As AI Scales, Can Traditional Cooling Keep Up?
Data Center Knowledge, Thursday, June 4th, 2026
AI workloads demand new cooling strategies as air cooling struggles with higher rack densities.
AI and high-performance computing are straining traditional data center cooling infrastructure. Accelerator servers draw far more power than conventional systems while running at sustained high utilization, making heat removal critical to performance and reliability.
Removing more heat with air requires moving much larger volumes of air, raising fan energy consumption. Rising global data center electricity demand, projected to reach 945 terawatt-hours by 2030, plus water scarcity concerns are pushing operators toward liquid and waterless cooling, especially two-phase direct-to-chip systems using dielectric fluids.
Industry data shows 22% of organizations now use direct liquid cooling, driven mainly by higher rack densities, making cooling central to infrastructure planning.