News

CDUs: The brains of direct liquid cooling

As air cooling reaches its limits with AI and HPC workloads exceeding 100 kW per rack, hybrid liquid cooling is becoming essential. To this, coolant distribution units (CDUs) could be the key enabler for next-generation, high-density data centre facilities. In this article for DCNN, Gordon Johnson, Senior CFD Manager at Subzero Engineering, discusses further the importance of CDUs in direct liquid cooling: Cooling and the future of data centres Traditional air cooling has hit its limits, with rack power densities surpassing 100 kW due to the relentless growth of AI and high-performance computing (HPC) workloads. Already, CPUs and GPUs exceed 700–1000 W per socket, while projections estimate that to rise to over 1500 W going forward. Fans and heat sinks are just unable to handle these thermal loads at scale. Hybrid cooling strategies are becoming the only scalable, sustainable path forward. Single-phase direct-to-chip (DTC) liquid cooling has emerged as the most practical and serviceable solution, delivering coolant directly to cold plates attached to processors and accelerators. However, direct liquid cooling (DLC) cannot be scaled safely or efficiently with plumbing alone. The key enabler is the coolant distribution unit (CDU), a system that integrates pumps, heat exchangers, sensors, and control logic into a coordinated package. CDUs are often mistaken for passive infrastructure. But far from being a passive subsystem, they act as the brains of DLC, orchestrating isolation, stability, adaptability, and efficiency to make DTC viable at data centre scale. They serve as the intelligent control layer for the entire thermal management system. Intelligent orchestration CDUs do a lot more than just transport fluid around the cooling system; they think, adapt, and protect the liquid cooling portion of the hybrid cooling system. They maintain redundancy to ensure continuous operation, control flow, and pressure, using automated valves and variable speed pumps, filtering particulates to protect cold plates, and maintaining coolant temperature above the dew point to prevent condensation. They contribute to the precise, intelligent, and flexible coordination of the complete thermal management system. Because of their greater cooling capacity, CDUs are ideal for large HPC data centres. However, because they must be connected to the facility's chilled water supply or another heat rejection source to continuously provide liquid to the cold plates for cooling, they can be complicated. CDUs typically fall into two categories: • Liquid to Liquid (L2L): Large HPC facilities are well-suited for high-capacity CDUs known as L2L. Through heat exchangers, they move chip heat into the isolated chilled water loop, such as the facility water system (FWS). • Liquid to Air (L2A): For smaller deployments, L2A CDUs are simpler but have a lower cooling capacity. By utilising conventional HVAC systems, they transfer heat from the returning liquid coolant from the cold plates to the surrounding data centre air by using liquid-to-air heat exchangers rather than a chilled water supply or FWS. Isolation: Safeguarding IT from facility water Acting as the bridge between the FWS and the dedicated technology cooling system (TCS), which provides filtered liquid coolant directly to the chips via cold plate, CDUs isolate sensitive server cold plates from external variability, ensuring a safe and stable environment while constantly adjusting to shifting workloads. One of L2L CDUs' primary functions is to create a dual-loop architecture: • Primary loop (facility side): Connects to building chilled water, district cooling, or dry coolers • Secondary loop (IT side): Delivers conditioned coolant directly to IT racks CDUs isolate the primary loop (which may carry contaminants, particulates, scaling agents, or chemical treatments like biocides and corrosion inhibitors - chemistry that is incompatible with IT gear) from the secondary loop. As well as preventing corrosion and fouling, this isolation offers operators the safety margin that operators need for board-level confidence in liquid. The integrity of the server cold plates is safeguarded by the CDU, which uses a heat exchanger to separate the two environments and maintain a clean, controlled fluid in the IT loop. Because CDUs are fitted with variable speed pumps, automated valves, and sensors, they can dynamically adjust the flow rate and pressure of the TCS to ensure optimal cooling even when HPC workloads change. Stability: Balancing thermal predictability with unpredictable loads HPC and AI workloads are not only high power; they are also volatile. GPU-intensive training jobs or changeable CPU workloads can cause high-frequency power swings, which - without regulation - would translate into thermal instability. The CDU mitigates this risk by controlling temperature, pressure, and flow across all racks and nodes, absorbing dynamic changes and delivering predictable thermal conditions. The CDU absorbs fluctuations by stabilising temperature, pressure, and flow across all racks and nodes, regardless of how erratic the workload is. Sensor arrays ensure the cooling loop remains in accordance with specifications, while variable speed pumps modify flow to fit demand and heat exchangers are calibrated to maintain an established approach temperature. Adaptability: Bridging facility constraints with IT requirements The thermal architecture of data centres varies widely, with some using warm-water loops that operate at temperatures between 20 and 40°C. By adjusting secondary loop conditions to align IT requirements with the facility, the CDU adjusts to these fluctuations. The CDU uses mixing or bypass control to temper supply water. It can alternate between tower-assisted cooling, free cooling, or dry cooler rejection depending on the environmental conditions, and it can adjust flow distribution amongst racks to align with real-time demand. This adaptability makes DTC deployable in a variety of infrastructures without requiring extensive facility renovations. It also makes it possible for liquid cooling to be phased in gradually - ideal for operators who need to make incremental upgrades. Efficiency: Enabling sustainable scale Beyond risk and reliability, CDUs unlock possibilities that make liquid cooling a sustainable option. By managing flow and temperature, CDUs eliminate the inefficiencies of over-pumping and over-cooling. They also maximise scope for free cooling and heat recovery integration such as connecting to district heating networks and reclaiming waste heat as a revenue stream or sustainability benefit. This allows operators to simultaneously lower PUE (Power Usage Effectiveness) to values below 1.1 while simultaneously reducing WUE (Water Usage Effectiveness) by minimising evaporative cooling. All this, while meeting the extreme thermal demands of AI and HPC workloads. CDUs as the thermal control plane Viewed holistically, CDUs are far more than pumps and pipes; they are the thermal control plane for thermal management, orchestrating safe isolation, dynamic stability, infrastructure adaptability, and operational efficiency. They translate unpredictable IT loads into manageable facility-side conditions, ensuring that single-phase DTC can be deployed at scale, enabling HPC and AI data centres to evolve into multi-hundred kilowatt racks without thermal failure. Without CDUs, direct-to-chip cooling would be risky, uncoordinated, and inefficient. With CDUs, it becomes an intelligent and resilient architecture capable of supporting 100 kW (and higher) racks as well as the escalating thermal demands of AI and HPC clusters. As workloads continue to climb and rack power densities surge, the industry’s ability to scale hinges on this intelligence. CDUs are not a supporting component; they are the enabler of single-phase DTC at scale and a cornerstone of the future data centre. For more from Subzero Engineering, click here.

