Data centers in North America are running into pushback from local communities.
They voice concerns about rising electricity prices, pollution, and increasingly, water consumption. These issues have become more pronounced as AI is ramping up the data center footprint by an order of magnitude or more.
Enter atmospheric water harvesting (AWH) from a company known as Atoco.
The water technology startup has developed several technologies, including a system that creates an inexhaustible source of atmospheric water by capturing moisture from the surrounding air and converting it into potable water.
“The atmosphere contains seven times more fresh water than all rivers and lakes combined,” said Omar Yaghi, Founder and Chief Science Officer of Atoco, in a recent paper titled Sustainable AI in a Water-Scarce World: How to Tackle the Data Center Water Challenge with Atmospheric Water Harvesting.
“We realized that some of the world’s most pressing problems like climate change can be addressed through taking water from air to water the world. It will give independence to the people of the world to have their own water.”
For data centers, Atoco recommends using nano-engineered reticular (net-like) materials that utilize waste heat. By consuming low-grade waste heat, water can be generated. This solution becomes particularly attractive to next-generation AI factories, which will consume vast amounts of energy and so a) require plenty of water for cooling and b) produce enormous amounts of waste heat.
“Rather than relying on already stressed water sources or expensive imports, water can be generated exactly where it’s needed in a sustainable and cost-effective manner,” said Yaghi.
Traditional approaches
Water shortages and droughts worldwide have led to the development of various types of Atmospheric Water Generators (AWGs).
In essence, air is drawn into the system and processed in several ways. There are different approaches depending on climate and energy availability. This might be fog harvesting, cooling and condensation, adsorption, or desiccation to separate the moisture. The water extracted only needs a little filtration, purification, and sometimes mineralization to make it suitable for human consumption.
For example, fog harvesting captures water droplets using mesh structures. It requires no energy but depends on consistent fog, making it feasible only in certain climates. Condensation-based AWGs, on the other hand, cool air to its dew point, producing water. In dry climates, this requires significant energy, making them inefficient and prohibitively expensive to operate. Other AWG methods have similar limitations.
Harnessing waste heat to make water
For data centers, Atoco enables a new generation of AWH solutions that can be powered entirely by ambient energy, such as the temperature differential between cold air and low-grade waste heat from the data center. Even with a differential as low as 7 °C (13 °F), it is possible to power a water generation process.
Yaghi believes the system can operate at utility scale while off-grid, even in ultra-dry and arid environments.
If so, this approach could alleviate water contention in already stressed areas of the US West and Southwest. Additionally, it could appease local communities who are, in some cases, up in arms about the potential for a new AI data center to deplete water tables and local rivers and lakes.
It further solves a problem data centers have been wrestling with — how to harness the low-grade heat produced by servers.
Also read: As AI infrastructure scales, energy limits are emerging as a hard constraint on new capacity, even before projects run into land, water, and permitting fights.
