Exascale, Compal to develop platform for next-gen GPU systems
The platform combines Compal’s production-grade GPU server platforms and liquid cooling systems with Exascale’s Modular Data Center (MDC) and SST power architecture to validate power, compute, cooling and controls as an integrated system.
AI infrastructure provider Exascale Labs has announced that it has entered into a Letter of Intent with Taiwan’s Compal Electronics to jointly develop a US-based platform for next-generation GPU systems. The platform will be built on a native 800-volt direct current (VDC) architecture using solid-state transformer (SST) technology to test how GPU computing systems perform in a fully integrated AI data center environment, according to a press release.
The platform combines Compal’s production-grade GPU server platforms and liquid cooling systems with Exascale’s Modular Data Center (MDC) and SST power architecture to validate power, compute, cooling and controls as an integrated system. The project is designed to bring both companies’ production-grade hardware into a shared testing environment, allowing electrical, thermal, mechanical and software interfaces to be engineered and validated together before deployment, the press release said.
Under the anticipated division of responsibilities, Compal will provide the U.S. project site, GPU servers, and liquid cooling systems. Exascale will design the modular data center system, SST-based power architecture and related software systems, and will lead the end-to-end deployment, commissioning, and testing of the platform.
“The goal is to validate native 800 VDC across a complete data center system,” said Dr. Hoansoo Lee, Chief Executive Officer of Exascale. “Compal’s planned contribution of the site, GPU servers and cooling systems will allow us to test our SST power architecture and modular infrastructure alongside the equipment they are intended to support. That integration is an important step toward commercial deployment.”
“By integrating our GPU servers and cooling systems into the development process from the outset, we can validate the complete infrastructure under real operating conditions and build a repeatable architecture that can support future high-density AI deployments at scale,” said Alan Chang, VP and GM of ISBG of Compal.
