AI Data Center Liquid Cooling Design, CDUs, GPU Cooling, and Engineering Checklists

This book is a working reference for engineers who have to design, review,  build, commission, or operate liquid cooling for AI data centers. It is organized  the way a project unfolds: why liquid cooling is needed, which architecture to  choose, how the GPU and rack interface works, how to size the loops, how to  specify and control Coolant Distribution Units (CDUs), how to reject the heat,  how to protect the facility when something fails, and how to prove the system  works before the first training job runs. 

It is written for MEP engineers, commissioning professionals, facility  managers, and consultants. It assumes you already understand chilled water  plants, pumps, heat exchangers, and controls. It does not assume you have  worked on a direct-to-chip system before. Where the book gives numbers, they  are design starting points, not specifications. The binding requirements  always come from the accelerator, server, rack, and CDU manufacturers and  from your project's design basis.