AI data centers are on track to drain enough water annually by 2030 to supply 10 million American households, while pumping 44 million metric tons of carbon dioxide into the atmosphere, according to a new Cornell University study that used data analytics and artificial intelligence to map the environmental footprint of the computing boom. The research offers both a stark warning and a roadmap for making the AI infrastructure buildout more sustainable, but it hinges on decisions being made right now.

What Does the Cornell Study Actually Reveal About AI Data Center Impact?

Researchers at Cornell developed a state-by-state analysis of data center environmental impact by combining industrial data, manufacturing records, financial information, and power system data to understand where AI computing is expanding and what resources it will consume. The findings are sobering. At current growth rates, data centers will emit carbon equivalent to adding roughly 10 million cars to America's roads by 2030, making it practically impossible for the United States to meet its net-zero climate targets without intervention.

The water consumption story is equally concerning. AI data centers are projected to drain around 1,125 million cubic meters of water annually by 2030, equivalent to the yearly household water usage of 10 million Americans. This is particularly alarming in regions already facing water stress, such as northern Virginia and areas surrounding the Great Salt Lake in Utah, where hyperscalers are actively building new facilities.

"Artificial intelligence is changing every sector of society, but its rapid growth comes with a real footprint in energy, water and carbon. Our study is built to answer a simple question: Given the magnitude of the AI computing boom, what environmental trajectory will it take? And more importantly, what choices steer it toward sustainability?" the researchers explained.

Cornell University researchers, Environmental Impact Study

The researchers acknowledge that gathering this data was extraordinarily difficult. Sustainability metrics like energy and water consumption are often public, but industrial data is fragmented and incomplete. The team used AI itself to fill gaps where companies were not reporting complete information, creating a more comprehensive picture than industry self-reporting alone could provide.

How Can Communities and Policymakers Reduce the Environmental Footprint of AI Infrastructure?

• Strategic Geographic Siting: Locating data centers in regions with abundant water supplies, such as the Midwest and the so-called "windbelt" including Nebraska, Texas, and Montana, could reduce water demands by over 50 percent. This single change, when combined with improved cooling efficiency, could cut water usage by approximately 86 percent.
• Accelerated Grid Decarbonization: Faster transition to renewable energy sources is critical. The study estimates that by 2030, meeting net-zero targets would require 28 gigawatts of wind capacity or 43 gigawatts of solar capacity, even with high levels of renewable energy adoption. Without accelerated decarbonization, emissions could rise by around 20 percent as AI demand outpaces grid improvements.
• Operational Efficiency and Coordination: Data centers must be operated with greater efficiency, and hyperscalers need to coordinate closely with utilities and regulators. The researchers stress that no single intervention is a silver bullet; instead, a combination of decarbonization, efficient operations, and smart siting collectively makes the difference.

The Cornell team emphasizes that this is a critical moment. "This is the build-out moment," the researchers concluded. "The AI infrastructure choices we make this decade will decide whether AI accelerates climate progress or becomes a new environmental burden". If the recommended steps are taken, carbon emissions could drop by 73 percent and water usage by 86 percent, transforming what appears to be an environmental catastrophe into a manageable challenge.

Why Are Communities Fighting Back Against Data Center Development?

The environmental concerns outlined in the Cornell study are not abstract. On May 23, more than 600 people gathered at the Utah State Capitol to protest the Stratos Project, a proposed 40,000-acre AI data center campus in Box Elder County backed by Canadian entrepreneur Kevin O'Leary. The rally represents the largest organized community action against a single data center development in the United States this year, and it reflects growing public awareness of the real-world costs of AI infrastructure expansion.

The Stratos Project exemplifies the tensions between economic development and environmental protection. At full buildout, the campus will occupy 10,000 to 13,000 acres and will include a dedicated on-site power plant, with investors having already committed $20 million and total investment expected to exceed $100 billion. Developers argue the project will generate $30 million annually in new revenues for Box Elder County initially, rising to $108 million at full buildout, and that expanding U.S. computing capacity serves national security interests.

However, protesters raised specific and consistent objections centered on water. Box Elder County sits in one of the most water-stressed regions of an already water-stressed state, and the Great Salt Lake is at historically low levels. Critics argue that the cooling systems required for a data center of this scale will consume water that northern Utah cannot spare, and that the environmental review process did not adequately model that consumption. Chants of "you can't drink data" and "protect the water, the air, the land" reflected the rally's core argument that economic benefits do not justify environmental risks in a region facing a generational water crisis.

Shannon Barton, a Brigham City resident who formed the Box Elder Accountability Referendum (BEAR) specifically to challenge the project, is pursuing a mechanism to put the Stratos Project to a popular vote in November. A state bill proposed by Representative Doug Owens would require a comprehensive study of the environmental impacts of data centers in Utah before further approvals can proceed. The project still faces a lengthy permitting process, with state officials confirming that county approval does not bypass state-level environmental review.

Where Is Global AI Data Center Investment Heading Next?

While the United States grapples with environmental and community concerns, India is emerging as a major hub for AI data center expansion. Schneider Electric, a global leader in power management and automation, has announced that its India data center operations will become one of its most rapidly expanding segments over the next four to five years. The company stated that data centers now account for approximately 15 to 20 percent of its operations in India, but this sector is growing at a double-digit rate and is expected to capture a much larger portion in the coming years.

"This venture will lead to a significantly quicker growth rate compared to the rest of the core business," noted Sumati Sahgal, Vice-President of Secure Power and Data Centres at Schneider Electric's Greater India Zone, emphasizing that data centres and grid modernization are emerging as vital growth drivers for the company.

Sumati Sahgal, Vice-President of Secure Power and Data Centres, Schneider Electric Greater India Zone

India's data center market is projected to reach $31.36 billion by 2035, expanding at a compound annual growth rate exceeding 13 percent. The nation's current data center capacity is expected to increase significantly from approximately 1.5 gigawatts to between 6 and 7 gigawatts by 2030. India is also increasingly emerging as a key manufacturing hub for data center equipment, with rising local demand bolstering domestic production of power and cooling systems.

The global expansion of AI data center infrastructure reflects the massive computational demands of modern artificial intelligence systems. However, as the Cornell study makes clear, this expansion cannot proceed without careful attention to environmental consequences and community concerns. The decisions made in 2026 about where to build, how to power, and how to cool these facilities will reverberate for decades, determining whether AI becomes a tool for climate progress or a new source of environmental burden.