ZincFive introduces battery system designed for AI DCs

ZincFive, a producer of nickel-zinc (NiZn) battery-based solutions for immediate power applications, has announced a new nickel-zinc battery cabinet designed for data centres deploying artificial intelligence workloads. The system, named BC AI, is positioned as an uninterruptible power supply (UPS) battery platform that can support both high-intensity AI power surges and conventional backup requirements. The company says the new system builds on its existing nickel-zinc battery range and is engineered for environments where GPU clusters and rapid power fluctuations are driving changes in electrical infrastructure requirements. The battery technology is intended to respond to fast transient loads associated with AI training and inference, while also providing backup during power interruptions. The system includes a battery management platform and nickel-zinc chemistry designed for frequent high-power discharge cycles. The company says this approach reduces reliance on upstream electrical capacity by managing dynamic loads at the UPS level. Nickel-zinc battery design for transient load handling As well as incorporating a new nickel-zinc battery cell designed for high-intensity usage and long service life, ZincFive highlights the product's compact footprint and field-upgradeable design. Nickel-zinc chemistry offers power density characteristics that allow the system to accommodate rapid load spikes without significant footprint expansion. ZincFive says competing approaches may require substantially more physical space to manage similar peak loads, particularly where AI applications can generate power demands above nominal UPS levels. The system is targeted at hyperscale operators, colocation facilities, and UPS manufacturers integrating AI-ready backup capacity. The company also points to potential benefits related to infrastructure design, including reduced UPS sizing requirements and support for power-management strategies aimed at improving grid interaction. Tod Higinbotham, Chief Executive Officer of ZincFive, says, “AI is transforming the very foundation of data centres, creating new challenges that legacy technologies cannot solve. "With BC 2 AI, we are delivering a safe, sustainable, and future-ready power solution designed to handle the most demanding AI workloads while continuing to support traditional IT backup. "This is a defining moment not just for ZincFive, but for the entire data centre industry as it adapts to the AI era.” For more from ZincFive, click here.

