Powering Progress or Peril? The Hidden Environmental Costs of AI Data Centers
The short version: A water-policy specialist's colloquium talk is a reminder that AI data centers' thirstiest costs are local, hidden, and rarely disclosed upfront.
Our Take
Credit to the UW-Milwaukee School of Freshwater Sciences for giving Emilie Washer a platform to walk through what most siting agreements never spell out in plain language: how much water a hyperscale data center actually pulls, whether it's from a stressed aquifer or a Great Lakes tributary, and what happens to that water afterward. The Great Lakes region has been sold as a "safe" place to build because water seems abundant, but abundance is not the same as unlimited, and evaporative cooling losses don't just vanish — they show up in local water tables and utility infrastructure built for a different era of demand.
What strikes us most is how consistently this analysis happens in academic colloquiums and policy papers rather than in the public hearings where these deals actually get approved. Communities are asked to sign off on tax abatements and zoning variances long before anyone runs the numbers Washer is running here. That's backwards. Water usage, like power draw, should be disclosed and modeled before a shovel goes in the ground, not reconstructed by researchers after the fact.
If you want to see how this plays out facility by facility, check our facility map for the water and power footprint of projects near you, and visit take-action if you want to push your local officials to demand real disclosure before the next deal gets signed.
This is GridWatch the USA’s original commentary. The video above is the work of UW-Milwaukee Center for Water Policy, published on YouTube — full credit to the creator.