Data Centers' Hidden Cost: Water, Power and AI

Every AI query you run, every video you stream, every photo you back up passes through a data center — and data centers are thirsty, hungry beasts. A single large facility can consume as much electricity as a small city and evaporate millions of liters of water a day for cooling. As AI demand explodes, the hidden resource cost of the digital world is becoming impossible to ignore.

The numbers are sobering, and the industry knows it.

Quantifying the thirst and hunger

Start with power. Data centers already account for a meaningful share of global electricity use, and AI training and inference are accelerating demand — a single large AI training run can consume as much energy as hundreds of homes use in a year. Tech companies, once proud of efficiency gains, now watch their emissions rise as AI scales faster than optimization.

Water is the less-discussed crisis. Cooling towers evaporate vast quantities to shed heat, often in regions where water is already scarce. Communities near new data center developments have begun pushing back, asking why their aquifers should cool someone else's chatbots. The industry's water usage effectiveness metrics, once an afterthought, are now front-page news.

The efficiency race

The response is a full-spectrum efficiency drive. Hyperscalers site new facilities where cold climates and clean grids reduce both cooling needs and carbon: the Nordics, with abundant hydropower and chilly air, have become data center havens. Liquid cooling — submerging servers in engineered fluids — slashes water use. AI itself optimizes cooling systems, finding efficiencies human operators miss. And the long game is cleaner power: nuclear deals, geothermal pilots, and 24/7 carbon-free energy pledges.

The honest accounting: efficiency is improving fast, but demand is growing faster. The digital world's footprint will keep rising until clean energy scales to meet it.

• Data centers rival small cities in power use; cooling evaporates vast water.
• AI training and inference are accelerating both demands.
• Liquid cooling, cold climates, and clean power are the industry's answers.

Transparency is improving, unevenly. Some operators now publish water and carbon figures per facility; others disclose little. As regulators and communities demand more, the industry's license to build will increasingly depend on honest accounting. The facilities powering the AI age will have to justify their appetites — in public, with numbers.

The cloud was sold as weightless. Its weight is measured in megawatts and megaliters — and the AI boom is putting both on the scale.