AI Water Consumption Sparks Debate as New Study Reveals Environmental Impact Estimates

AI Water Consumption Reaches Unprecedented Levels

AI water consumption has become a flashpoint in debates over the technology's environmental footprint. New research by Alex de Vries-Gao from the VU Amsterdam Institute for Environmental Studies estimates that AI systems could use between 312.5 and 764.6 billion liters of water in 20252

2

3

3

. This figure exceeds the volume of water people consume globally in bottles annually, marking a significant milestone in AI's resource usage. The study, published in the journal Patterns, represents the first comprehensive attempt to isolate AI's environmental impact from general data centers operations4

4

.

Data centers require massive amounts of water for cooling systems to prevent servers from overheating. Water is also consumed by power plants that generate electricity for these facilities, accounting for the majority of a data center's water footprint2

2

. The AI electricity and water consumption extends beyond direct operational needs to include semiconductor manufacturing and electricity generation throughout the supply chain5

5

.

Environmental Concerns Surrounding Data Centers Intensify

The AI environmental impact extends to CO2 emissions and use of water on a scale comparable to major cities. De Vries-Gao's research indicates that AI could generate between 32.6 and 79.7 million tons of carbon pollution annually, roughly equivalent to New York City's entire carbon footprint of 50 million tons2

2

4

4

. These greenhouse gas emissions now represent more than 8% of global aviation emissions4

4

.

Power demand for AI globally could reach 23GW this year, surpassing the energy consumed by Bitcoin mining in 20242

2

3

3

. Across the US, which hosts more data centers than any other country, local opposition to new projects has surged, driven largely by concerns about water and power usage2

2

. More than 230 environmental groups sent a letter to Congress warning that AI threatens Americans' water security1

1

.

Lack of Transparency From Tech Companies Fuels Debate

A critical challenge in assessing AI's true environmental costs stems from the lack of transparency from tech companies. While major tech firms publish annual sustainability reports, they typically don't break down figures to show how much AI specifically consumes2

2

. Google, OpenAI, and other companies often exclude key details like indirect water consumption from electricity demand2

2

4

4

.

This opacity has sparked competing narratives. OpenAI CEO Sam Altman claimed each ChatGPT query uses only about 0.000085 gallons of water5

5

, while Morgan Stanley projects data centers' global annual water consumption will reach 1,068 billion liters by 20285

5

. The discrepancy depends on whether calculations include only direct operational water use or account for the entire supply chain, including semiconductor manufacturing and electricity generation5

5

.

Data Center Cooling Drives Resource Demands

Data center cooling represents a major driver of AI energy usage. GPUs and AI servers generate orders of magnitude more heat than traditional data centers, making air-based cooling insufficient5

5

. Traditional evaporative cooling towers consume large volumes of water, while dry coolers use less water but require more electrical power5

5

.

MIT estimates that two liters of water might be needed for every kilowatt hour of energy5

5

. As Nvidia promises higher-performance processing units, deployment will require increasingly advanced cooling solutions. Companies are developing direct-to-chip cold plates and immersion systems where servers operate in dielectric fluids5

5

.

Conservative Estimates and Future Implications

De Vries-Gao's figures are considered conservative because they capture only operational impacts, excluding environmental costs across the supply chain and at end-of-life disposal2

2

3

3

. Professor Shaolei Ren noted the projections exceed his 2023 study that estimated 600 billion liters by 20272

2

.

The environmental costs are currently borne by society rather than tech companies. "At the moment society is paying for these costs, not the tech companies," de Vries-Gao stated4

4

. A University of Chicago survey found 4 in 10 U.S. adults are extremely worried about AI's environmental impact5

5

. Meanwhile, some advocates argue concerns are overstated, with one Substack post titled "The AI Water Issue Is Fake" gaining traction among tech commentators1

1

. Greater transparency in sustainability reports will prove essential for informed public discourse about AI's true environmental footprint.