aceModerate evidence
More
Model finds AI productivity gains add 0.5–1.8 gigatonnes of CO₂ a year
Study title: AI-driven productivity gains enable more CO₂ emissions than they avoid in a global energy–economy model
- Evidence type
- Benchmark or lab evaluation
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-2593cite or check it
Study title: AI-driven productivity gains enable more CO₂ emissions than they avoid in a global energy–economy model
Researchers modeled AI as a productivity booster that helps both fossil fuel and renewable energy supply, using a global computable general equilibrium model. When both sectors adopt AI at the same tier, net global CO₂ emissions rise by 0.5–1.8 gigatonnes a year, equal to 1.2–4.8% of 2024 energy-related emissions.
Read the study abstract
Abstract The net climate impacts of artificial intelligence (AI) depend largely on how its applications propagate through competing energy pathways. Predominant analyses examine the relationship between datacenter energy demand, renewables optimization, and demand-side efficiencies, but insufficiently address how AI also reshapes fossil fuel supply economics. We instead model AI as a bidirectional productivity amplifier in a global computable general equilibrium model, quantifying both enabled emissions from fos…
The problem
A global economic model suggests that AI productivity gains applied equally to fossil fuel and renewable energy would raise net annual CO₂ emissions by 0.5–1.8 gigatonnes, 1.2–4.8% of 2024 energy-related emissions.
In brief
The driver is upstream oil, gas and coal extraction. Small extraction gains outweighed larger gains in renewables generation, and renewables gains had to be 4–5 times larger than fossil gains for emissions to break even. Net reductions appeared only when fossil gains were zero.
The authors say the modeled indirect emissions are 3.3–13.3 times the International Energy Agency's 2025 datacenter emissions estimate of 0.18 gigatonnes. They argue that governance focused on datacenter footprints and avoided emissions misses this pathway. The results are comparative-static equilibria, not time-resolved forecasts.
Main points
- Under parallel adoption, net annual CO₂ emissions rose by 0.5–1.8 Gt, or 1.2–4.8% of 2024 global energy-related emissions.
- Enabled emissions of 0.6–2.4 Gt a year exceeded the IEA's 2025 datacenter estimate of 0.18 Gt by 3.3–13.3 times.
- Each 1% productivity gain in fossil fuels required 4–5% gains in renewables for net emissions to break even.
- The asymmetry held across 64 scenario combinations; emissions rose even with 15–20% renewables generation gains paired with 2–5% fossil extraction gains.
- A high fuel-neutral scenario covering grid and some demand-side efficiency cut net emissions by only about 0.1 Gt a year.
- At a carbon price of $80 per tonne, enabled emissions of 1.0 Gt still exceeded avoided emissions of 0.3 Gt; at $308 the net increase fell to 0.1 Gt.
- AI productivity raised CO₂ faster than GDP in parallel adoption scenarios, which the authors describe as negative decoupling.
The full story
The team used GTAP-E-Power, a global model of 20 economic sectors across five regions, calibrated to a 2017 base year and adjusted so renewables' share of electricity matched 2024 levels. AI adoption was represented as productivity shocks: 0–20% in fossil upstream extraction, 2–15% in refining, 2–10% in fossil power generation and 2–30% in renewables generation. Adoption tiers were calibrated to IEA 2035 estimates.
The authors cite evidence that fossil gains are already visible. Goldman Sachs analysts estimate AI could expand U.S. shale reserves by 8–20% and cut drilling costs by about 30%. An industry study found 44% of upstream organizations already use AI in exploration. For renewables, the authors found limited evidence of industry-wide realized gains beyond academic studies and pilots, yet still modeled gains at the upper bounds of documented technical potential.
Sensitivity tests varied two elasticities by ±50%, which moved the enabled-to-avoided ratio from roughly 2–3 times to about 7 times without reversing direction. Using the original 2017 baseline changed the ratio only marginally. The authors propose five governance priorities, including treating enabled emissions as a distinct policy category, constraining AI-enabled fossil productivity and pairing carbon pricing with supply-side limits.
Sources
1 on file- npj Climate Actiondoi:10.1038/s44168-026-00411-0
Talk it over
Who carries this cost, and what would have to change to prevent it?
Model finds AI productivity gains add 0.5–1.8 gigatonnes of CO₂ a yearhttps://truvace.com/item/ai-driven-productivity-gains-enable-more-co-emissions-than-they-avoid-in-a-globa
Who carries this cost, and what would have to change to prevent it?
The problem A global economic model suggests that AI productivity gains applied equally to fossil fuel and renewable energy would raise net annual CO₂ emissions by 0.5–1.8 gigatonnes, 1.2–4.8% of 2024 energy-related emissions.
Answer in the threadHow do we know? What can this kind of evidence show, and what can’t it?
The evidence This rests on a benchmark or lab test, from npj Climate Action.
Answer in the threadWhat does it leave open? What else would you want to know before acting on it?
The record’s own caveat The model is static and calibrated to a 2017 base year, so results show the direction and approximate scale of economic incentives rather than time-resolved forecasts, and they exclude datacenter emissions, non-CO₂ gases, infrastructure lock-in and political…
Answer in the threadHave you seen AI used this way where you study, work or live? Did it go the same way?
Answer in the thread
The questions are the same on every record. The details under them are taken from this record as written.
What readers say