AI for the grid and AI on the grid: Implications, challenges and opportunities

The increasing integration of artificial intelligence (AI) into the electricity grid presents both significant opportunities and challenges. This paper examines two related dimensions: “AI for the Grid,” which explores the use of AI to enhance grid planning, operations, and market participation, and “AI on the Grid,” which considers the rising electricity demand from AI-driven data centers and its implications for grid infrastructure and cost allocation. AI offers opportunities for the grid through advanced tools to improve system planning, operations, and resilience while also promoting affordability and reducing barriers to market entry. However, challenges remain regarding the reliability and trustworthiness of AI models, regulatory complexity, coordination across stakeholders, and potential market power concerns. Rapid load growth caused by AI on the grid will stress the bulk power system, creating challenges for reliability, infrastructure planning, and managing unpredictable loads. However, if incentivized and managed appropriately, data centers can also enhance grid flexibility and responsiveness. We establish eight specific research needs to help unleash AI innovation, while maintaining system reliability and affordability. For the grid these include, 1) evaluating productivity gains and risks of AI-assisted services, 2) advancing AI methods for operations and planning, 3) evaluating the potential of private, local, small language models (SLMs), and 4) assessing AI-driven market power risks. Research needs on the grid include, 1) establishing incentives for large load flexibility, 2) enhancing market structures and designs, 3) developing new cost-allocation strategies, and 4) balancing the costs of over- and under-planning for load growth in the face of uncertainty.