A new artificial intelligence facility planned for Wyoming is expected to consume more electricity than all residential households in the state combined, according to utility provider Rocky Mountain Power. It’s part of a broader national trend in which AI infrastructure is accelerating faster than energy systems can adapt, posing new challenges for utilities, governments, and consumers.
As AI model training and inferencing become increasingly resource-intensive, large-scale data centers are reshaping America’s energy and water usage. Industry estimates suggest data centers already account for 4.4% of US electricity consumption and could reach 12% by 2028. Some utilities warn that national electricity demand may rise as much as 50% within five years, with no state currently positioned to manage that level of growth.
AI infrastructure is outpacing utility capacity
Some of the new AI data centers being planned include one-gigawatt facilities, which is enough to power every home in a major US city like Philadelphia. Power consumption per AI query is also substantially higher than traditional internet use. A typical ChatGPT request uses approximately 2.9 watt-hours of electricity, compared to 0.3 watt-hours for a standard web search.
The growing energy burden is already prompting shifts in utility strategy. Pacific Gas & Electric, for instance, reversed plans to retire three coal plants. Grid operators in Texas have issued emergency alerts in anticipation of rising demand.
Meanwhile, Wyoming has emerged as a key location for new infrastructure due to its geographic stability and connectivity to the national internet backbone via Interstate 80. Companies including Microsoft and Meta are expanding in the state. Meta’s planned 900-acre hyperscale data center near Cheyenne is expected to come online within three years.
Water use raises new resource concerns
AI-related energy consumption is only part of the infrastructure equation; water usage is also a growing concern, especially in drought-prone regions. Many AI data centers use evaporative cooling systems, which require significant water resources. Estimates suggest each facility may use 2 million liters of water per day — roughly the equivalent of daily consumption for 6,500 US households.
This is particularly challenging for Western states like Nevada, Utah, and Colorado, where water is already scarce. In Nevada’s Reno area, for example, projected energy demand could require a 40% expansion of the state’s electricity sector within a decade. Colorado’s Front Range is planning to triple data center capacity, with a single new facility expected to draw 160 megawatts — surpassing even steel plants and mining operations in total electricity use.
Costs and emissions may fall to consumers
The economic and environmental impacts of this growth are significant. Data centers in the US may soon use more power than entire countries; one estimate compares their future energy footprint to Poland’s total consumption in 2023. Amazon alone could consume up to 48 terawatt-hours per year if it follows through on all planned facilities, and that roughly matches Nevada’s total electricity use last year.
States are beginning to respond. Utah recently passed Senate Bill 132 to develop alternative energy provisions for high-demand customers. Still, meeting the infrastructure requirements could cost up to $2 trillion by 2030. The resulting power generation is projected to cause $5.7 to $9.2 billion per year in public health-related costs due to increased emissions.
These costs are expected to be passed on to consumers through higher utility bills and environmental tradeoffs. As AI continues to expand across sectors, its energy footprint may soon impact every household — not just through digital services, but on monthly energy statements.
