Press Release · Compal Electronics, Inc.
Compal Unveils Rack-Scale Liquid Cooling Solutions Powered by NVIDIA at SC26
Distributed by GrowTech Host · Originally issued via PR Newswire

CHICAGO, Nov. 18, 2026 — Compal Electronics (Compal; TWSE: 2324) is showcasing its AI infrastructure solutions powered by NVIDIA's next-generation technologies at Supercomputing 2026 (SC26), taking place November 17–19 at McCormick Place in Chicago. At Booth #5008, Compal is highlighting its rack-scale liquid-cooled solution built on NVIDIA HGX Rubin NVL8 platform and the NVIDIA Vera CPU Compute Tray, demonstrating its engineering and integration capabilities spanning individual compute modules to complete NVIDIA Vera CPU rack systems. NVIDIA Vera is the fastest CPU for agents and simulation, delivering 2x faster agentic CPU performance for longer-thinking agents, faster-learning RL systems, and data-driven AI factories. By integrating high-performance computing, power infrastructure, advanced cooling technologies, and system-level engineering, Compal delivers deployment-ready rack-scale AI infrastructure solutions designed to address the growing compute density, energy efficiency, and cooling requirements of generative and agentic AI.
As AI workloads evolve from large language model training toward complex reasoning and autonomous agentic workflows, data centers face increasing demands for compute density, energy efficiency, and system integration. To address these requirements, Compal brings together GPU-accelerated computing, CPU-intensive processing, high-bandwidth interconnects, advanced cooling technologies, and rack-scale system engineering to help customers build efficient, scalable, and deployment-ready AI infrastructure.
Compal's rack-scale solution built on HGX™ Rubin NVL8 platform integrates up to nine OG231-2-L1 direct liquid-cooled servers in a single rack, supporting a maximum of 72 NVIDIA Rubin GPUs for large-scale AI training, inference, and compute-intensive workloads. Each server features eight Rubin GPUs and NVIDIA NVLink™ 6 technology. According to NVIDIA's published specifications for the NVIDIA HGX™ Rubin NVL8 platform, each eight-GPU system delivers up to 400 petaflops of NVFP4 inference performance and 24 TB/s of aggregate NVIDIA NVLink™ 6 Switch bandwidth. Compared with NVIDIA HGX™ B200, NVIDIA projects up to 10 times higher token throughput for NVIDIA HGX™ Rubin NVL8 platform under specific agentic AI inference workloads.
To address the power and thermal challenges of high-density GPU deployments, the rack architecture integrates direct liquid cooling (DLC), rack-level power infrastructure, centralized system management, and leak protection mechanisms. This coordinated approach to compute, power, and cooling helps data center operators improve deployment flexibility and operational efficiency.
Beyond GPU acceleration, Compal also offers a complete NVIDIA Vera CPU Rack solution designed for agentic AI, reinforcement learning, and CPU-intensive workloads. At SC26, Compal is showcasing the NVIDIA Vera CPU Compute Tray, highlighting its engineering and integration capabilities from individual compute modules and system design to fully integrated rack-scale infrastructure, supporting the growing demand for high-density CPU computing resources in next-generation AI applications.
Compal's OX130-8N-L1 of NVIDIA Vera CPU rack integrates up to 32 NVIDIA Vera CPU Compute Trays, supporting a total of 256 NVIDIA Vera CPUs. Each compute tray features eight Vera CPUs and one NVIDIA BlueField-4 DPU for networking and infrastructure acceleration. According to NVIDIA's published technical specifications, the Vera CPU Rack can support more than 22,500 concurrent CPU execution environments. For specific agentic AI and simulation workloads, NVIDIA Vera CPUs deliver up to 1.8 times the performance of NVIDIA's designated x86 CPU baseline and up to four times the energy efficiency under specified comparison conditions. This high-density CPU architecture supports complex workloads involving AI agent execution, code processing, tool invocation, and data management.
"As AI advances from model training toward large-scale inference and agentic applications, customers need more than greater computing performance. They need complete, deployment-ready infrastructure solutions," said Alan Chang, Vice President of the Infrastructure Solutions Business Group (ISBG) at Compal. "Guided by our One Integrated Solution approach, we bring together compute, power, and cooling through our NVIDIA HGX™ Rubin NVL8 platform and Vera CPU rack-scale solutions to deliver integrated system architectures designed for next-generation AI. We will continue to deepen our technology collaboration with NVIDIA to advance compute density and energy efficiency, accelerating the deployment and adoption of advanced AI technologies."
Leveraging its expertise in rack-scale system engineering and integration, Compal continues to translate advanced computing technologies into deployment-ready AI infrastructure solutions supporting a broad range of applications, including generative AI, agentic AI, scientific research, and high-performance computing. This approach also reflects the spirit of SC26's theme, "HPC Unites," underscoring the importance of cross-disciplinary collaboration and technology integration in advancing the next generation of AI and HPC infrastructure.
SOURCE Compal Electronics, Inc.
About Compal
Established in 1984, Compal has grown into a leading global manufacturer of computers and smart devices, partnering with top-tier brands worldwide. Compal was recognized by CommonWealth Magazine as one of Taiwan's top 7 manufacturers and has consistently ranked among the Forbes Global 2000 companies. Compal has actively expanded into new growth areas, including cloud servers, automotive electronics, smart medical and healthcare, and advanced communication solutions. Headquartered in Taipei, Taiwan, Compal operates design and production facilities in the United States, Taiwan, China, Vietnam, Mexico, Brazil, and Poland. Learn more at https://www.compal.com