As artificial intelligence reshapes the demands placed on digital infrastructure, data centres face mounting pressure to accommodate increasingly dense, high-performance workloads without sacrificing sustainability or uptime. A new cooling solution announced by ZutaCore points to how the industry may reconcile these conflicting pressures.
The company’s newly launched End-of-Row (EOR) Coolant Distribution Unit (CDU) family introduces a waterless, two-phase liquid cooling system designed specifically for AI and high-performance computing (HPC) environments. The 1.2MW and 2MW units offer a centralised approach that supports multiple racks from a single CDU while preserving rack-level control and monitoring. The system operates entirely within the white space of the data hall, eliminating the need for water infrastructure and reducing total cost of ownership.
As data centre operators contend with the energy and thermal implications of AI-scale compute, solutions that integrate performance, safety and sustainability are becoming essential. ZutaCore’s approach is based on its patented direct-to-chip cooling architecture, which uses a low global warming potential (GWP) dielectric fluid to reject heat through a closed-loop system, one that mitigates the risks of leakage and condensation common to traditional water-based methods.
From thermal bottlenecks to scalable control
Unlike conventional systems, which often rely on per-rack or per-zone cooling with duplicated infrastructure, the EOR CDU consolidates thermal management into a streamlined distribution model. This allows data centres to scale cooling capacity more efficiently while maintaining the responsiveness and redundancy needed for critical workloads. According to ZutaCore, hot-swappable components and multiple deployment configurations, including active-standby options for enterprise environments, offer the flexibility required to meet a wide range of operational risk profiles.
For AI infrastructure, the thermal burden is especially acute. The high power densities associated with large-scale inference and training operations push traditional cooling to its limits. ZutaCore’s design seeks to address this by combining advanced cooling physics with remote management via RESTful APIs, enabling real-time visibility and control across multiple CDU units. This kind of telemetry is increasingly important as AI data centres transition from static infrastructure to dynamic, high-throughput environments.
“Our new End-of-Row CDU family gives operators the control, intelligence, and reliability required to scale sustainably,” My Truong, CTO of ZutaCore, said. “By integrating advanced cooling physics with modern RESTful APIs, we’re enabling data centres to unlock new performance levels without compromising uptime or efficiency.”
The future of AI infrastructure is dry
The shift away from water-based cooling solutions reflects a broader industry reckoning with the environmental and operational limits of existing infrastructure. In regions where water usage is tightly regulated or grid constraints limit power availability, alternative cooling strategies are becoming critical. ZutaCore’s waterless approach speaks directly to these concerns, providing a technology that aligns with sustainability goals without compromising on performance.
The new EOR CDUs are already being integrated into live infrastructure. EGIL Wings, a developer of sustainable AI data centres, has confirmed that ZutaCore’s units will be deployed across its 15MW AI Vault platform. “By pairing ZutaCore’s waterless cooling with our sustainable power systems, we can deliver data centres that are faster to deploy, more energy-efficient, and ready for the global scale of AI,” Leland Sparks, President of EGIL Wings, added.
ZutaCore’s technology has already been proven in over 40 global deployments, including projects with Equinix, SoftBank, and the University of Münster. Strategic partnerships with Mitsubishi Heavy Industries, Carrier, and ASRock Rack are helping extend its footprint across geographies and into new classes of AI hardware, including factory-assembled, warranty-backed servers for NVIDIA HGX B300 systems.
As AI compute continues to drive density and power consumption in the data centre, the need for resilient, scalable, and sustainable cooling solutions is becoming urgent. ZutaCore’s new release suggests that waterless, direct-to-chip cooling, once a fringe concept, may now be moving to the centre of the AI infrastructure playbook.
