Evidence-anchored synthesis paper reconciling AI's dual energy narrative: while efficiency gains reliably cut energy and carbon intensity per task, cheaper compute triggers Jevons rebound effects that expand absolute energy demand.
Public debates treat AI energy impacts in silos, claiming AI either saves the planet via green optimization or destroys power grids through data-center expansion. No unified framework existed reconciling task-level efficiency with macro-level rebound.
Synthesized empirical literature into a unified 4-mechanism model (efficiency progress & green innovation vs. compute demand & rebound) governed by 6 critical moderators (regulation, grid timing, spatial siting, AI maturity, human capital, and finance).
Demonstrates that AI energy impact is not a fixed trade-off but a governed relationship: AI lowers energy/carbon intensity while absolute emissions are dictated by regulatory, spatial, and temporal grid dispatch policies.
Ask Bilal’s AI Clone