As AI accelerates the need for data center design, the focus tends to gravitate toward what goes along with this: new challenges, structural limitations and cutting-edge technologies like next-generation chips and increasingly sophisticated controls. But, beneath these advances is the infrastructure that determines whether they perform reliably and efficiently.

Back to basics: Engineering fundamentals

Consider advanced liquid cooling. By transferring heat more efficiently than traditional air-cooling methods, liquid cooling can support higher-density computing environments while, in many designs and depending on facility, local climate and nearby heat demand, reduce reliance on energy-intensive chillers. It can also create opportunities for waste heat recovery and reuse, along with wastewater treatment and recycling, returning water to the closed-loop environment, and supporting progress toward net zero water usage in data centers. These benefits have created a surge in adoption, with liquid cooling penetration for high-end AI chips surpassing the 50% threshold in 2026, according to recent analysis.

While liquid cooling is often framed as a new innovation, the principles behind it are rooted in one of the foundational concepts of heating and cooling design: hydronics.

Long considered the backbone of HVAC heating and cooling applications, hydronic systems are directly relevant to data centers. Just as every component in a data center functions as part of an interconnected ecosystem, a hydronic system relies on the seamless interaction of its individual components. Every balance valve, fitting, connection point, heat exchanger, tank and pump must work in harmony throughout the cooling loop to ensure continuous performance.

And whether heat is removed through a chilled-water system, rear-door heat exchanger or direct-to-chip loop, the underlying task remains the same: move thermal energy from where it is generated to where it can be rejected or reused.

That requires more than circulating fluid through the system. Designers must account for things like flow, pressure, temperature, system, redundancy and equipment balance, which all help determine whether heat can be moved efficiently from the chip to its ultimate point of rejection. 

Back to where it all began

In 1954, Xylem’s Bell & Gossett engineer Gil Carlson developed the concept of primary/secondary pumping, which became the standard technique adopted by contractors, engineers and boiler manufacturers alike. In this piping arrangement, two separate water circulation loops are employed so that the equipment loop and building distribution loop can operate somewhat independently.

Today, this same concept directly informs many liquid cooled data center designs, whereby primary and secondary loops work together to move heat from IT equipment to heat rejection systems and illustrating an inherent hydronic principle: connected cooling loops operate most effectively when they maintain hydraulic independence yet remain thermally linked.

When densities are high, the stakes are higher

AI densities are creating intense heat loads that are changing the way facilities respond to these shifting dynamics and as new computing comes online. In response, cooling loops must scale in accordance with the capacity required to manage high-density AI networks.

Hydraulic separation can support this flexibility. Instead of designing every component around a single, fixed operating point, for instance, engineers can create circuits that accommodate additional cooling capacity as racks and data halls are deployed.

But hydraulic independence does not mean the loops operate without affecting one another. Rather, they remain thermally connected. If secondary flow exceeds what the primary side can effectively support, recirculation and mixing can increase the temperature of the fluid supplied to the load. If the primary side circulates more water than necessary, the system may consume excess pumping energy and compromise its intended temperature differential.

Similarly, decisions made at the rack can affect everything from piping and pumping to heat rejection, water use and power demand throughout the facility. Hydronic infrastructure is designed with this compounding efficiency in mind, knowing that each individual piece of equipment has an important role to play toward the performance of the greater whole.

The intelligence of any cooling system begins with sound hydronic system design, all around a common objective to deliver the right flow, temperature and pressure at the right time. The technology may evolve, but performance is ultimately determined by engineering fundamentals, practical application and the expertise of facility managers, engineers and operators who bring these systems to life.