Red Hat adds support for OpenShift on NVIDIA BlueField DPUs

Red Hat, a US provider of open-source software, has announced support for running Red Hat OpenShift on NVIDIA BlueField data processing units (DPUs). The company says the development is intended to help organisations deploy AI workloads with improved security, networking, and storage performance. According to Red Hat, modern AI applications increasingly compete with core infrastructure services for system resources, which can affect performance and security. The company states that running OpenShift with BlueField aims to separate AI workloads from infrastructure functions, such as networking and security, to improve operational efficiency and reduce system contention. It says the platform will support enhanced networking, more streamlined lifecycle management, and resource offloading to the DPU. Workload isolation and resource efficiency Red Hat states that by shifting networking services and infrastructure management tasks to the DPU, CPU resources can be used for AI applications, also highlighting acceleration features for data-plane and storage-traffic processing, including support for NVMe over Fabrics and optimised Open vSwitch data paths. Additional features include distributed routing for multi-tenant environments and security controls designed to reduce attack surfaces by isolating workloads away from infrastructure services. Support for BlueField on OpenShift will be offered initially as a technical preview, with broader integration planned. Red Hat notes that ongoing work with NVIDIA aims to add further support for the NVIDIA DOCA software framework and third-party network functions. The companies also expect future capability enhancements with the next generation of BlueField hardware and integration with NVIDIA’s Spectrum-X Ethernet networking for distributed AI environments. Ryan King, Vice President, AI and Infrastructure, Partner Ecosystem Success at Red Hat, comments, “As the adoption of generative and agentic AI grows, the demand for advanced security and performance in data centres has never been higher, particularly with the proliferation of AI workloads. "Our collaboration with NVIDIA to enable Red Hat OpenShift support for NVIDIA BlueField DPUs provides customers with a more reliable, secure, and high-performance platform to address this challenge and maximise their hardware investment.” Justin Boitano, Vice President, Enterprise Products at NVIDIA, adds, “Data-intensive AI reasoning workloads demand a new era of secure and efficient infrastructure. "The Red Hat OpenShift integration of NVIDIA BlueField builds on our longstanding work to empower organisations to achieve unprecedented scale and performance across their AI infrastructure.” For more from Red Hat, click here.

Sabey achieves 25% carbon emissions cut

Sabey Data Centers, a data centre developer, owner, and operator, has announced a 25.2% reduction in Scope 1 and Scope 2 carbon emissions from a 2018 baseline, even as electrical load under management has continued growth in the same interval. The company’s 2024 Sustainability Report details progress in environmental performance, technology innovation, and clean energy partnerships intended to rival the global data centre sector. 2024 report highlights The 2024 report shares data on Sabey’s emissions reductions, energy efficiency improvements, and external partnerships. The company says it continues to align its emissions reductions with its science-based targets and is working to achieve net-zero carbon emissions across Scope 1 and Scope 2 by 2029. Key developments from the report include: · Carbon emissions slashed 25.2% from 2018 baseline· Pioneering MOU with TerraPower to explore integrating next-generation nuclear energy· Nine buildings earn ENERGY STAR certification with scores over 90; five receive a score of 99/100 Clear path to net zero The report outlines the steps Sabey is taking to meet its net-zero goal. These include continued investment in carbon-free energy, improving building operations to reduce energy use, reducing emissions from HVAC and fuel sources, and helping customers better understand their own energy footprints. Casey Mason, Senior Energy & Sustainability Manager, says, “Data centres are the backbone of the digital economy and [the] AI revolution, but must become stewards of global decarbonisation. “We are not just on track for net zero by 2029; we're reimagining how critical digital infrastructure can be both scalable and sustainable for the world’s fastest-growing industries. "Our work with TerraPower, local utilities, and SBTi showcases the kind of bold collaboration needed for a climate-secure future.” In alignment with the Greenhouse Gas Protocol, Sabey reports on emissions and sustainability efforts annually, engaging with external organisations in the process, including CDP, GRESB, EcoVadis, Atrius, and data centre tenants. The company’s emissions reporting includes both location-based and market-based accounting methods. For more from Sabey, click here.

