AI workloads are redefining data center design, pushing power densities beyond what air alone can handle. While air cooling will continue to serve vital roles across existing infrastructure, liquid cooling is becoming essential to sustain the performance and efficiency demanded by AI-class systems. The most future-ready operators aren’t replacing one with the other; they’re strategically integrating both—guided by partners who understand the entire thermal chain from silicon to chiller.
The current discussion on air cooling vs. liquid cooling revolves around the need to dissipate heat generated by artificial intelligence (AI) infrastructure. AI requires new, denser hardware — high-density chips and Graphics Processing Units (GPUs) — that generate far more heat than legacy workloads. As a result, air cooling is reaching its thermal limits. Air cooling systems cannot push enough air to dissipate all the heat generated by GPUs. You would have to place high-density AI racks in a wind tunnel for air cooling to handle that much heat.
Compounding the challenge, air-based cooling already accounts for up to 40% of a typical data center’s total electricity use, according to a recent Deloitte insight report. It’s one of the reasons liquid cooling is gaining favor, as it provides better thermal transfer and reduces energy consumption.
Is liquid cooling really more energy-efficient than air cooling?
Liquid cooling reduces cooling energy use by up to 90%, compared with air-based systems. So, in addition to addressing the extreme heat generated by AI hardware, it makes data center operations more efficient. Liquids are much better heat conductors than air, and liquid cooling solutions use less energy than the fans required for air cooling. As such, liquid cooling and hybrid solutions are crucial components for efficient AI and HPC workloads.
A liquid cooling data center for optimized AI workloads
To optimize AI workloads, liquid cooling is a necessity, not merely an option. It enables high-density chips in AI servers to run faster and longer, and, as such, has become a priority in new data center construction and retrofits. But switching from air cooling isn’t as simple as flipping a switch. Air cooling will remain necessary for the foreseeable future, handling about 20% of heat dissipation.
Chip innovation now moves at the speed of consumer tech. As semiconductor leaders AMD and NVIDIA release new AI processors, data centers must evolve, adopting liquid cooling at the core of their thermal strategy.
How can data centers transition to liquid cooling?
There are several liquid cooling methods. For example, Motivair by Schneider Electric systems use the single-phase direct cooling method, which involves circulating coolants through a closed-loop system to dissipate heat. This method has been in the market longer than other approaches, such as single-phase immersion, two-phase immersion, and two-phase direct liquid cooling. It provides the most scalable, efficient form of liquid cooling.
As data centers become GPU-focused, operators need to strategize for the adoption of liquid cooling:
Rear-door heat exchanger (RDHx) – Motivair’s ChilledDoor® enables hybrid approaches for the transition to liquid cooling, such as air-assisted liquid cooling. Fans mounted on a door at the back of a rack transfer heat from servers to a closed-loop water coil, which absorbs the heat, releasing cooler air into the data center. Cooling occurs directly at the heat source with minimal efficiency loss. This method is easily deployable in existing data centers without requiring a full retrofit.
Heat Dissipation Unit (HDU) – This air-to-liquid heat exchanger provides another transitional approach for data centers. An HDU can be deployed in data centers without plumbing or without existing access to water. The unit circulates coolant from CPUs and GPUs in a closed loop, with fans blowing on the coils to dissipate heat. These units have cooling capacity limitations but provide a bridge to liquid cooling in legacy, air-cooled data centers.
Coolant Distribution Unit (CDU) – This liquid-to-liquid method is more suitable for new construction or existing facilities with existing access to water. It requires a plumbing infrastructure to run a closed-loop system that circulates coolant from servers and other equipment to the facility’s chilled water plant, where the heat is removed.
Balancing air and liquid data center cooling for the AI era
Liquid cooling offers the performance and efficiency needed to manage the intense heat of GPU-driven workloads, reducing energy use while improving reliability. The transition won’t happen overnight. Many facilities will adopt air-liquid hybrid systems through solutions like rear-door heat exchangers and HDUs. These enable a gradual shift toward AI-ready, sustainable operations. Discover end-to-end cooling solutions that redefine performance for AI & HPC workloads. You can also contact a Schneider Electric expert to help you evaluate and design a cooling strategy that supports your data center’s future growth.
Add a comment