DATA CENTRES


Cooling gets clever A


I workloads are fundamentally changing how data centres generate, transfer, and reject heat. While rising rack densities are often cited as the primary challenge,


Maurizio Frizziero


the more profound shift lies in how thermal behaviour has become dynamic, uneven and closely coupled to workload patterns. For cooling professionals, this means traditional, static design assumptions are no longer suffi cient. Instead, eff ective cooling for AI infrastructure increasingly depends on viewing heat management as an end-to-end system. From capture at the rack and room level through to plant-side rejection or reuse, each stage of the thermal chain must work in concert. Cooling strategy is becoming adaptive by design, evolving alongside workloads rather than being fi xed at commissioning.


Why AI breaks conventional cooling models Conventional data centre cooling strategies were developed for applications with relatively stable electrical and thermal loads. Air-based cooling could be engineered around predictable airfl ow patterns, with incremental growth accommodated through conservative capacity margins. AI workloads disrupt this model. Training and inference


introduce sharp, high-amplitude swings in power consumption that translate directly into rapid thermal changes. These


Maurizio Frizziero, vice president of Chilled Water Systems at Vertiv, explores how rising AI workloads are pushing datacentre cooling toward smarter and more adaptive system designs.


fl uctuations can occur across heterogeneous hardware estates, where diff erent accelerators and processors exhibit distinct thermal characteristics. The result is greater variability at the rack and room level, with temperature stability depending not only on peak capacity but on responsiveness and control. In this context, cooling systems can no longer be treated as passive infrastructure. They must respond dynamically to workload behaviour, with closer integration between IT load, airfl ow management and heat rejection.


Capturing heat closer to the source As power densities increase, removing heat as close to its point of generation becomes essential. Traditional reliance on bulk air movement struggles to keep pace with high-density AI racks, where airfl ow requirements rise rapidly and fan energy consumption becomes a growing overhead. Direct-to-chip liquid cooling, particularly single-phase, has


therefore gained traction as an eff ective way to intercept heat directly at the processor level. By employing cold plates and dedicated liquid loops, these systems dramatically reduce thermal resistance, enabling high heat loads for artifi cial intelligence racks and Graphics Processing Units (GPUs). For data centre operators, this approach supports signifi cantly higher rack densities while greatly reducing reliance on high- volume airfl ow - although auxiliary air cooling may still be needed for the electrical components of the racks. To enable this liquid-cooled architecture, Coolant Distribution Units (CDUs) play an important role. CDU off erings span a wide range of capacities and include both in-rack and in-row confi gurations. These systems support liquid-to-air and liquid-to-liquid heat-exchange designs, making them versatile and suitable for diverse data centre layouts, including both existing facilities and new builds.


Immersion cooling uses a distinct thermal approach by


submerging servers in dielectric fl uid, eliminating air as a cooling medium. This enables uniform heat extraction across components, supporting very high thermal loads in a compact footprint - making it ideal for dense deployments like AI training clusters or high-performance computing pods. It also facilitates waste heat reuse, as the captured energy is available at relatively high and stable temperatures. However, immersion cooling requires specialised hardware,


fl uid management, and unique maintenance practices. It is most eff ective as part of a tiered cooling strategy rather than a one-size-fi ts-all solution.


14 April 2026 • www.acr-news.com


Download the ACR News app today


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36