Danfoss to showcase DC technologies at SuperComputing 2025

Danfoss, a Danish manufacturer of mobile hydraulic systems and components, plans to present its data centre cooling and power management technologies at SuperComputing 2025, taking place 18–20 November in St Louis, Missouri, USA. The company says it will demonstrate equipment designed to support reliability, energy performance, and liquid-cooling adoption in high-density computing environments. Exhibits will include cooling components, liquid-cooling hardware, and motor-control equipment intended for use across data hall and plant-room applications. Danfoss notes that increasing data centre efficiency while maintaining uptime remains a central challenge for operators and developers, particularly as AI and high-performance computing drive increases in heat output and power usage. Peter Bleday, Vice President, Specialty Business Unit and Data Center at Danfoss Power Solutions, says, “Danfoss technologies are trusted by the world’s leading cloud service providers and chip manufacturers with products installed in facilities around the world. "We look forward to welcoming visitors to our booth to discuss how we can help them achieve smarter, more reliable, and more sustainable data centre operations.” Cooling and power management focus Danfoss will present liquid-cooling components including couplings, hoses, and valve assemblies designed to support leak-tested coolant distribution for rack-level and direct-to-chip cooling. A smart valve train system providing plug-and-play connection between piping and server racks will also be shown, designed to help optimise coolant flow and simplify installation. The company's HVAC portfolio will also feature, including centrifugal compressor technology engineered for high efficiency and low noise in compact installations. Danfoss states that this equipment is designed to support data centre cooling requirements with long-term performance stability. In addition, the manufacturer will highlight its power-conversion and motor-control portfolio, including variable-frequency drives and harmonic-mitigation equipment intended to support low-PUE facilities. The business says its liquid-cooled power-conversion modules are designed to support applications such as energy storage and fuel-cell systems within data centre environments. Danfoss representatives will also discuss the company’s involvement in wider sustainability initiatives, including the Net Zero Innovation Hub for Data Centers, where industry stakeholders such as Google and Microsoft collaborate on energy-efficiency and decarbonisation strategies. For more from Danfoss, click here.

Arista unveils 800G R4 series networking portfolio

Arista Networks, a provider of cloud and AI networking systems, has introduced a new generation of R4 Series networking platforms designed for artificial intelligence, large-scale data centre environments, and routed backbone deployment. The new systems are intended to support high-performance compute clusters, low-latency operation, and large routing backbones. According to Arista, the portfolio is designed to provide high port density and support for 800-gigabit ethernet networks, with integrated security features for encrypted traffic. Seamus Crehan, President of Crehan Research, says, “The 800GbE market is growing explosively with port shipments more than tripling sequentially in Q2 '25, and Arista led in branded market share for both 800GbE as well as overall data centre ethernet switching. "These new 800GbE products from Arista are well-timed to capitalise on this segment’s projected 90% five-year average annual growth rate driven by AI, storage, and general compute workloads.” Arista states that the new systems are designed to reduce operating costs and energy consumption in AI and data centre environments, while supporting routing technologies such as EVPN, VXLAN, MPLS, and SR/SRv6. The company also highlights engineering for predictable latency and packet-handling performance. Focus on high-density 800G networking for AI Arista says the platforms are aimed at workloads including AI training, inference, data centre interconnect, and large-scale routing. The new generation supports a range of 800-gigabit configurations, with capacity options designed for large-enterprise, cloud, and service provider networks. Tim Smith, Senior Vice President of Technical Infrastructure Engineering and Operations at Magnite, comments, “When Magnite needed to build our next-generation data centre solution for AI and other advanced computing needs, Arista was the clear choice given their high quality offering. "We've deployed a dense 800G spine using the modular Arista platform with both AI-optimised and high-scale multiservice routing linecards, providing an ideal foundation for the future.” Arista notes that the systems include options for secure traffic handling with wirespeed encryption across all ports, including MACsec, IPsec, and VXLANsec. The highest-capacity chassis in the range supports hundreds of 800-gigabit ethernet ports in one system. Arista also introduces its HyperPort interface, which the company says can simplify scale-across network designs and reduce AI workload completion times compared with traditional multi-link configurations. Supporting spine and leaf deployments Arista has also expanded its fixed-form systems designed for use as either data centre spines or leaf switches. According to the company, the systems offer flexible port combinations for 800-gigabit and 100-gigabit ethernet environments. Leaf systems in the portfolio are positioned for direct server connectivity and mixed-workload data centres. These switches include copper and fibre options, uplink ports, and hardware-based encryption support. Arista says its larger modular systems and several associated linecards are available now, alongside new fixed-format switches. Additional platforms and configurations are scheduled for release in early 2026. For more from Arista, click here.

