ZutaCore OmniTherm builds on the company’s patented two-phase, waterless liquid cooling platform. Instead of relying on water or air, the system uses a non-conductive dielectric fluid that instantly absorbs heat at the chip surface. The fluid vaporises on contact, carrying heat away without risk to surrounding electronics.
The technology operates safely in any position, vertical, horizontal, or fully inverted, a key capability as data centres adopt more modular and high-density rack configurations. It can also cool components located on the reverse side of processors, such as power delivery systems and PCIe cards, addressing one of the newest design challenges in AI hardware.
“High ampacity in rack DC busbars, now integral to AI architectures, is introducing new challenges to operators,” Alex Cordovil, Research Director at Dell’Oro Group, said “These architectural shifts are accelerating demand for direct-to-chip liquid cooling solutions capable of managing not only extreme heat flux, but also providing the design flexibility required to support modular, orientation-agnostic system architectures.”
This flexibility allows AI infrastructure designers to deploy compute clusters without being limited by gravity or mechanical layout. As the power density of chips continues to rise, orientation-agnostic systems are becoming essential for ensuring consistent performance across varying rack and chassis designs.
From rack to edge computing environments
ZutaCore describes OmniTherm as a drop-in compatible solution that fits within existing server footprints, avoiding the need for mechanical redesigns or new cooling infrastructure. This approach shortens deployment times and reduces the cost of introducing liquid cooling into legacy environments.
The design is aimed at a wide spectrum of environments, from hyperscale cloud data centres to edge AI systems, where cooling challenges are compounded by smaller enclosures and limited airflow. The ability to operate in any orientation means that the same cooling platform can support multiple architectures, from traditional racks to compact modular systems deployed closer to the network edge.
My D. Truong, ZutaCore’s Chief Technology Officer, said OmniTherm gives system designers the “thermal flexibility to design and run tomorrow’s compute loads on today’s infrastructure without the need to compromise.”
The company’s two-phase liquid cooling technology, known as HyperCool, already powers several commercial deployments, using dielectric fluid to absorb heat directly at the chip and dissipate it through a closed-loop system. OmniTherm extends that approach to address the latest challenges emerging from the rise of AI accelerators and new chip designs that require cooling on multiple surfaces.
Redefining the limits of AI performance
AI workloads are redefining the limits of data centre design. Chips now regularly exceed 700 watts per unit, and rack densities are increasing by as much as 50% per generation. These trends are forcing operators to rethink not just cooling technologies, but the entire architecture of the data centre.
Two-phase cooling, which allows liquid to change state from fluid to vapour and back again, is viewed as one of the most efficient ways to handle these thermal loads. By maintaining stable chip temperatures, it can also improve performance, reduce power usage effectiveness (PUE), and extend hardware lifespan, factors that are becoming critical in an energy-constrained AI economy.
As the global race to scale AI infrastructure continues, innovations such as ZutaCore’s OmniTherm underline the growing intersection between thermal science and computational performance. What was once a mechanical concern is now a central pillar of AI strategy, because in the era of machine intelligence, the ability to manage heat may determine who leads the next wave of computing innovation.
