Avoiding Waste Heat through AI Infrastructure Thermal Integration
National Academy of Engineering, The Bridge vol. 56 no. 1 · Jennifer Kurtz, Jibran Zuberi et al.
An article arguing that rising AI chip power and liquid or two-phase cooling raise reject-heat temperatures enough to integrate data centres into heating and water-efficient energy systems.
On this page
Overview
This article appears in The Bridge, the National Academy of Engineering (NAE) journal, volume 56 number 1, and is dated 2026; its authors are Jennifer Kurtz, Jibran Zuberi, Luke Wierl and Steve Hammond. According to the abstract in the LBNL Center of Expertise listing, rising AI chip power and liquid or two-phase cooling raise waste-heat temperatures enough to integrate data centres into heating and water-efficient energy systems.
The article supports this view with a few figures. AI chip power has risen from nearly 200 W to over 1,000 W. Liquid and two-phase cooling raise waste-heat temperatures from about 30 C to over 50 C. About 70-80% of data centre heat could be recovered.
In efficiency terms, average PUE is stated to have fallen from about 1.6 in 2014 to about 1.4 in 2023. This implies that infrastructure efficiency has improved but that waste heat is still largely unexploited. The article's argument is that avoiding waste is possible not only by cutting consumption but by treating heat as a useful resource.
As a limitation, this page rests on the abstract in the LBNL listing, not on the article itself; method, case studies and economic analysis are not covered here.
Key points
- About 70-80% of data centre heat could be recovered.
- Average PUE fell from about 1.6 in 2014 to about 1.4 in 2023.
- AI chip power rose from nearly 200 W to over 1,000 W.
- Liquid and two-phase cooling raise waste-heat temperatures from about 30 C to over 50 C.
- The article appears in The Bridge, volume 56 number 1.
- Higher waste-heat temperature makes integration with heating systems possible.
Using it in design
- In the heat-reuse calculator, take coolant return temperature as the key input.
- Show reuse potential rising for liquid and two-phase systems (above 50 C).
- Use a recoverable fraction of about 70-80% as the default.
Monthly newsletter · free
Keep your reading list current
Once a month: new library sources, changes to Turkish regulations and upcoming data centre events.
SoSo CAD
Check cooling capacity alongside your design
Review zone heat loads, available capacity and the capacity balance after the largest unit is lost in SoSo CAD.
Request a demo
Who should read it
Data centre designers planning heat recovery and liquid-cooling strategy, and energy policy specialists.
Publication details
- Authors
- Jennifer Kurtz; Jibran Zuberi; Luke Wierl; Steve Hammond
- Publisher
- National Academy of Engineering, The Bridge vol. 56 no. 1
- Year
- 2026
- Document language
- English
- Source type
- Expert / community
- Access
- Free to read
- Licence
- Copyright retained; citation only — the publisher’s terms apply to the original document
Original licence note
Journal article; cited only.
Citation
Jennifer Kurtz; Jibran Zuberi; Luke Wierl; Steve Hammond (2026). Avoiding Waste Heat through AI Infrastructure Thermal Integration. National Academy of Engineering, The Bridge vol. 56 no. 1. https://www.nae.edu/349844/avoiding-waste-heat-through-ai-infrastructure-thermal-integration-
Events on this subject
- Data Centre World Frankfurt
- SC26 – International Conference for High Performance Computing, Networking, Storage and Analysis
- iMasons Climate Accord Summit
All content and rights belong to the respective authors and publishers. SoSo CAD provides only the citation, its own summary and a link to the source, and is not responsible for the sources' content.