Sparkle's BlueMed submarine cable lands in Cyprus

Sparkle, the first international service provider in Italy, and Cyta, a provider of integrated electronic communications in Cyprus, have announced the arrival of the BlueMed submarine cable at Cyta’s Yeroskipos landing station in Cyprus. BlueMed is Sparkle's new cable connecting Italy with several countries bordering the Mediterranean and up to Jordan. It's part of the Blue & Raman Submarine Cable Systems - built in partnership with Google and other operators - that stretch further in the Middle East up to Mumbai, India. With four fibre pairs and an initial design capacity of more than 25 Tbps per pair, the system delivers high-speed, low-latency, and scalable connectivity across Europe, the Middle East, and Africa. A new PoP in Cyprus With the branch to Yeroskipos station, Sparkle secures a key point of presence (PoP) in Cyprus, while Cyta gains access to the BlueMed submarine cable system, enhancing connectivity between Cyprus, Greece, and other Mediterranean countries. This initiative aims to enable Cyta to better meet the growing demand for advanced internet services and digital content in the country, while strengthening Cyprus’ role as a strategic digital hub - providing direct connectivity to Greece, to Western and Central Europe via Genoa and Marseille, and to the Levant through other neighbouring eastern Mediterranean countries. Enrico Bagnasco, CEO of Sparkle, comments, “This is a new, important stage for BlueMed, a project that embodies our commitment to innovation and collaboration, linking Europe with Africa and the Middle East through state-of-the-art infrastructure. “Today, we are also particularly glad to celebrate this milestone with our long-standing partner Cyta, confirming our shared commitment to strengthening connectivity in the Mediterranean basin.” George Malikides, Chief Technology Officer at Cyta, adds, “The connection of Cyta to BlueMed will further enhance the Cyprus digital ecosystem and reinforce the island’s position as a key digital hub in the Eastern Mediterranean." George Metzakis, Chief Commercial Officer at Cyta, concurs, stating, “The arrival of BlueMed in Cyprus marks a pivotal step forward in our ongoing mission to strengthen the island’s international connectivity.” BlueMed has received funding from the European Commission under the Connecting Europe Facility (CEF) programme, highlighting its strategic relevance for improving digital resilience and connectivity across Europe and beyond.

