Running Meta's Data Centers with AI Agents - Sustainably

​As demand for AI Data Center grows, operating a large hardware fleet requires new approaches to reliability, power efficiency, and capacity. This fireside chat explores practical principles for building trustworthy AI agents and how they can support fleet operations through fault diagnosis, power modeling, and continuous feedback from production systems.

​We will also examine the sustainability side of the hardware lifecycle: measuring embodied carbon, incorporating sustainability into design decisions, extending equipment life, and enabling reliable reuse. Drawing on Meta’s experience, the session will show how automation, engineering discipline, and industry collaboration can help make hyperscale infrastructure both more efficient and more sustainable.

​Emre Tepedelenlioglu, PhD, Production Systems Engineer at Meta, leading work on hardware sustainability, embodied carbon, and circularity across Meta’s data-center fleet.

​Xun Jiao, Hardware Systems Researcher at Meta, building diagnostic tests and AI agents for reliable, efficient hyperscale AI infrastructure.