The rapid proliferation of AI-driven data centers (AI DCs) presents unprecedented challenges for their integration into bulk power grids. These facilities exhibit highly volatile, computation‑driven power demand profiles, pushing the limits of traditional grid planning, operations, and reliability frameworks. In this talk, I will outline the key technical and operational challenges posed by large‑scale AI DC integration, including load forecasting uncertainty, stability concerns, resource adequacy, and forced oscillation implications. Building on these challenges, I will discuss the spectrum of emerging research opportunities that span power systems engineering, data center architecture, control and optimization, and cross‑sectoral coordination. These opportunities highlight the need for new paradigms in flexible demand management, demand‑response mechanisms tailored to AI workloads, improved modeling of large computing clusters, and enhanced grid‑interactive data center designs.
Finally, the talk will conclude with a deep dive into one specific subproblem within this broader landscape, illustrating current methodologies, open questions, and future research directions.