High-Performance Computing Data Center Cooling System Energy Efficiency (ESIF)

Overview

The source is a web page from the National Laboratory of the Rockies (formerly NREL) describing the cooling system of the ESIF high-performance computing (HPC) data center; it was last updated on 17 February 2026. The page is a schematic walk-through that explains each step of the system in order. The first principle of the system is that most of the heat is captured by liquid.

Operators require 90% or more of IT heat to be captured by liquid cooling, and fan walls handle the residual air-cooled load. Cooling distribution units (CDUs) supply coolant at the temperature, pressure and chemistry the IT needs. This shows that the liquid capture fraction can be a design target in liquid-cooled halls. A hierarchy is defined for heat rejection. Heat is first reused as building process hot water, then sent to a thermosyphon dry cooler and finally to cooling towers.

The system also includes a closed energy-recovery water loop. The ordering follows the logic of using the most efficient available option first, to reduce energy and water consumption. The source is a presentation of a single facility, which means it is the practice of one research facility rather than a general standard or calculation method. The 90% capture rate should therefore be read as this facility's operating requirement and not as a universal rule. The page gives no detailed capacity, flow or temperature values, so this record contains none either.

Key points

1. Operators require 90% or more of IT heat to be captured by liquid cooling.
2. Cooling distribution units supply coolant at the temperature, pressure and chemistry the IT needs.
3. Fan walls handle the residual air-cooled load.
4. A closed energy-recovery water loop is used.
5. Heat rejection hierarchy: reuse as building process hot water, then thermosyphon dry cooler, then cooling towers.
6. The page was last updated on 17 February 2026.
7. The source describes one facility's practice; it is not a general standard.

Using it in design

• Model heat rejection as a priority chain (reuse, dry cooler, evaporative tower) with hourly switching.
• Allow a 'minimum liquid capture fraction' design parameter for liquid-cooled halls.
• Design the layout so the residual air-cooled load is met by fan walls or a similar solution.
• Keep temperature, pressure and chemistry requirements as catalog fields when selecting CDUs.
• Assess heat reuse by matching it to the building's process hot water demand.

Who should read it

Mechanical engineers designing cooling and heat rejection for liquid-cooled, high-density HPC data centers.

Publication details

Authors

National Laboratory of the Rockies

Publisher

National Laboratory of the Rockies (formerly NREL)

Year

2026

Document language

English

Source type

Government / public

Access

Free to read

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation