AI’s Hidden Environmental Cost: Massive Water Consumption

As artificial intelligence (AI) continues to “revolutionise” industries, it’s leaving an unexpected environmental footprint: water usage. Research shows that AI-powered data centres, which run the vast AI models used by businesses worldwide, are using staggering amounts of water for cooling and energy production.

According to International Business Machines Corporation (IBM), about 82% of large enterprises have either deployed AI or are experimenting with doing so, signalling a rapid AI adoption. But this comes at a cost, as every AI prompt in data centres consumes around 16 ounces of water. Experts say that a single data centre can have cooling towers that need millions of gallons of water annually to prevent the critical computing infrastructure from overheating.

This growing demand for water poses a problem. Google reported a nearly 22% increase in its water usage from 2021 to 2022, reaching over 5.56 billion gallons. Microsoft saw a 34% jump, consuming almost 1.7 billion gallons.

The water consumption of AI data centres is driven by the need to cool the massive servers that power AI models. These servers generate extreme heat, which is managed through industrial-scale cooling systems that rely heavily on water. In addition, these systems often use water for humidification to keep the equipment running smoothly.

The environmental costs of AI extend beyond water, with energy consumption and carbon emissions also raising concerns. For instance, generating the electricity needed to run AI servers consumes vast amounts of water at thermal and nuclear plants. AI’s water footprint is thus a combination of both “onsite” and “offsite” water consumption.

In a world facing growing water scarcity, balancing technological growth with responsible resource use poses a serious sustainability challenge.