AI Is Forcing a Rethink of Data Centre Cooling and Power Why cooling matters now? Data Centre Cooling trends to watch Rapid growth of AI is beginning to change the model What liquid cooling actually looks like? Why speed is becoming critical? Existing data centres are being redesigned for AI Cooling and power now go together AI infrastructure is becoming more industrial Billy Linehan

Schneider Electric and Motivair showcase how AI is reshaping cooling and power systems inside modern data centres

Irish Tech News joined a group of international technology and infrastructure media in Buffalo, New York last week for a Schneider Electric-hosted briefing and site visits focused on the infrastructure emerging around AI computing.

The programme included presentations from Schneider Electric engineers and executives, a visit to Motivair, the liquid cooling company acquired by Schneider Electric in 2024, and tours of TeraWulf's expanding AI infrastructure campus in New York State, where Schneider Electric technologies are being deployed as part of the wider infrastructure buildout.

A major theme throughout the visit was the speed at which new cooling, power and infrastructure solutions are now being deployed as AI systems dramatically increase computing density inside modern data centres.

For years, most data centres relied mainly on air cooling. Chilled air circulated through server halls to remove heat generated by computing equipment. That model worked reasonably well for earlier generations of enterprise computing and cloud infrastructure where heat densities were lower.

Companies such as NVIDIA have dramatically increased the processing power packed into these systems, particularly for AI training and inference workloads. The result is that modern AI racks can generate far greater levels of heat than previous generations of computing infrastructure.

According to Tuan Hoang, Head of Product Development and Innovation for Schneider Electric's cooling business, traditional air cooling systems are now approaching practical limits at very high rack densities.

"Liquid cooling allows us to do that," Hoang told me during the visit. "It has 4,000% more heat capacity than air."

At the Motivair facility and later at the TeraWulf site, journalists were shown examples of liquid cooling systems now being deployed beside high-density AI infrastructure. See photo above.

Instead of relying entirely on chilled air circulating through server halls, the systems use liquid circulating through cooling units connected into server environments to remove heat more efficiently and closer to the chips themselves.

The visual impression is quite different from the public image many people still have of data centres. Large pipework systems, cooling units, power systems and engineering infrastructure increasingly dominate these environments as AI deployments scale. Hoang stressed repeatedly that air cooling is not disappearing entirely.

"It is still necessary," he explained, particularly for wider facility environments and supporting infrastructure.

What is changing is the balance between air and liquid cooling as AI workloads become denser and more power-intensive.

Another major theme during the Buffalo visit was speed. AI infrastructure operators increasingly want facilities operational as quickly as possible because expensive GPU systems only generate returns once deployed and running. That pressure is driving growing interest in modular infrastructure, prefabricated systems and repeatable engineering designs which can be deployed more rapidly than traditional bespoke builds.

The scale and pace of construction at the Lake Mariner TeraWulf campus reflected that urgency. One AI-focused facility is already operational while further expansion continues across the wider site.

Hoang also discussed the growing challenge of adapting existing data centres for AI workloads rather than building entirely new facilities from scratch.

"A lot of customers are trying to retrofit existing data centres," he said during the interview, explaining that many operators are now attempting to adapt infrastructure originally designed for lower-density cloud computing.

That pressure is one reason modular cooling systems and repeatable infrastructure designs are becoming increasingly important as AI deployments scale.

The discussions also highlight