a global expansion of supply, while in solar activity forecasting, AI can increase production efficiency from existing facilities," the study states.
According to the researchers, the findings suggest that AI adoption by the fossil fuel industry will lead to an annual increase of between 0.5 and 1.8 gigatons in carbon emissions, equivalent to an increase of 1.2% to 4.8% compared to the energy industry's global emissions volume in 2024.
"In this scenario, the increase in renewable energy output must be about 4 or 5 times greater than the increase in fossil fuel output in order to balance it. Minor improvements in fossil fuel extraction outweigh major improvements in renewable energy."
According to the findings, the improvements expected from AI adoption in renewable energy do mitigate the increase in carbon emissions, but only to the extent of 0.1 gigatons per year.
"Our findings challenge the dominant framing of AI's climate impact as a balance between the energy it consumes and the emissions it helps prevent. Instead, we characterize AI as a two-way amplifier: the same capabilities that improve renewable energy production also expand the economic feasibility of fossil fuels, in an asymmetric way. AI by its nature does not reduce or increase emissions. Its productivity improvements strengthen existing economic and institutional structures in the system in which it operates. Without policy guidance, AI will continue to entrench and preserve existing energy systems, rather than redirecting them."
