Stanford University Invited Seminar: Powering AI—Meeting the Electricity Demands Needed for Frontier AI

Overview

About this event

On October 30, 2026, the Center hosted an international expert seminar at Yonsei University featuring Dr. Liang Min, Managing Director of the Bits & Watts Initiative at Stanford University. Titled “Powering AI: Meeting the Electricity Demands Needed for Frontier AI,” the seminar examined growing AI data center electricity demand in the United States and power supply strategies that jointly address reliability, affordability, and sustainability.

The presentation explained how expanding AI service use and computational demand can increase total electricity consumption despite improvements in computational efficiency. It also explored the geographic concentration of data centers and rapid load variations associated with synchronized GPU activity, emphasizing the need to consider grid hosting capacity and dynamic behavior alongside supply capacity. Generation and transmission interconnection delays, together with bottlenecks in individual lines and transformers, were discussed as constraints on connecting new data centers.

Proposed strategies included identifying available transmission capacity to inform siting and connection arrangements, and using dynamic line ratings and power flow control to address network bottlenecks. The seminar also explored flexibility from UPS batteries, on-site energy storage, and the scheduling of computational workloads. Coordinated planning of data center locations, grid connections, and clean generation capable of providing dependable supply was presented as another key approach.

From a policy and operational perspective, the session addressed fair allocation of grid upgrade costs attributable to large new loads, alongside long-term capacity commitments or minimum-payment agreements to reduce investment uncertainty. The discussion also covered low-voltage ride-through requirements and research directions for high-density power supply using HVDC and MVDC technologies. Participants considered data centers both as electricity consumers and as potential contributors to grid flexibility, gaining insight into the technical and institutional challenges of integrating large loads.