The Full Story
Data centers are the physical infrastructure of the internet. They house thousands of servers—the computers that store your emails, stream your videos, and run Google searches—all packed into climate-controlled buildings. These servers generate enormous amounts of heat, and for decades, the standard solution has been simple: pump water through cooling systems to prevent the machines from overheating, then discharge the heated water back into local sources like rivers and lakes.
This system works, but it created an environmental crisis that went largely unnoticed by the public until recent years. In 2023 and 2024, news reports revealed that hyperscale data centers—the term for facilities operated by major tech companies—were consuming water at rates that alarmed local water authorities. Google's facilities in The Dalles, Oregon consumed so much water that the city began rationing for residents. Microsoft's data center projects in the Midwest raised concerns about aquifer depletion. Meta's facilities in Iowa drew scrutiny from agricultural communities that depend on the same water sources.
The crisis intensified because of timing and geography. Many data centers were built in areas with cheap electricity (often from hydroelectric or nuclear plants) but were now competing with growing populations and shrinking water supplies exacerbated by climate change. The Colorado River, which supplies water to 40 million people across seven states, is at historically low levels. As data center operators are trying to fix their water use problems, they are pursuing multiple technological and operational strategies simultaneously: recycling cooling water, deploying direct air cooling instead of water-based cooling, relocating facilities, and investing in water reclamation projects.
Why This Matters
Water is not equally distributed, and scarcity hits hardest where it is already scarce. When a data center in Arizona or Nevada consumes millions of gallons daily, that water is no longer available for agriculture, which uses 80 percent of the West's water supply. It is no longer available for residential use in cities facing restrictions. For communities downstream of data centers, water quality deteriorates when heated discharge returns to rivers, harming fish populations and ecosystems that depend on cool water temperatures.
The human cost is immediate and tangible. Farmers in the American West have watched groundwater levels drop as data centers and other industrial users pump from aquifers built over millions of years. Some wells have gone dry entirely. Residents in towns near data centers face municipal water restrictions—rationing showers and lawn watering—while massive computing facilities use water with fewer constraints. This creates a visible inequality that generates political pressure and community backlash.
The problem extends beyond the American West. Data center operators are trying to fix their water use problems globally because water stress affects facility locations worldwide. Ireland, which hosts major data centers operated by Apple, Google, and other tech companies, has experienced unprecedented water shortages. Singapore has invested heavily in data center cooling innovations specifically because the island nation must desalinate water at enormous cost and environmental impact. Taiwan, home to advanced chip manufacturing facilities that support data centers, faces water scarcity that affects both tech and agriculture.
Background and Context
To understand the scale of the problem, numbers matter. A single large data center can consume 300,000 to 750,000 gallons of water per day under typical cooling scenarios. For perspective, an average American household uses about 300 gallons daily. A major data center, in other words, uses as much water as an entire city of 1,000 to 2,500 people. And these are not isolated facilities; Google, Microsoft, and Amazon operate dozens of major data centers each, sometimes in the same region.
The growth of artificial intelligence has dramatically increased the water problem. Training large language models like GPT-4 requires vast computational power sustained over weeks, consuming more electricity and water than traditional data center operations. A single large AI training run can require millions of gallons of water for cooling. As AI deployment accelerates—with every tech company racing to build AI capabilities—water consumption is projected to grow exponentially through the rest of the decade.
Historically, data center operators had little incentive to minimize water use. Water was abundant and cheap in many locations where facilities were built. Regulations were lax compared to emissions standards or waste disposal rules. This changed only when local communities and environmental groups began measuring actual water consumption and connecting it to public water crises.
Key Facts
- Google used 5.6 billion gallons of water globally in 2023, with consumption concentrated in drought-prone regions including the Southwest and the Middle West.
- Microsoft committed in 2024 to reduce water use at its data centers by 95 percent by 2030 through advanced cooling and water recycling.
- The average large data center uses as much water daily as 1,000 to 2,500 American households.
- AI model training consumes significantly more water than traditional data center operations, with single training runs potentially requiring millions of gallons.
- Data center operators are trying to fix their water use problems using direct air cooling (air-to-air heat exchangers that require no water), recirculation systems that reuse water multiple times, and water reclamation technologies that treat and purify used water for reuse.
- Hyperscale data centers now face permit delays and community opposition in regions facing water stress, forcing companies to reconsider facility locations.
- Some facilities are experimenting with wastewater reclamation—using treated municipal wastewater instead of fresh water for cooling—a practice that is spreading across the industry.
What People Are Saying
Communities affected by data center water use have shifted from apathy to activism. City councils in Iowa, Oregon, and Arizona have passed restrictions on data center water consumption or required environmental impact assessments before granting permits. Farmer organizations have filed lawsuits against some operators, arguing that data centers should not have priority access to shared water resources during shortages.
We cannot allow the digital economy to consume resources at the expense of agriculture and basic human needs for water. This is a choice we are making as a society about what matters more—the infrastructure of the internet or the sustainability of communities that depend on water every day.
Environmental groups frame the issue as a climate justice problem: tech companies are externalizing the costs of their operations onto communities already vulnerable to climate change impacts. Some researchers argue that data center operators are trying to fix their water use problems, but doing so at a pace too slow given the acceleration of AI development.
Broader Implications
The data center water crisis represents a larger tension in the tech industry: growth without regard for environmental limits. As artificial intelligence becomes more central to technology company business models, water consumption will become a critical constraint on expansion. Companies cannot simply build data centers where electricity is cheap if water is unavailable or prohibited.
This is already reshaping corporate strategy. Google and Microsoft are investing heavily in water-efficient cooling technologies, but they are also beginning to locate new facilities in regions with abundant water—including Northern Europe, Canada, and Scandinavia. This shifts costs and environmental impacts geographically, raising questions about whether relocating the