New

Viewpoint − Water efficiency gains won’t resolve datacentres’ contested future

Sameer H. Shah
School of Environmental & Forest Sciences, University of Washington, Seattle, United States of America; shs89@uw.edu

Nicole J. Wilson
Department of Environment and Geography, and the Centre for Earth Observation Science, University of Manitoba, Winnipeg, Canada; nicole.j.wilson@umanitoba.ca

Cassandra L. Workman
Department of Anthropology, University of North Carolina Greensboro, Greensboro, United States of America; clworkma@uncg.edu

ABSTRACT: Datacentres are increasingly being built to service artificial intelligence (AI) – and large volumes of water are often required to power and cool these facilities. Today, public opposition against new datacentre development is sharply increasing in the United States, with local water insecurity concerns – including institutional distrust, ecosystem risks, and escalating utility rates – featuring prominently. The scientific community participates in these debates by quantifying datacentres’ volumetric water use – what we call "assessment-driven" research. Our viewpoint argues that such assessments can inadvertently produce an information ecosystem in which water overuse becomes framed as a resource management problem to be solved with more efficient water-use technology. To us, this will not be sufficient to overcome opposition – and could, paradoxically, inflame existing tensions. This is because the "water problem" is an existential one for local communities, not a techno-managerial one. Water access shapes the futures of local communities, their perceived character, and the lives people want for themselves – and water use, including more efficient use, by these industrial facilities can enable the very kinds of existential changes that are often the bases of local resistance movements. Assessing who (if anyone) benefits from the water used for AI can challenge existing managerial approaches that often separate co-constitutive water-society relationships and assume efficiency improvements will, at least partly, quell public opposition.

KEYWORDS: Datacentres, artificial intelligence, resistance, water security, United States of America