AI is scaling rapidly, and power infrastructure needs to keep up. As GPU-intensive workloads create sustained, highly dynamic demand, traditional approaches based on predictable loads, installed capacity, and backup availability are being challenged. In this session, Pierre-Adrien Bel will explore what it truly means to support AI, and why power performance under real-world load conditions is becoming a new measure of resilience. 

Pierre-Adrien Bel will examine how an integrated, system-level approach can help maintain stable, predictable performance as AI infrastructure scales.

Drawing on Rehlko’s latest research and its e-POD architecture, this session will demonstrate how designing, modeling, and validating the entire power system can reduce uncertainty, strengthen resilience, and give data center operators greater confidence in building to meet today’s demands.