AI’s Water Problem: Can Rainwater Harvesting Make Data Centres Sustainable?

The rapid advancement of artificial intelligence (AI) has revolutionised industries, but its environmental costs are becoming harder to ignore. One of the most pressing concerns is AI’s immense water consumption, primarily for cooling the data centres that power machine learning models. A recent report by the National Engineering Policy Centre (NEPC) calls for urgent action to ensure AI’s growth does not come at the expense of sustainability.

However, amidst the concerns, there is an opportunity: could rainwater harvesting provide a sustainable water source for data centre cooling? With their vast rooftop areas, data centres are ideally positioned to collect and store rainwater, potentially reducing their reliance on mains water and easing pressure on local resources.

AI’s Unsustainable Water Footprint

Data centres, the backbone of AI operations, require significant water and energy for cooling. Major tech companies, including Google and Microsoft, have reported sharp increases in water usage, with Google consuming 19.5 million cubic metres in 2022—a 20% increase from the previous year. Microsoft saw an even greater rise, using 6.4 million cubic metres, up by 34%.

The issue is compounded by predictions that AI’s energy consumption could match that of the Netherlands by 2027. As AI adoption grows, so too does the demand for data centres, amplifying the strain on water supplies, particularly in regions where droughts and water scarcity are already a concern.

A Solution Above Our Heads: Rainwater Harvesting

One overlooked solution is rainwater harvesting. Most data centres are large, warehouse-like structures with expansive rooftops—ideal for collecting rainwater. Instead of relying on mains water for cooling, these facilities could use harvested rainwater, significantly reducing their environmental impact.

How Rainwater Harvesting Could Work for Data Centres

1. Collection – Rainwater would be captured from the vast rooftops of data centres and directed into storage tanks.

2. Filtration and Treatment – The collected rainwater would undergo filtration to remove debris and bacteria, ensuring it meets the required standards for cooling systems.

3. Integration with Cooling Systems – Once treated, rainwater could be used directly in the cooling process, reducing reliance on potable water supplies.

4. Excess Water Use – Any surplus could be used for other onsite needs, such as irrigation, fire suppression, or even recharging local aquifers.

Given that many data centres are located in areas that experience high rainfall, this approach could drastically reduce their water footprint while making them more self-sufficient.

Sustainability Recommendations from the NEPC Report

The NEPC report outlines key measures to make AI more sustainable, including:

• Mandatory Environmental Reporting – Tech firms should be required to disclose their water and energy consumption to encourage greater transparency and accountability.

• Sustainability Standards for Data Centres – Policies should mandate the reduction of potable water use for cooling and promote the adoption of alternative water sources, such as rainwater harvesting.

• Reevaluating Data Practices – Optimising how data is collected, stored, and processed can significantly reduce the strain on infrastructure and resources.

• Government Leadership and Investment – Policymakers should support research and implementation of sustainable AI solutions, ensuring the UK becomes a global leader in eco-friendly technology.

Balancing Innovation with Sustainability

While AI is set to transform economies and societies, ignoring its environmental impact could have dire consequences. Professor Tom Rodden from the University of Nottingham warns that AI’s unchecked growth risks exacerbating environmental damage. Implementing practical solutions, like rainwater harvesting, alongside broader regulatory measures, can ensure AI’s expansion does not come at the cost of sustainability.

Additionally, integrating sustainability principles into AI research and development and computer science education can foster long-term environmental responsibility within the tech sector.

Conclusion

AI’s water-intensive infrastructure is a growing challenge, but solutions exist. By leveraging natural resources available—like rainwater—data centres can reduce their dependence on potable water and mitigate their environmental impact. Combined with regulatory measures and sustainable data practices, this approach can help the UK and the world lead the way in responsible AI development.

The key question is whether tech giants will seize this opportunity or continue down an unsustainable path. The future of AI must be as innovative in its resource management as it is in its technological advancements.