As demand for artificial intelligence grows, a new report raises questions about the electricity and water needed to support it.
Artificial intelligence may feel like a digital service, but it depends on physical infrastructure. Data centers need electricity to run their equipment, cooling systems to manage heat, and suitable places to expand. Those needs can create difficult tradeoffs for states and the communities around them.
A September 2026 study from TRG Datacenters compared 24 states with existing AI infrastructure. It ranked the states by factors including data center capacity, estimated emissions, electricity use, and the share of mapped data centers in areas facing high water stress.
North Dakota ranked as the state facing the greatest challenges to further expansion in the study. It has one large AI data center site with 68 megawatts (MW) of capacity. The report estimates that AI data centers account for 1.7% of the state’s electricity use and produce 219,100 tonnes of CO₂ annually.
How the States Compare
| State | Large AI data center sites | Capacity | Estimated share of state electricity use | Mapped data centers in high water stress areas |
|---|---|---|---|---|
| North Dakota | 1 | 68 MW | 1.7% | 0% |
| Wyoming | 1 | 152 MW | 6.7% | 100% |
| Minnesota | 1 | 178 MW | 2.1% | 44.4% |
| Wisconsin | 2 | 369 MW | 4% | 53.8% |
| New Mexico | 2 | 87 MW | 2.1% | 100% |
| Oklahoma | 2 | 53 MW | 0.5% | 0% |
| North Carolina | 1 | 65 MW | 0.4% | 94.1% |
| Utah | 1 | 133 MW | 2.9% | 90% |
| Arizona | 3 | 356 MW | 3% | 100% |
| Alabama | 1 | 153 MW | 1.3% | 0% |
Figures are from the TRG Datacenters study supplied for this article. The water stress measure applies to mapped data centers and should not be read as the share of a state’s water supply used by data centers.
Where Expansion Could Be Difficult
Wyoming ranked second. Its one large site has 152 MW of capacity, while the study estimates that AI data centers account for 6.7% of the state’s electricity use—the highest share among the ten states shown. All of its mapped data centers are in high water stress areas.
Minnesota ranked third. The report places 44.4% of its mapped data centers in high water stress areas and estimates annual emissions of 573,400 tonnes of CO₂. Its large AI data center site has 178 MW of capacity.
Wisconsin ranked fourth, with two large sites and 369 MW of capacity. The study estimates that AI data centers account for 4% of the state’s electricity use; 53.8% of its mapped data centers are in high water stress areas.
New Mexico rounded out the top five. Its two large sites have a combined capacity of 87 MW, and all of its mapped data centers are in high water stress areas, according to the report.
Oklahoma offers a different point of comparison: its two large sites have a combined capacity of 53 MW, the lowest in the table. North Carolina’s single large site has 65 MW of capacity, while the study places 94.1% of its mapped data centers in high water stress areas.
Why This Matters for Household Budgets
A data center project can bring investment to a community, but residents also have reason to ask how its electricity and water needs will be met. When a utility plans major upgrades, the practical question for households is who will pay for them and how those costs could affect future bills.
The TRG rankings identify potential pressure points; they do not establish that residents in these states will see higher rates or water costs. Local utility plans, project agreements, and public hearings would provide a clearer picture of the impact of any proposed site.
For anyone living near a planned data center, three questions are worth bringing to a public meeting: How much electricity and water will the project require? What upgrades are needed? And who will cover their cost?
