First Full-Sector Study Finds AI Raises Net Carbon Emissions by Up to 1.8 Gigatonnes Annually
Summary
- • New study is the first to quantify AI's full power-sector climate impact — including fossil fuel productivity gains
- • Across 64 scenarios, net yearly carbon pollution rose 0.47–1.8 GT, roughly 1–5% of energy sector annual emissions
- • Emissions only fell in scenarios where AI did not boost fossil fuel sector productivity
- • Fossil fuel AI applications already at commercial scale; renewables AI still largely at pilot/academic stage
Details
First study to quantify AI's full power-sector climate impact
Previous research examined only datacenter energy use vs. renewable benefits, ignoring AI's role in boosting fossil fuel productivity — this study models both sides of the ledger.
Net carbon emissions rise 0.47–1.8 GT/year across 64 scenarios
This represents approximately 1–5% of the energy sector's annual emissions — a significant additional burden on top of existing fossil fuel output.
Emissions only fall when AI does NOT boost fossil fuel productivity
In every scenario where AI is adopted by both clean and dirty energy sectors, net emissions increase. Emission reductions only appear if the fossil fuel sector is excluded from AI adoption.
Renewables must outpace fossil fuels by 4x for emissions to break even
If AI adoption rates are similar across clean and dirty energy, renewable productivity gains must be at least 4x greater than fossil fuel gains just to achieve net-zero impact.
Fossil fuel AI at commercial scale; renewables AI at pilot stage
Fossil fuel operators (Saudi Aramco, Equinor, Adnoc) have real contracts and deployed AI at scale. Renewable AI applications remain largely speculative, blocked by permitting and interconnection barriers.
IEA: AI could boost recoverable oil/gas reserves by 5%, cut deepwater costs 10%
The International Energy Agency's estimates underscore the material economic and production impact of AI on fossil fuel output.
Rystad: AI to create ~$500B in fossil fuel E&P value by 2026–2030
Independent energy intelligence firm Rystad Energy projects AI and digitalization will deliver ~$500B in cumulative value for fossil fuel exploration and production companies in five years.
Fossil fuel AI gains emit at least 3x current datacenter estimates
Researchers found that AI-enabled fossil fuel productivity gains generate emissions at least three times current estimates for AI datacenter energy demand alone.
Datacenter energy demand not included in calculations
The study deliberately excluded datacenter energy use from its model, meaning the true net emissions impact of AI is even larger than reported here.
Key findings from the first study quantifying AI's full power-sector climate impact, published August 2026.
What This Means
This study fundamentally challenges the AI-as-climate-solution narrative by counting both sides of the ledger for the first time. The finding that AI's fossil fuel productivity gains already operate at commercial scale — while renewable AI benefits remain speculative and pilot-stage — means the emissions asymmetry is structural, not a temporary lag that market forces will self-correct. For tech companies promoting AI's green potential, policymakers weighing AI investment subsidies, and ESG investors assessing AI exposure, this research demands a serious reckoning: the net climate effect of unconstrained AI deployment across both clean and dirty energy sectors is, according to this model, consistently negative. Policy interventions conditioning AI subsidies on clean-energy sourcing, or restricting AI use in fossil fuel extraction, may be required to change the trajectory.