AI's Hidden Thirst: A Decision-Support Tool for Water Governance in Minnesota
Location
Oyate Hall
Event Website
https://2026undergraduateresearchsy.sched.com/event/2Ix8z/ais-hidden-thirst-a-decision-support-tool-for-water-governance-in-minnesota
Start Date
15-4-2026 6:00 PM
End Date
15-4-2026 8:00 PM
Description
The rapid growth of artificial intelligence (AI) data centers in Minnesota is creating strain on local water and energy resources. Current water consumption estimates are inadequate because they provide only coarse, facility-level annual averages, which fail to capture the high-frequency thermal transients inherent to modern AI workloads. For example, AI workloads include a energy-intensive computational phase and a lower-intensity communication phase, each placing distinct demands on Graphic Processing Unit (GPU) cooling systems. Current water consumption tools cannot capture these unique characteristics of AI workloads. To bridge the gap, this project will develop a first-of-its-kind, bottom-up predictive model to precisely evaluate the water footprint of GPUs. We are currently implementing a fine-grained water consumption modeling tool. This tool is designed for Minnesota’s data center operators to optimize water consumption and for state regulators, such as the Department of Natural Resources (DNR), to enforce the Water Appropriation Permit required under recent state legislation. This work will provide an essential decision support tool to empower communities and enable science-based environmental governance for a sustainable technological future.
Publication Date
2026
AI's Hidden Thirst: A Decision-Support Tool for Water Governance in Minnesota
Oyate Hall
The rapid growth of artificial intelligence (AI) data centers in Minnesota is creating strain on local water and energy resources. Current water consumption estimates are inadequate because they provide only coarse, facility-level annual averages, which fail to capture the high-frequency thermal transients inherent to modern AI workloads. For example, AI workloads include a energy-intensive computational phase and a lower-intensity communication phase, each placing distinct demands on Graphic Processing Unit (GPU) cooling systems. Current water consumption tools cannot capture these unique characteristics of AI workloads. To bridge the gap, this project will develop a first-of-its-kind, bottom-up predictive model to precisely evaluate the water footprint of GPUs. We are currently implementing a fine-grained water consumption modeling tool. This tool is designed for Minnesota’s data center operators to optimize water consumption and for state regulators, such as the Department of Natural Resources (DNR), to enforce the Water Appropriation Permit required under recent state legislation. This work will provide an essential decision support tool to empower communities and enable science-based environmental governance for a sustainable technological future.
https://digitalcommons.morris.umn.edu/urs_event/2026/posters/13