Delta’s Unique Capability to Provide Cooling Solutions for Next-gen AI GPUs, Servers, and Racks Highlighted at SC25

ST. LOUIS, MO, November 18, 2025 — Delta, a global leader in power management and smart green solutions, is presenting a portfolio of cutting-edge liquid and air-cooling solutions to meet the evolving thermal management demands at every level of AI infrastructure –from GPUs to  high-density servers and racks– at Supercomputing 2025 (SC25). The showcase’s main features is the new in-row 2MW Coolant Distribution Unit (CDU) compliant with Google’s Project Deschutes +/-400 VDC 5th-generation OCP-spec infrastructure, a new liquid cooling plate with up to 6,200W of cooling capacity, and newly-developed integrated heat spreader (IHS) lids for next-generation GPUs.

Franziskus Gehle, Vice President, Delta Electronics (Americas), said, “As computing density and performance continue to accelerate, so does the need for next-generation thermal management strategies and solutions. As a global leader in thermal management technologies for almost 20 years, Delta is leveraging its innovation capabilities and strong collaboration with industry leaders to drive the development of advanced cooling solutions. SC25 is the ideal platform to share insights on how our highly diversified portfolio helps customers enhance the productivity and energy efficiency of their AI systems.”

AI Chip to AI System Cooling 

Delta’s exhibit at SC25 represents its most comprehensive demonstration yet of end-to-end thermal solutions — spanning every level of AI infrastructure, from chip and board to subsystem and system.

At the chip level, Delta will demonstrate its new thermal management solutions for AI GPUs: the IHS Lids and the Micro Channel Lid (MCL). The MCL structure design increases heat exchange area by 20%, enhancing overall heat dissipation performance.
At the board level, Delta’s cold plate technology features optimized flow paths for GPU and CPU modules, delivering highly efficient liquid cooling for the latest-generation AI chips. Delta will also showcase its lineup of air-cooling solutions, including high-efficiency EVAC thermal modules and DC fans, offering powerful alternatives for air-cooled system designs.

At the subsystem (rack) level, Delta’s latest innovations — including the In-Rack Coolant Distribution Unit (CDU), modular rack manifold, DC pump, and high-efficiency HVDC fans — enable seamless integration of liquid and air cooling in both existing and new data center environments. The In-Rack CDU efficiently manages up to 250 kW of heat load, while the rack manifold’s modular design simplifies coolant routing across multiple cold plates. The 800 W and 2,000 W DC pumps, operating at 48 V, ensure steady circulation under dynamic AI workloads, and the high-efficiency 400 VDC and 800 VDC HVDC fans provide targeted air-assist cooling to manage residual heat — together delivering scalable, energy-efficient cooling for high-density AI infrastructure.

At the system (row) level, Delta will showcase the 2MW In-Row CDU that supports Google's “Project Deschutes” open architecture. The unit delivers a flow rate of 500 GPM (1,890 LPM) and features redundant pump and power feed designs to provide large-capacity coolant distribution across multiple AI racks.

Together, these innovations reflect Delta’s “Chip-to-System” approach — integrating thermal solutions across every level of AI infrastructure to enhance performance, reliability, and sustainability.

Visit Delta at SC25

Attendees are invited to visit Delta Booth #2412 at Supercomputing 2025 in St. Louis, Missouri, November 16–21 to experience live demonstrations of Delta’s AI chip cooling, in-rack liquid systems, and 2MW in-row CDU. Delta’s experts will be on site to share insights into how these innovations advance cooling efficiency, scalability, and sustainability across the AI landscape.