Wolong introduces efficient motors for DC cooling applications

Wolong Electric America, a developer of industrial motor and drive technology, has introduced its Permanent Magnet Direct Drive (PMDD) motors, highlighting their ability to improve energy efficiency and reduce heat generation in high-demand environments such as data centres. Designed to operate without belts or sheaves, PMDD motors use a direct drive system that reduces mechanical complexity and common failure points, improving reliability and minimising maintenance requirements. The approach should also reduce mechanical stress and radial load on bearings, contributing to a longer service life. Lower heat output and energy use in data centres At the core of each motor is a rare earth magnet design that delivers stronger magnetic fields in a compact form factor. This aims to enable higher efficiency and cooler operation compared with traditional induction motors, which would be a key advantage in temperature-sensitive environments such as data centres, where controlling internal heat and power consumption are constant priorities. The motors are operated via a variable frequency drive (VFD), enabling precise speed control, smooth acceleration and deceleration, and reduced electrical strain on supporting systems. A 4:1 turndown ratio allows the motors to maintain torque and efficiency - including at low speeds - supporting variable airflow demands within cooling systems. Wolong Electric says its PMDD motors can be integrated directly into fan assemblies, reducing overall system losses and eliminating inrush current at startup. With reported efficiency improvements of around 20% over conventional induction motors, the design should contribute to measurable reductions in both energy use and waste heat. Flexible configurations for critical environments Wolong Electric says the PMDD motors can be tailored to specific applications, including data centre cooling systems, refineries, and OEM equipment such as heat exchangers. The motors are designed to operate at lower temperatures and with reduced maintenance demands, supporting long-term reliability and stable thermal management across facility operations. The company’s design approach hopes to enable easy integration with OEM and packaged system configurations, helping operators meet efficiency goals while aligning with evolving energy standards.

Datacentre Innovation Series 2025 announced

In an age where digital transformation is pivotal for every industry, TraiCon Events says it is proud to announce the Datacentre Innovation Series 2025, taking place on 10 December 2025 at the Park Hotel in Hyderabad, India. This premier event will focus on the future of data centres, exploring new technologies, sustainability practices, and cutting-edge innovations that are reshaping the industry. Transforming India’s data centre landscape India’s data centre industry is on the brink of a major revolution, driven by cloud adoption, data sovereignty, AI, and the growing need for digital infrastructure. With a rapidly expanding digital economy and increased reliance on cloud computing, India is investing heavily in modernising and expanding its data centre ecosystem. The focus is on scalability, resilience, sustainability, and ensuring that data protection and privacy are at the forefront. The Datacentre Innovation Series 2025 will delve into the most pressing topics such as edge computing, data centre automation, 5G infrastructure, green data centers, AI-driven operations, and cybersecurity. Who should attend? This high-level strategic forum will bring together more than 250 senior-level professionals, government officials, industry experts, technology innovators, and decision-makers to discuss the future of India’s data infrastructure and the rapidly evolving landscape of digital transformation. Attendees will include: • Data centre operators & providers• Cloud service providers (CSPs)• Telecom operators• Government & regulatory officials• Technology innovators & solution providers• Data security & privacy experts• Sustainability leaders• Data infrastructure investors• Designers & engineers of data centres• Academia & research institutions• Telecom & IT analysts Why this series and why now? India’s data centre industry is witnessing unprecedented growth, with an increasing focus on sustainability, cloud-first strategies, and the future of edge computing. As the nation drives towards digital India initiatives and addresses data sovereignty concerns, it’s essential to stay ahead of the curve with the latest technologies and best practices. With the rise of 5G, AI, and IoT, the demand for robust, scalable, and secure data centres is higher than ever before. The Datacentre Innovation Series 2025 will equip professionals with the knowledge and strategies required to tackle the evolving challenges and ensure that India’s data centre ecosystem is ready for the demands of the future.

GNM-IX launches new PoP in Bucharest

GNM-IX, a Dutch internet exchange and backbone operator, has announced the launch of a new point of presence (PoP) in Bucharest, Romania, located in NXDATA-1, one of the country’s major carrier-neutral data centres and a digital gateway to Southeast Europe. This marks GNM’s first presence in Romania, expanding the company’s distributed interconnection platform into another region of Europe. The new Bucharest site provides access to GNM’s core connectivity services - internet exchange (GNM-IX), VLAN-based interconnections, and global IP Transit - enabling Romanian operators and content networks to exchange traffic locally and optimise international routes through GNM’s multi-terabit platform. For existing GNM members, the new PoP should strengthen connectivity across Southeast Europe, creating additional redundancy and more efficient routing options towards the Balkans, Central Europe, and beyond. Alex Surkoff, Business Development Director at GNM, comments, “Our goal is to make high-performance connectivity available wherever networks grow. “Expanding to Bucharest enhances our distributed architecture and gives both local and international operators new ways to interconnect - staying local in traffic exchange while remaining part of the global internet fabric.” GNM-IX now has more than 10 Tbps of aggregated traffic and 650+ connected networks. For more from GNM, click here.