The rapid acceleration of AI workloads has fundamentally transformed data centre thermal design — demanding a shift from traditional air cooling toward hybrid liquid/air architectures. This session examines the critical cooling challenges associated with high-density GPU deployments across the full thermal chain — from cold plate technology and secondary fluid networks to large-scale heat rejection.

Four key challenges on the road to densification are addressed: mission-critical secondary fluid networks requiring uninterrupted operation; flexibility as the air-to-liquid cooling mix evolves; growing constraints on heat rejection as IT footprints shrink; and the imperative to reduce mechanical power at constrained sites. Liquid cooling system design is explored — including supply water temperature management, advanced feed-forward control strategies for dynamic GPU load transients, and comparative thermal solution architectures demonstrating Peak PUE reductions. Emerging strategies are also presented, positioning operators to meet both performance and sustainability objectives in an era of rapidly evolving AI infrastructure.