Their loads can also be dynamically volatile, highly concentrated, and difficult for traditional utility monitoring systems to see in real time. That combination is changing the way utilities and grid operators need to think about large-load interconnections and system reliability.
Unlike more conventional industrial loads, AI training and inference workloads can create rapid changes in demand that affect the transmission system in real time. As individual facilities and data center campuses continue to grow, understanding what is happening between the transmission grid and the GPU rack becomes increasingly important.
The consequences are already becoming visible. In July 2024, 1,500 MW of data center load disconnected from the grid in under one second during an event in Virginia — an incident the white paper examines alongside the broader reliability implications of rapidly expanding AI infrastructure. EGM's new white paper, Electrical Power Architecture, Characteristics & Requirements for Large AI-Based Data Centers, examines the electrical infrastructure required to support these facilities and the points at which their scale and operating characteristics begin to create wider grid reliability concerns