<p>Achieving low-carbon energy and information networks requires coordination between variable renewable energy supply and fluctuating traffic demand. Artificial intelligence (AI) systems offer tools both to optimize networks and to coordinate between them. Yet AI systems generate substantial carbon emissions — from model training, deployment and use, and the hardware life cycle — creating a paradox in which the solution contributes to the problem. This Review analyses the dual role of AI systems. We examine AI applications in energy supply networks and information networks, the capabilities required for supply–demand coordination, and the carbon emissions of AI systems themselves. Achieving low-carbon AI systems has become a central challenge to ensure that their environmental benefits outweigh their costs. Looking forward, we outline three research directions: low-carbon AI goals, energy-efficient large models, and AI-driven life-cycle management. Furthermore, we call for policy frameworks and industry strategies to achieve low-carbon energy and information networks.</p>