We spend a lot of time on this podcast talking about what AI does to information: deepfakes, synthetic media, manipulated reality. But there’s a physical cost to all of it that rarely gets discussed: water.

Every AI-generated image, every deepfake video, every ChatGPT query runs on servers that need millions of litres of fresh water to stay cool. And those data centers are increasingly being built in regions already facing drought and water stress.

Dr. Kevin Grecksch is Associate Professor of Water and Environmental Governance at the University of Oxford, where he directs the MSc in Water Science, Policy and Management. His research focuses on who gets water, who controls it, and what happens when there isn’t enough. In this episode, he helps us understand the hidden environmental footprint of the AI tools we discuss every week on this show, and what smarter governance could look like.

We discuss:

* The physical infrastructure behind AI-generated content, and why it needs so much water

* Why data centers are being built in drought-prone regions like Spain and the American Southwest

* What “water positive by 2030” actually means (Kevin doesn’t hold back)

* The irony of AI as climate solution while worsening water stress

* Integrative planning: circular systems, waste heat reuse, and localized solutions

* What those of us using generative AI should keep in mind

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit thedeepfakedialogues.substack.com