Can Wastewater Help Fuel AI Growth?

The continuous growth of artificial intelligence (AI) and high-performance computing (HPC) is driving a proportional demand for data centers that can manage their substantial infrastructure requirements. This, in turn, presents a significant challenge: balancing the increasing need for computing resources with crucial sustainability goals, particularly regarding water usage. Data centers require substantial amounts of water to prevent their equipment from overheating, which can impact reliability and performance.

However, water scarcity concerns are becoming increasingly prominent in different parts of the world, including in several key regions of the UK. Anglian Water, which serves Northamptonshire, Bedfordshire, and parts of Cambridgeshire, has struggled to provide enough water for large data centers in water-stressed areas. Regulations also limit how much drinking water can be used for such industrial purposes, including data centers. Cambridge Water has expressed similar concerns, stating they would support AI growth zones only if water could be supplied sustainably without impacting current customer supplies.

In response to these challenges, Anglian Water has proposed a bold solution: cooling large data centers with “treated sewage effluent” instead of drinking water. Geoff Darch, head of strategic asset planning at Anglian Water, recently told the BBC that companies looking to establish new data centers should be strategically locating them near water recycling plants to facilitate easier access to these alternative supplies.

While the hope is that this could significantly reduce the demand for drinking water for these facilities, the initial impact is expected to be minimal. Darch noted that recycled wastewater may account for only a single-digital percent of overall consumption, although even that will amount to “tens of millions of liters per day.”

The concept has garnered support; the Data Centre Alliance did not object to using treated sewage, and John Booth, who chairs its energy efficiency committee, stated he “would not see a problem” with using “pumped effluent from the very last stage of a sewage plant” for cooling. This water, of course, would be treated by the data center before entering the cooling system to reduce corrosion or damage. Elizabeth Orchard from the Institution of Civil Engineers confirmed that wastewater-powered cooling is as a “known, viable technology” and “very sensible” if a suitable source is nearby.

Beyond wastewater, other innovative cooling strategies contribute significantly to water conservation and efficiency. Many large data centers in the UK already employ closed-loop cooling systems, which do not need continual supplies beyond an initial “charge up.” Digital Realty, for example, uses DCI electrolysis for water preservation in one facility, saving 1.24 million liters monthly by extending water usage and eliminating chemicals.

Ultimately, there is no one-size-fits-all approach to cooling. Data centers utilize diverse methods like direct liquid cooling and immersion cooling for high-density AI/HPC workloads, evaporative cooling for specific inferencing deployments, and free cooling, which leverages cooler outside air or water to minimize carbon footprint. Hybrid cooling strategies that combine multiple technologies are also becoming increasingly common. Regardless, it’s clear that recycled wastewater has the potential to reduce water stress while supporting growth of AI and HPC.

Featured Image Photo Credit:https://www.pexels.com/photo/full-shot-of-robot-toy-8566474/

Categories

Recent Blogs