Water Footprint of AI Industry: A Critical Perspective

The water footprint of the AI industry has gained significant attention following the release of a study titled “The carbon and water footprints of data centres and what this could mean for artificial intelligence.” This research, led by Dutch academic Alex de Vries-Gao, uncovers alarming statistics regarding the sector’s water use, which now exceeds the total volume of bottled water consumed around the world. The study indicates that the extensive data centres necessary for large language models and generative AI tools are consuming immense amounts of water to cool their high-performance servers.

AI Industry’s Water Consumption Levels

The report outlines that the annual water consumption in the AI sector has escalated to levels surpassing approximately 450 billion litres of bottled water utilized by the global population annually. This increase is primarily due to the thermal requirements of sophisticated chips, such as those manufactured by Nvidia, which produce substantial heat during the training and inference processes involved in AI development. To avert hardware malfunctions, data centres commonly employ evaporative cooling systems or liquid cooling circuits, many of which source water from local municipal supplies or already stressed aquifers.

Transparency Issues in Water Management

This study emphasizes a widening “transparency gap” concerning the management of water resources by major tech firms. While corporations like Microsoft, Google, and Meta have committed to achieving “water positive” status by 2030, their actual water usage has surged in tandem with their advancements in AI technology. For instance, a single interaction with a chatbot could “consume” the equivalent of 500ml of water through evaporation at the data centre, depending on factors such as the server’s location and the local temperature conditions.

Geopolitical and Local Infrastructure Challenges

Geopolitical factors and local infrastructure challenges are increasingly becoming significant in this ongoing crisis. Numerous data centres are located in areas that are already experiencing water scarcity, which creates tension with local populations and agricultural sectors. The findings suggest that unless there is a substantial improvement in hardware efficiency or a shift to closed-loop cooling systems that do not depend on evaporation, the water footprint of the industry will continue to grow at an alarming rate.

Emerging Standards and Metrics for Sustainability

As regulatory agencies worldwide begin to formulate environmental reporting standards for the AI industry, experts assert that “carbon footprints” are no longer the only vital measure for sustainability. The “water footprint” of artificial intelligence is now an essential element in assessing the long-term sustainability of the current trajectory of AI development.