AI’s dirty secret: it runs on water

8 Sept 2026 · 1 min

Watch on YouTube

0:46

In this YouTube monologue, Dan Novaes explains that AI’s rapid growth is not only an electricity problem but also a water problem. He describes how massive AI data centers house thousands of high‑performance chips that generate huge heat loads, and many facilities rely on water‑based cooling systems to keep the hardware at safe temperatures. This creates significant demand for freshwater, putting pressure on local water supplies and prompting community concerns in regions where water is already scarce. Novaes then outlines engineering responses: newer data centers are adopting closed‑loop cooling systems and other designs that dramatically cut freshwater consumption. He frames the core challenge as not just building smarter AI, but operating the required compute infrastructure in an environmentally responsible way, highlighting the hidden physical resource footprint behind every digital AI interaction.

Chapters

AI doesn't just use electricity; it uses water too. A lot of it happens inside the massive data centers running AI. Thousands of powerful chips generate enormous amounts of heat, and some facilities use water-based cooling systems to stop those machines from overheating. That means the AI boom isn't just increasing demand for computer chips; it's putting new pressure on local water supplies too. In some regions, communities have already raised concerns about new data centers competing for limited water. But engineers are working on alternatives. Newer facilities can use closed-loop systems, or cooling designs that dramatically reduce freshwater consumption. So, the challenge isn't simply making AI smarter, it's figuring out how to run enormous amounts of computing without putting unnecessary pressure on the environment. Because every AI response may feel completely digital, but behind the screen, there are physical machines using real-world resources to make it happen.