AI may point to a boundless digital future, but the data centers that power it rely on a limited resource: water.
For investors eager to capitalize on AI, that gap could be critical. Yet, when it comes to water, they are working with incomplete information. Data center companies have been slow to provide a complete view of their water use, leaving investors without the information they need to fully understand financial risks below AI’s surface.
With disclosure lagging far behind the industry’s rapid growth, more investors can call for greater transparency from these companies on water risks linked to their operations, including through shareholder resolutions. A recent proposal filed by NorthStar Asset Management with Digital Realty Trust, which goes to vote on May 29, is an example.
There has been a lot of attention given to the large amounts of water used on-site for cooling in data centers. But water risks extend well beyond their walls. As Ceres’s research spotlights, water used to generate electricity for these centers can be an even larger, and often unacknowledged, risk.
Water risk generates opposition
Clusters of data centers in dry regions such as Phoenix, Arizona can significantly strain already stressed water supplies. There, the combined water demands of existing and planned facilities for use on-site and to make electricity could increase water stress — that is, when water demand exceeds supply — by as much as 32%. Another concern is whether current water utilities can supply enough water during peak demand on the hottest days.
These potential impacts are fueling concern throughout the US and beyond.
In 2025, water was a top driver of local opposition to data centers in the United States, derailing some projects and putting companies’ reputations at risk. In the first three months of 2026, at least 20 proposed data center projects, accounting for more than $41.7 billion in investment, were canceled due to local pushback.
Growing concerns about the water that data centers discharge after it has been used may further complicate the risk picture. Wastewater from data center cooling and water-treatment processes can be laced with chemical additives, heavy metals and other pollutants that even municipal treatment plants may be ill-equipped to fully remove. This has the potential to harm local waterways. The elevated temperature of data center wastewater can also cause ecological harm.
Given these escalating risks, robust disclosure that accounts for local water conditions is a critical step in giving investors the decision-useful insight they need to assess the data center water risks in their portfolios — and ensuring that communities and policymakers know what’s happening in their backyards.
Investors can help move the needle, deepening their understanding of the complex water risks that span data centers and their supply chains and how to address these risks in decision-making and strategy. Shareholder resolutions are one tool, but investors can also engage directly with companies for better visibility into the challenges they face and solutions being put to work. Ceres’s Valuing Water Finance Initiative lays out a framework to help investors support companies in addressing all dimensions of water risk including water use, pollution, and impacts to ecosystems and communities.
Raising the bar on disclosure
There are encouraging signs the industry is recognizing the full scope of water risks linked to AI and data centers, with some companies providing more comprehensive information about their water use.
Among these is CyrusOne, a global data center developer and operator. The company publicly reports how much water is used from beginning to end inside of its facilities — including water taken in, water consumed or lost through evaporation and water returned. Critically, this reporting includes the 30 sites it runs that are in areas that face high or extremely high water stress.
The level of detail the company is providing on water usage also stands out. As part of its reporting, the company discloses water usage effectiveness, a key industry standard that measures how much water a facility uses per unit of energy used by its servers. It does this for both the expected efficiency under ideal conditions and actual performance. The gap between these two numbers can help stakeholders understand if the company is meeting its efficiency commitments. Most notably, the company reports that the water used to generate the power that its facilities run on has decreased 67% since 2018, driven by increasing use of renewable energy.
Google also breaks down its overall water use. The company specifies its withdrawal, consumption and discharge by facility type — distinguishing data centers from offices and other facilities — and reports data center water use by individual location. It also details the types of water sources it draws from, differentiating between potable, non-potable and reclaimed wastewater. It discloses the percentage of freshwater its data centers withdraw from vulnerable water sources, offering valuable insight for investors concerned with data centers operating in water-stressed regions.
Meta also provides some useful information, disclosing water withdrawal from individual facilities. It also reports water withdrawal by source in areas of high water stress, as well as aggregate information about water consumption and discharge across all its facilities. The company extends its disclosure beyond direct operations, reporting the total amount of water consumed in the electricity it purchases and contracted renewable energy across all of its facilities.
The road ahead
These practices help give a baseline of where companies can reasonably be expected to provide more comprehensive water risk disclosure. But progress can happen faster if companies across the industry go beyond reporting aggregated data — providing more facility-level information as data centers proliferate in areas of high water stress — while expanding disclosures to capture indirect water risk tied to electricity generation for their facilities. Policymakers are helping speed up progress, introducing and passing bills requiring greater transparency so communities and stakeholders can have a voice in identifying and addressing potential water impacts.
Investors also play a key role advancing momentum, engaging with data center companies to encourage uptake of leading practices for disclosure and using other tactics to call for more thorough reporting of water risks that could impact long-term business value.
AI holds great promise, but without strong corporate disclosure, investors are left guessing at the true costs to water supplies that companies, communities and stakeholders depend on. Working with companies to bridge the information gap is essential to protecting long-term business and shareholder value and ensuring the AI buildout proceeds responsibly.
Kirsten James is senior program director for water at Ceres, a nonprofit advocacy organization.
Guest posts on ImpactAlpha represent the opinions of their authors and do not necessarily reflect the views of ImpactAlpha.
