The Hidden Water Cost of AI: Is Artificial Intelligence Draining Our Resources?
What many don’t realize is that AI itself has a significant water footprint, one that could be worsening the very crisis it aims to solve.
Artificial Intelligence (AI) is often seen as a solution to global challenges, including optimizing water usage, predicting droughts, and improving agricultural efficiency. However, what many don’t realize is that AI itself has a significant water footprint one that could be worsening the very crisis it aims to solve.
How AI Consumes Water
The rapid advancement of AI, particularly large-scale models like Open AI’s GPT series, Google’s Bard, and Microsoft’s AI-powered tools, requires massive computational power. These models rely on data centres, which house thousands of high-performance servers. To prevent overheating, data canters use cooling systems that consume billions of litres of water annually. According to a 2023 study by UC Riverside, training just one AI model like GPT-3 requires over 700,000 litres of clean water enough to meet the drinking water needs of 320 people for a year.
A 2023 report from Microsoft revealed that its water usage jumped 34% in a single year due to AI expansion. Researcher Shaoli Ren, who specializes in AI’s environmental impact, warned that “each AI query has a hidden cost, and it’s often water.”
Case Studies: AI’s Impact on Water Resources
Several high-profile cases highlight the growing conflict between AI and water availability. Google’s Water Footprint in 2022, Google reported consuming 5.6 billion litres of water, with a significant share used to cool AI-powered data centres. Its facility in The Dalles, Oregon, drew over 1 billion litres, leading to public outcry as local communities faced water shortages.
Uruguay’s Water Crisis: Protests erupted in Montevideo, Uruguay, in 2023 after severe drought left residents struggling with water shortages. Meanwhile, Google’s local data centre continued operations, using large amounts of the region’s scarce water resources.
Microsoft’s Surge in Consumption: In Iowa, a Microsoft data centre consumed 11.5 million gallons (43.5 million litres) of water in 2022. This increase is directly linked to Microsoft’s partnership with OpenAI to expand AI capabilities.
These examples illustrate an ethical dilemma should AI development be prioritized over essential public and global water needs?
Can AI Be More Sustainable?
With AI adoption skyrocketing, the environmental toll is only expected to grow. Experts suggest several solutions to reduce AI’s water footprint:
Better Cooling Technologies: Companies like Meta and Google are experimenting with liquid cooling and advanced airflow management to cut down on water usage.
AI Model Efficiency: Training more efficient AI models with lower computational demands can significantly reduce energy and water consumption.
Renewable-Powered Data Centres : Some companies are moving data centres to locations where they can run on hydroelectric or solar power, reducing strain on local water supplies.
Regulatory Policies: Governments could introduce water usage limits for AI companies, ensuring responsible consumption in water-scarce regions.
The Future: Balancing AI and Water Sustainability
AI is here to stay, but its long-term success must include environmental responsibility. If companies and policymakers fail to address AI’s growing water footprint, we may solve one global crisis while worsening another. Transparency, innovation, and ethical AI development will be key to ensuring that AI doesn’t come at the cost of our planet’s most precious resource… water.
Follow for more on my Blusky account and Instagram