AI now consumes more water than the global bottled water industry: New study reveals

The environmental cost of the AI boom has come under renewed scrutiny after a study called “The carbon and water footprints of data centres and what this could mean for artificial intelligence” was published recently. Led by Dutch academician Alex de Vries-Gao, the study reveals that the industry’s water consumption now exceeds the total volume of bottled water consumed globally. Research indicates that the massive data centres required to power large language models and generative AI tools are withdrawing vast quantities of water to cool high-performance servers.

According to the report, the AI sector’s annual water withdrawal has reached a scale that surpasses the approximately 450 billion litres of bottled water used by the global population each year. This surge is attributed to the thermal demands of advanced chips, such as those produced by Nvidia, which generate intense heat during the training and inference phases of AI development. To prevent hardware failure, data centres rely on evaporated water cooling systems or liquid cooling loops, many of which draw from local municipal supplies or stressed aquifers.

The study highlights a growing "transparency gap" regarding how tech giants manage these resources. While companies such as Microsoft, Google, and Meta have pledged to become "water positive" by 2030, their actual consumption metrics have risen sharply alongside the deployment of AI. For instance, a single exchange with a chatbot can "consume" the equivalent of 500ml of water through evaporation at the data centre level, depending on the location of the server and the prevailing local temperature.

Geopolitics and local infrastructure are increasingly at the centre of this crisis. Many data centres are situated in regions already facing water shortage, leading to friction with local communities and agricultural sectors. The research suggests that without a fundamental shift in hardware efficiency or a transition to closed-loop cooling systems that do not rely on evaporation, the industry’s water footprint will continue to expand exponentially.

As regulatory bodies around the world begin to draft environmental reporting standards for AI, experts argue that "carbon footprints" are no longer the sole metric for sustainability. The "water footprint" of digital intelligence is now a critical factor in determining the long-term viability of the current AI trajectory.