Connecting Data Centers to Water Crisis

Think of a data center like a giant kitchen that never turns off its ovens. The computers that train and run AI give off a lot of heat, and something has to carry that heat away. Many data centers use water for this, often through evaporative cooling, where water evaporates to remove heat, much like sweat cools your skin.

Evaporated water does not come back to the same river or aquifer right away, so that water is considered consumed. A watershed is the land area that drains into a particular river or lake, and it is the shared water supply for everyone living in it. Let's hold onto that picture: one shared pitcher, many people pouring.

Where the pitcher is already low, we call it regional scarcity, meaning the local supply is small compared with what people, farms, and nature need. A data center in a wet region may barely be noticed, but the same center in a dry, drought-prone region can compete with homes and agriculture.

Now the thermal feedback loop enters. Hot weather means cooling systems work harder and use more water, while hot, dry conditions also shrink the water available. Heat can also warm rivers, which holds less oxygen and stresses fish. Power plants that supply the electricity can use water too, adding to the demand.

The cascade looks like this: more cooling demand, less local water, higher stress on ecosystems, and then tougher choices for communities, such as higher bills or restrictions. Not every data center causes this. Designs using little water, recycled water, or cooler climates reduce the risk, though some of those options use more electricity instead.

So the connection is a chain: AI heat requires cooling, cooling can consume water, and in already scarce, warming watersheds that use can ripple into ecosystems and communities. The risk depends on place, design, and how heat and drought feed each other.