Compal Scales AI Infrastructure with Next-Generation NVIDIA Platforms

The challenge for modern cloud providers and enterprise operators is no longer just securing the latest accelerators, but rapidly integrating them into production-ready infrastructure. Compal is addressing this by bundling server engineering, direct liquid cooling, and rack-level validation for NVIDIA’s newest technologies, including the Blackwell Ultra platform and the forthcoming Rubin GPU and Vera CPU architectures.

For current deployments, the company is highlighting the OG430-2-L1, a 4U liquid-cooled server powered by eight NVIDIA Blackwell Ultra GPUs. Beyond mere hardware, Compal is positioning itself as a systems integrator capable of handling the thermal and connectivity complexities inherent in high-density computing. This strategy extends to the next generation of AI workloads, with early engineering support for NVIDIA HGX Rubin NVL8 systems, including a 44U rack-scale configuration.

Recognizing that agentic AI demands more than just GPU power, Compal is also integrating NVIDIA Vera CPU compute trays and Spectrum-X SN6600 Ethernet switches into its portfolio. By synchronizing CPU, GPU, and networking components into unified architectures, the company aims to reduce the engineering bottlenecks that typically delay the rollout of advanced AI clusters. As Alan Chang, Vice President of Compal’s Infrastructure Systems Business Group, noted, the true value of new silicon is only realized once it is transformed into scalable, deployable infrastructure.