AI Workloads Are Reshaping Data Center Cooling — How Munters Is Responding
The data center industry is in the middle of a thermal inflection point. For most of the last two decades, standard enterprise server racks operated at 5 to 15 kW of heat output — well within the capacity of conventional raised-floor CRAC cooling. The deployment of GPU-dense AI infrastructure has changed that calculation fundamentally. A single rack of current-generation AI accelerators can generate 60 to 100 kW of heat, and next-generation configurations are being designed for 150 kW and above. At these densities, conventional air cooling cannot move heat away from the rack fast enough to maintain safe operating temperatures.
Munters Data Center Technologies is expanding its product portfolio in direct response to these thermal requirements. The strategy has two components: enhancing air-side efficiency for the majority of racks in a mixed-use facility that still run conventional compute workloads, and developing precision cooling solutions for the high-density AI clusters that generate the majority of the heat load. The acquisition of Geoclima in 2024 — which added liquid-cooled chillers and precision dry coolers to the Munters portfolio — was a key element of this strategy.
On the air side, Munters indirect evaporative cooling technology remains highly relevant even in AI-era facilities. While liquid cooling handles the rack-level heat for GPU clusters, the facility-level thermal management — managing the heat rejection from cooling distribution units, maintaining appropriate ambient conditions in white space, and controlling humidity to prevent electrostatic discharge — continues to rely on air-side systems. Munters' modular indirect evaporative cooling units are designed to integrate with liquid cooling infrastructure and maintain overall facility PUE in the range of 1.1 to 1.2.
For data center operators in South India — including NTT, Yotta, STT GDC, and emerging regional colocation providers — the question of how to handle AI workloads is becoming increasingly operational rather than theoretical. Bengaluru and Chennai are seeing active discussions around GPU-as-a-service deployments and enterprise AI infrastructure. Arvi Fluid Systems can provide early-stage thermal design input and connect operators with Munters' data center engineering resources. Contact us to start a conversation about your cooling requirements for AI-era infrastructure.