The water crisis of AI: how training chatbots drains lakes and rivers

Artificial intelligence is hungry — not for data alone, but for water. While most debates about AI focus on energy consumption, a less visible but equally alarming crisis is quietly unfolding: AI’s thirst for freshwater. Training large AI models, like those behind popular chatbots and image generators, requires enormous amounts of electricity to cool the servers in sprawling data centers. The process isn’t just energy-intensive — it is water-intensive. Many data centers rely on water-cooled systems to prevent overheating, and in regions with limited freshwater resources, this can have devastating consequences.

In places like parts of the U.S., Europe, and Asia, rivers and reservoirs are being tapped to cool machines running billions of calculations. Estimates suggest that some of the largest AI training runs can use millions of liters of water per day, effectively putting digital development in direct competition with local communities and ecosystems.

The consequences are startling. Lakes are shrinking, groundwater is being drained faster than it can be replenished, and fish populations are under pressure from rising water temperatures caused by cooling runoff. Communities near these facilities face water shortages, even as tech giants tout their AI achievements as “green” or “sustainable.”

The scandal deepens when companies claim efficiency while quietly shifting the environmental costs elsewhere. Some tech firms locate their water-intensive data centers in regions already experiencing droughts, putting profits over public good. Local governments may see jobs and investment, but residents often see empty taps, restricted water usage, and ecological degradation.

Experts warn that as AI grows, so too will its thirst. “Every new generation of AI models is bigger, faster, and more resource-hungry,” says Dr. Lena Hart, a researcher in environmental technology. “If we don’t account for water use, we risk creating a global digital drought — one that affects millions of people and ecosystems worldwide.”

AI’s water problem is largely invisible to the public, buried in corporate reports and technical jargon. Unlike CO₂ emissions, there’s little regulation or accountability, leaving communities and nature to shoulder the cost of innovation.

As AI continues to expand, the question becomes stark: Are we willing to sacrifice lakes, rivers, and livelihoods for the promise of smarter machines? Without urgent oversight, the answer may be yes.