AI infrastructure consumes far more water than tech companies report
Major technology companies such as Amazon, Google and Microsoft are investing almost $1 trillion (€874 billion) in AI infrastructure, but the true environmental costs of that expansion are often obscured, according to The Wall Street Journal.
Although these companies publish sustainability reports, most only disclose the water consumption that is used directly on site for cooling. They often leave out “indirect” water consumption required by the power plants that generate their electricity. Because there are no legal obligations to report this secondary water use, the full ecological footprint remains hidden.
The discrepancy between reported and actual water consumption is staggering. Research from the Lawrence Berkeley National Laboratory showed that indirect water consumption in the United States has historically been 12 times higher than direct consumption. Meta is a rare exception and reported both figures; in 2024 indirect consumption amounted to 19 billion gallons, overshadowing direct consumption by a ratio of 20 to one.
Google’s figures showed a similar trend, with total water consumption in 2025 up by 34%. Although Google claimed to offset its energy consumption with renewable energy sources – which use minimal water – critics argue that these offsets do not solve the local depletion of water resources in areas where fossil fuel power plants are still in operation.
The geographical location of these facilities exacerbates the crisis. To secure cheap land and electricity, many data centres are built in regions with water scarcity, such as Phoenix, Arizona. A report from Ceres suggests that while data centres currently account for 3% of Phoenix’s annual water consumption, that figure could rise to more than 20% by 2031.
In addition, the sharp rise in demand for AI has led some utilities to keep outdated, water‑intensive coal‑fired power plants running longer than planned. To avoid relying on the public power grid, companies such as Microsoft, Meta and Amazon are increasingly building their own dedicated natural gas plants, which may reduce competition on the energy market but continue to put pressure on local groundwater levels.
Technological efforts are being made to limit the impact. Nvidia and Microsoft are promoting closed cooling systems that eliminate the need for continuous water replenishment and reduce energy consumption. However, modernising older installations is often too costly, leaving the sector dependent on water‑guzzling evaporative systems.
Ultimately, the lack of transparency – characterised by the use of non‑disclosure agreements and incomplete reporting – has fuelled public distrust.
That opacity has contributed to around $170 billion (€150 billion) worth of data centre projects being postponed or cancelled since 2024. Experts warn that current corporate reporting reflects only a small fraction of actual resource consumption, leaving the true environmental costs of the AI boom largely invisible.
Business AM
Subscribers 0
Fans 0
Followers 0
Followers