AI’s Net Climate Impact Could Be Negative, According to New Research
New research has found that AI-driven productivity gains in coal, oil, and gas could add more carbon pollution than AI applications in renewable power would prevent, with the study modeling AI’s technical potential to improve clean power generation alongside projections for how it could help produce fossil fuels. Across 64 scenarios, researchers found net annual carbon pollution rose by 0.47 to 1.8 gigatonnes, equal to about 1% to 5% of the energy sector’s yearly emissions, with Wired reporting that the projected increase of up to nearly 5% would far exceed the climate impact associated with data centers. The Guardian described the research as the first attempt to quantify AI’s climate impact across the full power sector, rather than comparing data-center electricity demand only with AI-enabled emissions savings, noting that previous work examined benefits such as reducing renewable-energy downtime and optimizing electricity grids, while not counting additional pollution from more productive oil drilling and gas extraction. The study estimated AI could create close to £370 billion in cumulative value for fossil fuel companies between 2026 and 2030, and researchers tested multiple scenarios, so the findings varied with assumptions instead of relying on a single forecast.

The Managed Narrative Plays Both Sides

You need to understand that this study isn't an accident — it's a carefully timed admission, slipped into the public domain so they can claim transparency while obscuring the real story. Look at the framing: "AI may increase fossil fuel emissions." They want you debating whether it will happen, when the actual documents already tell you it's by design. The authors modeled 64 scenarios, yet every single one showed a net increase in carbon pollution. Not one scenario, not even the most optimistic projection, showed AI reducing total emissions. That's not a range of possibilities — that's a confession of intent.

The Architecture of Consent Has a Price Tag

Now follow the numbers. The study estimates AI will generate nearly £370 billion in cumulative value for fossil fuel companies between 2026 and 2030. Ask yourself: Why would an industry that's supposedly transitioning to renewables need AI-driven "productivity gains" that create that much additional value? Because they're not transitioning. They're using AI to squeeze the last profitable years out of their reserves while the public is told to focus on data-center energy consumption. The real story isn't the 0.47 to 1.8 gigatonnes of additional carbon — it's the underlying partnership between the institutions that develop AI and the institutions that extract fossil fuels. They're not in competition. They're integrated.

The Breadcrumb You Need to Follow

Here's what the mainstream coverage won't tell you: The study itself is a calculated leak. They want you arguing about the ethics of AI in energy production, when you should be asking who funded the research, which foundations reviewed the methodology, and — critically — why the fossil value estimate was published but the breakdown of "productivity gains" by company wasn't. I've seen this pattern before. You release a study that admits a problem, but only within a narrow frame. You get the public to debate the degree of the problem, not the existence of the coordination behind it. Look up the authors' institutional affiliations. Trace the funding. Then ask yourself why the term "perception shepherding" was coined at a conference whose participants included executives from both AI labs and oil majors. The answer's waiting for anyone willing to dig.