EVERYTHING
The increasing computational demands of Artificial Intelligence (AI) systems are creating a significant environmental footprint, particularly concerning water usage. AI servers function essentially as large electric heaters; the Nvidia chips that process requests consume between 300 and 700 watts each. To manage this heat, data centers utilize evaporative cooling, where water absorbs thermal energy and evaporates into the atmosphere. According to Unian, approximately 80% of the water taken for this purpose is not returned to the local supply.
Corporate Consumption and Growth
The scale of consumption by major tech firms has accelerated dramatically. Google reported that its water usage nearly doubled between 2021 and 2024, citing increased AI load as the primary driver in its ecological reports. Microsoft’s data centers alone consumed 1.7 billion gallons in 2022. Furthermore, the training run of GPT-4 in Iowa during July 2022 required 11.5 million gallons of water for a single month.
Real-World Impacts and Future Projections
The consequences of this demand are already visible in drought-stricken regions worldwide. Specific examples illustrate the conflict between technological growth and environmental stability:
- In Chile, an ecological court halted Google’s data center near Santiago, a country that has endured continuous drought for 15 years.
- Keretaro, Mexico, is planning 32 new data centers, coinciding with 2024 being the driest year on record for the region.
- A proposed Google facility in Uruguay would consume 7.6 million liters daily—an amount equivalent to the needs of 55,000 people every day.
Researchers from the University of California, Riverside, project that by 2027, the AI industry could annually consume between 4.2 and 6.6 billion cubic meters of water. The lower estimate alone matches the entire annual water intake of four countries in Denmark, while the upper range approaches half of Great Britain’s total yearly consumption.
Ethical Gaps in Infrastructure Planning
The trend indicates that AI cooling infrastructure is being built predominantly in areas where water is inexpensive and local populations are least equipped to negotiate resource allocation. While some companies disclose their usage—Microsoft admitted that 42% of its 2023 water intake occurred in regions officially designated as having a "water deficit"—others remain opaque. For instance, Amazon, which manages the world's largest cloud infrastructure, does not publish any data regarding its water consumption.
The escalating demand for computational power necessitates urgent global dialogue on sustainable cooling methods and responsible resource deployment to mitigate severe environmental risks.