Key Takeaways

  • Utah’s Stratos Project faces rising public concern about water use, heat load, and electricity demand.
  • Independent researchers say widely circulated claims about AI water consumption are based on outdated or misinterpreted data.
  • Evidence from national studies suggests data centers rarely drive residential energy rate increases.

Public concern around the proposed Stratos Project in Box Elder County has escalated quickly, even as the technical realities behind data center infrastructure remain far more nuanced than much of the online conversation suggests. The project, backed by an international investor and designed for a potential 9 gigawatts of capacity, has become a focal point for larger questions about AI, energy consumption, and rural land use. Local residents, who first heard about the project on social media, wonder why a data center should reshape the environment of an entire valley. This reflects a fair question that many communities are asking.

Fear of large-scale data centers is rising faster than public understanding of how they work. Recent national data from the 2026 Pew Research Center survey, as analyzed by ConstructConnect, shows only 25% of U.S. adults have heard a lot about data centers, and another 25% have heard nothing at all. Yet 39% believe they are mostly harmful for the environment, and 38% fear negative impacts on home energy costs. This knowledge gap often drives local opposition before technical details are even discussed.

Water use remains a primary flashpoint. A widely shared 2024 Washington Post calculation previously claimed that a single AI email prompt required a standard bottle of water. That estimate has been debunked as inaccurate. Independent researchers who dug into the methodology and connected with the original analysts at the University of California, Riverside, clarified that the estimate reflected older systems and included large amounts of indirect water use from hydropower evaporation. More recent assessments from specialists like EcoLogits place the real per-prompt cost at roughly 1 to 10 milliliters. The amount used directly inside a data center itself can be closer to 0.2 to 2 milliliters.

This kind of swing highlights how quickly computing efficiency evolves. Industry researchers studying data center energy patterns note that a three- or four-year gap is essentially an entire hardware generation. Cooling design, hardware utilization, and power sourcing shift constantly. Even within a single operator, differences can be stark. Meta’s Eagle Mountain site reported a vastly different water consumption profile than its Prineville, Oregon, site in 2024, despite being roughly the same size. Climate and cooling methods play large roles in these discrepancies.

Researchers caution that reducing onsite water use is not automatically beneficial. Air cooling typically consumes less direct water but increases electricity consumption, meaning more indirect water use at the power plant. Industry frameworks like ISO 50001 for energy management and ASHRAE thermal guidelines address this by standardizing environmental conditions. Optimizing across electricity, cooling, and water requires a complex balancing act. Because outside observers rarely see these operational tradeoffs, public perception often diverges sharply from engineering reality.

A number of industry analysts have tried to bring context to the debate. Reports from organizations such as Deloitte and McKinsey often describe data centers as critical economic infrastructure that tends to generate jobs and tax revenue while accounting for a small percentage of total water consumption. For perspective, data centers represent about 0.14% of daily U.S. water use. A single 100-acre walnut farm can consume roughly twice the water of Meta’s 100-acre Eagle Mountain site.

Thermal load is another point of confusion. Preliminary analysis from a physicist at Utah State University estimated that a 9-gigawatt natural gas plant dedicated to the Stratos Project could create a localized heat footprint in the valley. Those findings spread widely. However, developers have since signaled an interest in multiple power sources instead of gas alone, meaning the scenario will likely evolve. The idea that the entire state could warm by 5 degrees, a claim circulating on social platforms, does not reflect documented research. Still, communities understandably want thorough ecological review before construction begins.

Electricity pricing remains a politically sensitive issue. Many residents assume a facility drawing gigawatts of power will raise household rates, yet empirical data does not support that conclusion. Research from the Lawrence Berkeley National Laboratory shows rate impacts typically track natural gas prices, weather, and state policy rather than data center expansion. Additional analysis from the University of Southern California found rate effects in the range of 0.007% to 0.08%. Utah’s own 2025 legislation goes further by requiring that any power built for a data center cannot raise resident electricity costs, settling the policy question at the state level.

Public anxiety regarding data center development extends far beyond Utah. Delays and cancellations tied to community objections have impacted nearly $120 billion in data center investments across the United States and Europe, according to a 2023 STL Partners study cited through ConstructConnect. Much of the concern focuses on rural land loss (22%), water use (22%), and grid strain (21%). At the same time, data centers contributed an estimated $926.9 billion to U.S. GDP, generated 5.5 million jobs, and provided $204.4 billion in tax revenue in 2024. Hyperscale operators like Equinix, Digital Realty, and QTS are navigating similar dynamics across their respective markets.

The Stratos Project serves as a proxy for a broader conversation about AI, public trust, and infrastructure. While some residents worry the facility symbolizes a shift toward an AI-driven economy that misaligns with local values, others view it as an opportunity for long-term economic growth and tax revenue. Bridging this divide requires clear communication, transparent design decisions, and public access to the technical facts that shape how modern data centers actually function.