Ginelle Greene-Dewasmes on AI’s energy demand—and its potential to cut energy use

Ginelle Greene-Dewasmes on AI’s energy demand—and its potential to cut energy use

In an interview, sustainable-development specialist Ginelle Greene-Dewasmes argues that discussion of AI’s energy needs is too focused on centralized infrastructure: greater use of AI on phones and other devices also affects energy infrastructure, and is harder to measure. She says AI’s growing electricity demand must be weighed against opportunities to reduce energy use elsewhere. Her examples include AI that adjusts building heating, ventilation and air conditioning (HVAC) to occupancy and predicts maintenance needs; she says this has reduced HVAC energy use by 25% and emissions by up to 40% in some cases. Greene-Dewasmes describes work to improve AI efficiency and apply AI in energy-intensive sectors, alongside efforts to coordinate industry around grid flexibility. The interview presents these as ways to address AI’s energy impact, not as evidence that its wider benefits already outweigh its demand. It leaves a practical measurement challenge: how to account for energy use as AI tasks shift from data centres to devices.

Publication: 621d1a22-477f-487d-86e7-5861a1738e02

attribution — uncertain: The quotes are correctly attributed to Greene-Dewasmes in a Nature interview, but the draft omits two key facts. She speaks as the World Economic Forum's AI and Energy Lead. The 25%/40% HVAC figures come from work by the Centre for the Fourth Industrial Revolution Azerbaijan, which she describes. Calling them 'her examples' without that actor could imply her own project results. Smallest repair: name her WEF role and say these were Azerbaijan Centre results she cited.

The Centre for the Fourth Industrial Revolution Azerbaijan (a public legal entity) is using AI-powered solutions to improve efficiencies

summary — uncertain: The summary keeps her central argument: the energy debate is skewed toward data centres, and device use is harder to measure. The title and question fit the source. However, it drops the Azerbaijan scope of the HVAC figures. It also uses 'grid flexibility' without explaining it; the source defines dynamic load-shifting as changing when energy is used to match grid loads, green-energy availability and pricing. Adding a short gloss and the Azerbaijan context would let lay readers follow it.

I think the conversation is too skewed towards the energy needed to power centralized infrastructure, when there are lots of other things that will contribute to AI’s energy requirements.

date — pass: originalPublished is 2026-11-11, taken from the feed entry. This matches the 11 NOV 26 dates of same-Spotlight pieces listed on the page. No earlier publication or revision date is evident.

Can green technologies deliver energy security in an unstable world? Spotlight 11 NOV 26

support — pass: Each claim excerpt is verbatim and supports its claim. The device-measurement question is grounded in her statement that this is harder to measure than data-centre outputs. 'Must be weighed against' is a mild paraphrase, but it is consistent with her aim that AI reduce energy consumption overall. The limitations correctly note there are no case details and no evidence that benefits outweigh demand.

It’s much harder to measure this than the outputs from one data centre.

media — pass: The selected media is the verified direct link to the original Nature interview, which is the most direct access to the source.