The Artificial Intelligence–Power Grid Nexus for Sustainability: AI–Power Grid Sustainability Nexus: Challenges and Path Forward

Abstract

Artificial intelligence (AI) has shown significant promise in improving power grid sustainability; however, a co-evolutionary framework is needed for sustainable AI and sustainable grid operation. There is a need for “Grid friendly AI', that guarantees a sustainable power grid. AI is expected to increasingly rely on resilient and reliable energy infrastructure, while simultaneously serving as a critical enabler for the optimization, predictive control, and decarbonization of sustainable power grids. AI technologies are being leveraged to improve grid flexibility, enhance forecasting accuracy, enable real-time decision-making for resiliency, and support greater integration of distributed energy resources. Conversely, the exponential growth of AI workloads, particularly large-scale training and inference, poses escalating energy and carbon demands, necessitating the development of cleaner, smarter, and more adaptive power infrastructures to sustain digital systems. This article explores the nexus between AI and the power grid through the lens of sustainability. We first examine how AI can potentially help accelerate the transition to clean energy, enabling greater penetration of renewable and sustainable energy, improved demand-side flexibility, and more efficient grid operations. We then investigate the increasing energy footprint of AI itself and propose strategies to power data centers and compute-intensive workloads with sustainable electricity. We asked the question- “What can the AI-power grid community do at the intersection of AI and power to make things more sustainable?". These include geographically aligning AI infrastructure with renewable generation, leveraging edge and federated AI to reduce energy consumption, and employing carbon-sensitive workload scheduling. In addition, we discuss pressing challenges at this nexus, including limited data accessibility, lack of interoperability standards, model explainability requirements, and evolving regulatory frameworks. To address these, we present a forward-looking architecture for aligning AI development with power grid evolution, grounded in co-design principles, energy-aware AI practices, sustainable digital infrastructure, and collaborative policy making. Achieving a sustainable, intelligent future will depend on viewing AI and power systems not as isolated domains, but as dynamically co-evolving systems, wherein AI contributes to grid decarbonization, and the grid, in turn, enables environmentally responsible AI. However, AI needs to become more efficient - it is not just about allowing unbridled growth of AI, but of Green AI.