Data Centers Heat Reuse 101

Overview

The text begins from the observation that almost all the energy servers use turns into heat, and defines heat reuse as the beneficial use of any heat recovered from the data centre. In theory all of the energy used in a facility can be captured; the excess heat comes not only from IT equipment but also from power facility losses and the thermodynamic processes of cooling machines. Reuse has existed for over 10 years and a project map was assembled (a November 2022 snapshot), yet the early large examples have not led to many new projects, even in energy-conscious regions such as the Nordics.

The reasons for limited adoption are listed: local conditions are hard to replicate and piping to a heat host is costly, some climates need heat for comfort for long periods, legislation is not always supportive, there are no generalised subsidies, data centre and heat host find it hard to collaborate, and the investment in pipes, pumping stations, heat pumps and controls is significant. What has changed is government decarbonisation commitments and subsidies, ambitious corporate goals, sharply higher fossil fuel prices and wider direct liquid cooling, which lifts return temperatures from 27-28 C to 45-65 C and so raises heat quality.

A project needs a nearby heat host (or a utility as intermediary), a temperature level that interests the host, ideally the same ownership structure or willingness to collaborate, and supportive legislation. On cost, heat extraction is shared between the data centre and the host, while heat transformation (storage, temperature increase, heat-to-cooling) is paid by the host. The data centre almost always keeps a redundant cooling system in case the host is unavailable, and a project rarely lowers its investment but lowers its operating cost.

Heat hosts fall into two groups. The ideal ones have high, steady 24/7 demand: water pre-heating, industrial laundries, bio-ethanol plants, food and beverage, breweries, large hospital campuses, distillation (desalination and wastewater treatment), chemical/textile/metal industries and pharmaceuticals. Seasonal ones are district heating networks, biomass drying, greenhouses and fish farming. Avoided CO2 is credited to the heat host, which may compensate the data centre by agreement. Techno-economic analysis weighs heat quantity and temperature, distance, location and CO2 and water benefits on a life-cycle basis.

Key points

1. Almost all server energy becomes heat; in theory all energy used in the facility could be captured.
2. Typical equipment return-temperature bands: air cooling about 15-28 C, rear-door HX about 28-40 C, cold plate and immersion about 40-65 C; direct liquid cooling enables 45-65 C return versus 27-28 C.
3. Reasons for limited adoption: hard-to-replicate local conditions and piping cost, heating-season climates, unsupportive legislation, no general subsidies, collaboration difficulty and high investment.
4. Viability needs a nearby heat host (or utility intermediary), a useful temperature, aligned ownership or willingness to collaborate, and supportive legislation.
5. Containerised/modular data centres can exceed 0.5 MW, a significant heat quantity for many local applications.
6. Cost split: heat extraction shared; heat transformation (storage, temperature boosting, heat-to-cooling) paid by the host; the data centre keeps redundant heat rejection.
7. Ideal heat hosts have constant 24/7 demand; district heating, biomass drying, greenhouses and fish farming are seasonal.
8. A data centre is a reliable heat source with uptime over 99.8% per year, at most about 20 h of downtime in the worst case.
9. Techno-economic inputs: heat quantity and temperature (for example 1 MW at 40 C versus 100 MW at 60 C), distance to an existing heat network, location type, and CO2 and water benefits.

Using it in design

• A heat-reuse module should report recoverable heat (kW) and its temperature band by cooling technology (air, RDHx, DLC/immersion) and suggest compatible heat-host categories.
• Keep full-capacity heat rejection independent of any heat host, and warn if the heat-reuse loop is counted as N capacity.
• Add distance, temperature, quantity, CO2 and water savings as input fields in the heat-reuse calculator.
• Report costs split into heat extraction (shared) and heat transformation (host).
• For liquid-cooled designs compare the 45-65 C return temperature with the host's temperature need to decide between direct use and temperature boosting.
• When reporting carbon and water gains, separate emissions avoided by the heat host from the data centre's own savings.

Who should read it

Data centre owners, sustainability managers, energy service providers and prospective heat hosts evaluating heat reuse.

Document contents

1. Introduction
2. Why heat reuse and why now
3. What makes data center heat reuse possible and sustainable
4. Is it beneficial for a heat host to receive the heat from a data center?
5. Is it beneficial for a data center to provide the heat to a heat host?
6. Candidates for a heat host
7. Who gets the CO2 emissions reduction?
8. What is the impact on water usage?
9. How to collaborate for such a project?
10. Techno-Economic Analysis
11. What are the next steps?
12. Conclusion
13. References
14. License

Related standards

• EN 50600-4-6 (Energy Reuse Factor)
• ISO/IEC 30134-6 (ERF)
• EU Energy Efficiency Directive 2023/1791

Read the document

Publication details

Authors

Pecchioli, C. (Alfa Laval); Comella, J. (Cloud&Heat); Sickinger, D.; VanGeet, O. (NREL)

Publisher

Open Compute Project Foundation

Year

2023

Document language

English

Source type

Industry body

Access

Open licence — downloadable

Licence

CC BY-SA 4.0 — share with attribution under the same licence

Licence evidence

www.opencompute.org

Citation