Artificial intelligence is about to become a water guzzler on a global scale. By 2030, AI data centers will consume as much water annually as 1.3 billion people—the entire population of Sub-Saharan Africa.

This isn’t a minor detail lost in the carbon footprint debate. AI’s environmental cost extends beyond emissions to water and land footprints. Cooling and powering AI’s sprawling data centers demand trillions of liters of water and vast tracts of land.

Today’s focus on carbon emissions is misleading. For instance, replacing coal power with bioenergy cuts carbon by 70 percent but multiplies water use 30 times and land use 100 times. “Low-carbon” does not equal “low-impact.”

AI’s energy appetite is staggering. Data centers powering AI already consume nearly as much electricity as France. By 2030, that figure will triple, surpassing the combined usage of Pakistan, Bangladesh, and Nigeria—countries with over 650 million people.

Surprisingly, the bulk of energy consumption comes not from training new AI models but from inference—the everyday running of AI to answer queries. Inference accounts for 80 to 90 percent of total AI energy use. ChatGPT alone processes billions of prompts daily, consuming hundreds of gigawatt-hours annually.

Energy use varies wildly across AI tasks. A simple chatbot query uses 200 times more energy than sorting spam emails. Generating AI images burns 1,400 times the baseline, while short AI-generated videos can consume 200,000 times more energy. Viral video trends add fuel to the fire.

The notorious Jevons Paradox haunts AI’s future. Efficiency gains don’t reduce total consumption; they boost it. Cheaper, faster AI encourages more use, expanding total energy and water demand. Efficiency improvements are swallowed by relentless growth.

The environmental burden isn’t evenly spread. Ireland’s data centers sucked up 21 percent of the country’s electricity last year, forcing a moratorium on new facilities. In Uruguay and Mexico, water-intensive data centers have worsened drought conditions and strained drinking water supplies.

Only 32 countries host AI-specific cloud infrastructure. The US and China control 90 percent of it. Meanwhile, low-resource countries bear the brunt of electronic waste and mineral extraction needed to build AI hardware, deepening global inequalities.

By 2030, AI infrastructure will generate an estimated 2.5 million tons of electronic waste annually—the equivalent of discarding nearly 250 Eiffel Towers every year. Much of this toxic e-waste ends up in countries with weak environmental protections.

This report is not an anti-AI manifesto. AI drives innovation and improves lives. But its environmental footprint demands urgent governance. Transparency on water and land use must join carbon tracking. Efficiency alone won’t save us without limits on AI usage growth.

Governments and companies must embed sustainability into AI design, deployment, and regulation. Ignoring AI’s water and land costs risks shifting environmental stress onto vulnerable regions, often those not benefiting from AI’s advantages.

AI’s future depends on balancing technological progress with planetary limits. Without strict oversight, the world’s thirst for AI could drain resources faster than anyone expected. The clock is ticking.