Most talk about AI and climate change centers on data centers. These buildings use huge amounts of electricity to run AI systems.
That electricity use is growing fast. By 2030, it could reach nearly three times the combined yearly power use of Pakistan, Bangladesh, and Nigeria.
Even so, data center emissions make up less than 1% of global emissions today. That has let some groups argue AI's climate benefits, like improving renewable energy systems, cancel out the harm.
A new study says that framing misses the bigger problem. It focuses on how oil and gas companies use AI tools to find and extract fuel faster.
Holly Alpine used to work as a manager at Microsoft. She left her job in 2024 to study how tech companies help the fossil fuel industry.
Alpine and her husband teamed up with two researchers. Their peer-reviewed study came out last week.
The study found that AI use in the fossil fuel sector could add between 0.47 and 1.8 gigatonnes of CO2 each year. That range matches Mexico's yearly emissions on the low end and Russia's on the high end.
Those numbers are 3.3 to 13.3 times higher than the emissions from powering AI data centers. Alpine says this is the part of the story people are not talking about.
AI helps oil and gas companies process seismic and well data quickly. This lets them find fuel deposits with more certainty.
Alpine says this makes some drilling projects possible that would have been too slow or risky before. Rig counts have dropped, but production has gone up.
Why Enabled Emissions Aren't Counted
Alpine says Microsoft's fossil fuel teams have long been bigger than its renewable energy teams. She points to a 2019 example where ExxonMobil said Microsoft's technology helped it pump an extra 50,000 barrels of oil a day.
There was also a separate deal with Chevron. Alpine says the emissions tied to just those two deals equal 300% of Microsoft's own operational emissions, including its data centers.
Right now, there is no formal system for tracking these "enabled emissions." Alpine says that needs to change before real accountability can happen.
The study calls for more transparency around fossil fuel contracts. It also wants those contracts checked against climate targets tied to keeping warming under 1.5 degrees Celsius.
Alpine says the goal is not to ban AI use in the fossil fuel sector. Some AI tools, like methane leak detection, help lower emissions.
What Comes Next
The researchers are working to get these emissions counted in major climate frameworks. That includes the Greenhouse Gas Protocol and the Science Based Targets initiative.
In Europe, the AI Act once classified some AI uses as high risk, which would trigger more oversight. Alpine says a recent policy change called the Omnibus has limited that path somewhat.
Now, attention is turning to a new European framework called the Cloud and AI Development Act. Researchers hope it will include rules that account for how tech companies' AI tools are used by fossil fuel clients.
