AI Productivity Gains Drive Net CO₂ Increase in Energy–Economy Model

A paper in npj Climate Action models AI as a bidirectional productivity amplifier in a global computable general equilibrium model, finding that AI-driven productivity gains in fossil fuel extraction enable more CO₂ emissions than AI avoids through renewables optimization. Under parallel adoption scenarios, net annual CO₂ emissions increase by 0.47–1.8 gigatonnes — 1.2–4.8% of 2024 global energy-related emissions.

Key Findings

• Enabled vs. avoided emissions: Enabled emissions (from fossil fuel productivity gains) exceed avoided emissions whenever fossil-sector gains are nonzero. Net emissions reductions require renewables gains 4–5× greater than fossil fuel gains.
• Fossil fuel incumbency reinforced: AI lowers extraction costs and reduces operational risk, extending the economic viability of oil and gas. The paper notes the IEA's characterization that oil and gas production would fall ~8% annually without reinvestment — AI pushes back peak-supply expectations.
• Policy gap: Absent policy steering, AI's modeled effects increase the carbon intensity of the global economy, a dynamic current governance frameworks do not capture.

Methodology

The model treats AI as a general-purpose technology that amplifies productivity across both high-carbon and low-carbon energy supply pathways. It contrasts with predominant analyses that focus only on datacenter energy demand and renewables optimization, ignoring how AI reshapes fossil fuel supply economics.

Who It's For

Developers and analysts working on AI's climate impact models or energy-economy policy.