Delta Thailand Integrates Liquid Cooling With Modular AI Data Centers

Per InsiderMonkey, Delta Electronics (Thailand) unveiled a prefabricated AI modular data center solution at COMPUTEX 2026 on June 5, 2026, and InsiderMonkey reports the system is designed to cut deployment time by up to 60%. InsiderMonkey further reports the design integrates an 800VDC high-voltage power architecture with liquid-cooling components showcased at COMPUTEX, including a 3 MW GoCool liquid-to-liquid coolant distribution unit (CDU), a 2.4 MW in-row CDU, a modular 1.5 MW in-row CDU, and a cold plate configured for NVIDIA Vera Rubin NVL72 servers, and that the containerized solution supports up to 60 kW of IT load per rack. InsiderMonkey also reports Delta signed a memorandum of understanding with Daikin on April 22 to collaborate on next-generation CDU solutions for AI and HPC data centers across ASEAN-Oceania.

PR Newswire reporting on Delta's Data Centre World Asia 2025 presence lists a 20-foot AI Containerized Data Center, a liquid-to-air (L2A) CDU with up to 80 kW cooling capacity, and a modular power train scalable to 2,400 kW. PR Newswire includes a quoted statement from Jackie Chang, Delta's Head for the Southeast Asia Region: "The growth of AI and emerging digital workloads are reshaping how facilities are designed and operated," describing the showcase as an integrated set of power, cooling, and infrastructure solutions for high-density workloads.

Editorial analysis - technical context: Industry-pattern observations: Combining liquid cooling with higher-voltage DC distribution and prefabricated modules is a growing response to AI racks that exceed traditional air-cooling thresholds. Liquid CDUs and cold plates move thermal management closer to the chip or rack, which reduces thermal resistance and can improve cooling coefficient of performance relative to air systems, while HVDC architectures reduce the number of AC-DC conversion stages.

For practitioners: Prefabricated, containerized solutions that integrate power and cooling can shorten on-site commissioning and reduce integration work for cloud providers and enterprises adopting dense AI nodes. Observed patterns in similar deployments show shorter deployment timelines and tighter vendor co-design between power and thermal subsystems when OEMs ship integrated modules rather than retrofitting existing halls.

• •Adoption metrics, such as announced customer pilots or deployed racks, which would clarify real-world thermal and operational gains.
• •Technical benchmarks for PUE and rack-level cooling effectiveness from independent pilots.
• •Product interoperability statements or interoperability testing results with major GPU platforms beyond the NVIDIA Vera Rubin NVL72 configuration, and any follow-on technical details from the Delta-Daikin collaboration.

Editorial analysis: Overall, the reporting documents a continued industry trend toward liquid cooling and integrated power architectures for high-density AI deployments; observers and practitioners will want vendor-agnostic performance data and deployment case studies to validate vendor claims.