There has been a lot of talk lately about the implications of surging AI usage. While the focus has been on electric power consumption, the water intensity of data centers should also emerge as a major challenge going forward.
Indeed, AI is catalyzing a shift towards power-hungry “monster chips” that are becoming increasingly difficult to maintain at normal working temperatures (<80°C), with the risk that elevated temperatures make servers slow down or malfunction. Efficient cooling in data centers has then become critical, and data center operators are increasingly embracing various water and liquid cooling techniques that are less energy-intensive than air cooling (see our previous reports on data center cooling).
It is estimated that in the US, a mid-sized data center uses about 300,000 gallons or 1.3 million liters of water every day, equivalent to the consumption of 100,000 homes. According to water consultancy Bluefield Research, total water consumption by global data centers is on track to reach 450 million gallons per day by 2030, compared to 292 million gallons in 2022.
Even if the water consumption CAGR is not that impressive (+5.5% over 2022-30) thanks to efficiencies, the addition of a new demand driver to an already stressed demand/supply environment is likely to exacerbate the looming water crisis, specifically in areas where water resources are scarce.
To keep using water to cool data centers and reduce their environment footprint, cloud giants are then gradually switching from freshwater to wastewater, with notably Amazon and Google now using non-potable water in roughly 20% of their data center locations.
This move to recycled water can actually turn data centers into small water utilities as wastewater, already treated by the local utility, is further purified by data centers to make it reusable. Data centers can also switch to the local potable water supply in case of emergency and have on-site water storage and permit to treat their own wastewater.
A major opportunity is then opening up to water technology companies such as Xylem whose extensive suite of technologies can accommodate the diverse stages of a water treatment and purification process. Such purification systems may rely a single technology, or a combination of multiple technologies to achieve optimal results, including : carbon filtration (to remove organic material), reverse osmosis (in which water is forced through a semi-permeable membrane in the opposite direction of the natural flow with enough force to exceed the osmotic pressure, rejecting dissolved solids), ion exchange (resins remove ionic impurities from water and are regenerated through a reversible chemical process), or ultraviolet light (contaminants such as bacteria and viruses are exposed to UV light, damaging their DNA and rendering them inactive).
Another opportunity is emerging for water technology providers: water-as-a-service, in which the provider takes responsibility for operation and maintenance of the on-site water treatment system, and that could be well suited to data center on-site recycling initiatives should cloud giants decide to move on closed loop systems for their water.
Interestingly, water-as-a-service entails a business model shift from traditional equipment sales to highly recurring and more profitable usage-based revenue.
