SoftBank, ZutaCore, and Foxconn Develop AI Server Cooling Solution

SoftBank Corp., ZutaCore, and Foxconn have collaborated to design and develop a rack-integrated solution optimized for ZutaCore's two-phase Direct Liquid Cooling (DLC) technology, specifically for AI servers utilizing NVIDIA accelerated computing. This collaboration marks the world's first implementation of ZutaCore's two-phase DLC technology using NVIDIA H200 GPUs, announced in a press release.

The solution was tested at SoftBank's data center, achieving a partial Power Usage Effectiveness (pPUE) of 1.03 per rack, indicating high cooling efficiency. The rack-integrated solution includes a "Plug & Play" feature, allowing for immediate use by simply connecting the equipment. This innovative cooling system circulates an insulating refrigerant in a two-phase system, enhancing cooling efficiency while reducing power consumption.

Foxconn developed an AI server based on NVIDIA H200 GPUs, optimized with ZutaCore's technology, which passed NVIDIA's temperature test, confirming its compatibility, stability, and reliability. The rack is designed for high-density placement of AI servers, compatible with both 21-inch and 19-inch servers, and adheres to global standards set by the Open Compute Project (OCP).

The collaboration aims to enhance energy efficiency in AI data centers, addressing the growing demand for AI computing resources while reducing power consumption and carbon emissions. SoftBank, ZutaCore, and Foxconn plan to continue their efforts to commercialize and deploy this solution globally.