Satellite images show how data centers are changing America's landscape (www.businessinsider.com)

Business Insider’s investigation mapped 1,240 data centers built or approved across the U.S. by the end of 2024—about four times the 2010 total—driven largely by the AI arms race as Amazon, Meta, Microsoft and Google are expected to spend roughly $320 billion in capex this year to expand compute for large models. Using satellite imagery, public records and corporate disclosures, BI shows hubs concentrated in places like Northern Virginia, Stanton Springs (GA), Cheyenne (WY) and Goodyear (AZ), where individual campuses can draw as much power as a small city and use millions of gallons of water per day (Microsoft’s Wyoming site is estimated at ~84,725 MWh/year; Goodyear approvals permit up to 3 million gallons/day). The significance to the AI/ML community is practical and policy‑level: training and serving ever‑larger models will increasingly collide with regional energy, water and social constraints. BI finds 40% of facilities are in water‑stressed areas, utility customers in at least 41 states face higher bills or will soon, and neighbors report noise, vibration and housing/land‑use impacts. Major cloud providers have sustainability pledges—Microsoft and Amazon aim to be water‑positive by 2030, Google pledges 120% replenishment—but the report underscores a growing tradeoff between scaling AI compute and environmental, infrastructure and community costs, putting pressure on ML practitioners to prioritize efficiency, alternative cooling and smarter siting.