Compal Bets on Liquid Cooling for Next-Gen AI Servers, Driving Data Center Efficiency

NEW YORK, NY – November 19, 2025

The Rise of Immersion Cooling in a Power-Hungry Era

The demand for computing power, particularly for Artificial Intelligence and High-Performance Computing, is surging. This exponential growth, however, comes with a significant energy cost, putting increasing pressure on data centers to find more efficient cooling solutions. Compal Electronics is responding with a new line of high-density, liquid-cooled AI servers unveiled at Supercomputing 2025, signaling a growing industry shift towards immersion cooling and intelligent management.

The new servers, including the SG720-2A/OG720-2A and SG223-2A-I models, utilize advanced liquid cooling technologies – including immersion cooling in the SG223-2A-I – to dissipate heat more effectively than traditional air-cooled systems. This approach isn't just about performance; it’s becoming crucial for sustainability and managing operational costs.

“The industry is hitting a thermal wall with air cooling,” said one data center engineer. “We’re reaching densities where traditional methods simply can’t keep up, and the energy bills are becoming unsustainable.”

Compal’s servers are designed to support up to eight AMD Instinct MI325X/MI355X GPUs and AMD Pensando Pollara 400 AI NICs, catering to the demanding thermal profiles of these high-powered components. Crucially, the company claims the systems achieve a Power Usage Effectiveness (PUE) as low as 1.05 and a partial PUE around 1.01, figures considered exceptionally efficient for modern data centers.

A Partnership Forged in Innovation: Compal and AMD

Compal’s push into high-density AI servers isn’t a solo effort. A long-standing collaboration with AMD has been instrumental in shaping the new server line. The servers are powered by AMD EPYC processors and AMD Instinct GPUs, offering a powerful combination of processing and graphics capabilities.

“The integration of AMD’s processors and GPUs with Compal’s server design expertise is a winning formula,” explained a technology analyst. “It allows them to create highly efficient and powerful systems tailored to the needs of AI and HPC workloads.”

Looking ahead, Compal is jointly developing next-generation servers based on AMD’s upcoming ‘Venice’ EPYC CPUs, slated for release next year. The ‘Venice’ processors, built on a 2-nanometer process, promise even greater performance and energy efficiency. The use of PCIe 6.0 in these new servers will double the bandwidth between the processor and GPU.

AMD Pensando Pollara 400 AI NICs are also a key component of the servers, accelerating network performance for AI workloads.

Intelligent Management: Beyond Cooling

While advanced cooling is a critical element of Compal’s new server line, the company is also focusing on intelligent management capabilities. In partnership with American Megatrends International (AMI), Compal is integrating AI-powered tools for remote monitoring, automated maintenance, and dynamic resource orchestration.

“Cooling is only part of the equation,” said an AMI spokesperson. “You need to be able to proactively manage the server’s performance and identify potential issues before they impact operations.”

The integrated management platform uses machine learning algorithms to analyze server performance data, predict hardware failures, and automate routine maintenance tasks. This not only reduces downtime but also optimizes resource utilization and lowers operational costs.

The ability to predict and prevent failures is becoming increasingly important as data centers become more complex and densely populated. AI-powered management tools can help IT teams stay ahead of the curve and ensure the reliability and availability of critical applications.

“The future of data center management is about automation and proactive maintenance,” said one data center manager. “We need tools that can help us optimize performance, reduce costs, and ensure the reliability of our infrastructure.”

Compal and AMI are co-hosting a session at SC25 titled “Revolutionizing AI at Scale,” showcasing their combined approach to intelligent data center management. The session will focus on optimized server architecture and AI-powered tools for proactive monitoring and automated maintenance.