ZutaCore brings two-phase cooling to PCIe GPUs

Author: Joe Peck

ZutaCore, a developer of liquid cooling technology, has announced that its OmniTherm cold plate now enables waterless, two-phase cooling for manufacturers building servers with the NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs in a single-slot PCIe form factor, supporting full-power operation in standard enterprise and AI cloud server environments.

As AI inference expands across enterprise and cloud environments, PCIe GPU servers have become a common platform due to their relative ease of deployment, scalability, and compatibility with existing infrastructure.

However, as GPU power consumption rises, air cooling can become a limiting factor, restricting density, driving up fan power, and increasing the risk of thermal throttling during sustained workloads.

The company says OmniTherm addresses this by enabling a transition to two-phase liquid cooling without introducing water inside the server.

The single-slot design allows operators to increase accelerator density in standard server architectures while capturing heat into a liquid loop, reducing reliance on high fan speeds that can create excessive noise, waste power, and cause difficult operating conditions in the data centre.

“Enterprise and cloud operators want the flexibility of PCIe GPUs, but they also need density and sustained performance as power levels rise,” comments My D. Truong, CTO of ZutaCore. “OmniTherm delivers waterless, two-phase cooling in a single-slot form factor, helping data centres increase accelerator density while maintaining stable thermals for 24/7 AI workloads.”

Two-phase cooling for dynamic AI workloads

Production AI workloads – particularly inference – are rarely steady, fluctuating constantly and creating thermal swings that can affect performance and reliability.

ZutaCore says its two-phase approach is designed to respond to changing workloads, helping data centres maintain predictable performance under dynamic utilisation.

As racks move into higher power levels, the operational cost of air cooling also rises, with increased fan energy consumption and growing acoustic and facility pressures.

OmniTherm uses a sealed, non-conductive dielectric fluid system that captures heat without requiring facility water in the server, reducing cooling overhead and providing a path to scaling PCIe-based AI deployments.

Alongside this announcement, ZutaCore has also introduced HyperCool Cloud, a cloud-native operations platform designed to help data centres manage liquid cooling infrastructure.

The platform, the company says, provides “near-real-time” CDU telemetry, fleet-level monitoring, and alarm-to-resolution workflows, helping operators manage service response and uptime as deployments scale across sites and fleets.