Compal Showcases CXL and Liquid Cooling Innovations for AI Data Centers at OCP 2025

Compal Electronics showcased new AI data center technologies at the 2025 OCP Global Summit in San Jose, announced in a press release. The company introduced solutions featuring Compute Express Link (CXL)-based memory pooling, PCIe Gen5 interconnects, and advanced liquid cooling systems designed to enhance performance and energy efficiency.

Compal highlighted its ongoing collaboration with NVIDIA, integrating compute, networking, and cooling technologies to support scalable AI infrastructure. The company demonstrated multi-terabyte memory pooling and cache-coherency capabilities using CXL and RDMA architectures, enabling high-bandwidth data exchange and dynamic resource allocation across server nodes.

Two new liquid cooling servers were unveiled at the event: the OG720-2A-L2 ORv3 7OU model featuring direct-to-chip liquid cooling for eight AMD Instinct GPUs, and the SG223-2A-I 2U immersion-cooled server. Both systems are engineered to maintain partial Power Usage Effectiveness (pPUE) below 1.1 under full load, providing energy-optimized options for high-density AI and HPC deployments.

Visitors to Compal’s booth at OCP 2025 were able to view demonstrations of these technologies and discuss data center performance and cost optimization strategies with the company’s engineering team.