At the end of the movie The Big Short, which focuses on the 2008 global financial crisis, Christian Bale's character, Michael Burry, correctly predicts the collapse of the US housing market and focuses all his investment into one obscure asset class - water.

The investor’s interest in H20 was well-founded. Water is in some ways our most valuable commodity, not only in its ability to stave off death by dehydration but in its crucial use across basically every major industry. This is especially true within the data center sector, where water plays a crucial albeit understated role in ensuring the facilities can run continuously without overheating. However, with the power of compute increasing by the day, the subsequent heat generated is in turn increasing, requiring even larger volumes of water to maintain operations.

AI driving demand

As data centers increase in size and capacity, the demand for water is growing to match, as evidenced in a recent report by the International Energy Agency (IEA), which revealed that a typical 100MW data center in the United States uses approximately two million liters of water daily. On a global scale, the IEA estimates the data center industry currently consumes more than 560 billion liters of water each year, with that figure potentially rising to 1,200 billion liters annually by 2030.

Water demand is also being exacerbated by an emerging trend of data centers being built in high concentrations in water-scarce areas. In a recent report by Bloomberg, it was found that close to two-thirds of new data centers constructed or in development since 2022 have been in locations with high levels of water stress. Five states highlighted in the report - California, Texas, Arizona, Illinois, and Virginia - account for 72 percent of all new data centers in locations of acute water stress.

As a result, it is becoming increasingly clear that water sustainability concerns are falling by the wayside as operators prioritize power and fiber access when deciding where to site their data centers. A common argument made is that with the advances in cooling technology, water demand will reduce exponentially as we do away with evaporative cooling methods for more advanced technologies.

It is clear that cooling technology is advancing at a rate of knots, with several technologies promising significant savings relative to current methods. In a Microsoft study published this year in the journal Nature, it was found that cold plates and two forms of immersion cooling can reduce water usage by 31 to 52 percent over their entire life cycles when compared to traditional air cooling in data centers.

However, while advances in cooling will see a marked reduction in water usage, improvements in circularity may play an even greater role, allowing for water use to become a more sustainable process. This is even more important as the size and number of data centers continue to grow, with estimations of water usage within data centers supporting LLMs and AI applications expected to consume 30 to 40 percent more water annually than the UK by 2027.

Circularity

A company at the forefront of this drive is Gradiant. Founded by MIT alumni Anurag Bajpayee and Prakash Govindan, the company recently signed two major deals with data centers in the US market to implement sustainable water solutions.

The core innovation of the company is the ability to treat any water source, from seawater to industrial waste, and purify it for use in critical infrastructure. “We are feedwater-agnostic,” says Govindan. “We can take waters that are otherwise unusable and convert them into high-quality process water for data centers - without tapping into municipal or potable supplies.”

In water-scarce areas such as Texas and the Middle East, this could prove to be a game changer in allowing data centers to take advantage of the region's vast energy wealth whilst not adversely impacting fresh water reserves.

Gradiant has also placed the use of AI centrally in its operations, deploying a platform known as SmartOps to continuously optimize water operations using neural networks. “SmartOps makes real-time decisions to reduce energy use, maximize water recovery, and extend membrane life,” explains Govindan. “It’s like having an expert engineer watching every valve, 24/7.”

For Gradiant, the issue of water should be placed on par with access to land and power when considering the growth of the market. “Power gets all the attention, but water is just as essential - and equally under threat. If we don’t design data centers with water in mind, we risk turning water-rich regions into water-stressed ones,” he points out.

The biggest challenge for companies such as Gradiant is the roll-out and adoption of technology in time to alleviate the already rising tide of water scarcity. As a result, its becoming increasingly clear that unless there is a more concerted effort from data center operators to balance their growth with an acceptance that water sustainability should play a crucial role in the planning of new builds we could reach an inflection point where data centers begin to adversely impact water access and quality the world over.

However, at present, water remains one of the last considerations for data center operators, especially those in the AI space, when it comes to deciding where to site their operations. This is well reflected in the concentration of new data centers in water scarce, but energy dense locations such as Texas and the Middle East.

Major hyperscalers such as AWS, Meta, Microsoft, and Google have begun to make progress on water sustainability measures by setting clear water neutrality goals and supporting a range of community water reclamation projects globally. This is evidenced in Microsoft's latest sustainability report, where the hyperscaler reported a net drop in water consumption and withdrawal compared to last year, citing improvements in cooling and water sustainability measures.

Despite the progress, with the growth of AI not expected to abate anytime soon, more must be done to ensure that technological advancements such as Gradiants are more widely adopted to provide a bulwark to the growing issue of water scarcity.

Investing in water is arguably more important for data centers now than it was for Burry back in 2008. If they don’t do the right thing, humanity may end up paying the price.