Rethinking Water Usage in AI Data Centers: The Overlooked Issue

As the demand for power in AI data centers grows, the conversation around sustainability is increasingly focused on an often-overlooked resource: water. The question isn't just how to generate the necessary energy to power these facilities; it's also about whether the water required for that energy will be available. This underappreciated risk has raised alarms among experts.

A recent analysis by Ceres, alongside insights from Bluefield Research, reveals a striking reality: much of a data center's water footprint is linked to electricity consumption rather than direct cooling needs. Interestingly, there seems to be a substantial gap in how indirect water usage is reported and recognized within the industry.

The Hidden Water Costs of Power Generation

Power plants—whether coal, natural gas, or nuclear—utilize vast amounts of water not only to produce energy but also for cooling purposes. Typically, these plants heat water to create steam that drives turbines. Once the steam passes through the turbines, it cools back into water, which means that data centers sourcing electricity from such facilities are indirectly responsible for significant water consumption, despite their own low water usage.

Research from Lawrence Berkeley National Laboratory, cited by Janus Henderson, highlights that as much as 75% of a data center's total water footprint can be attributed to this indirect consumption. Meta, for instance, disclosed its indirect water usage for 2024, revealing it to be over twenty times the direct water consumption of its data centers—a staggering figure that lacks context in typical electricity bills.

When lenders finance data centers—often covering 70 to 90 percent of the borrowing for new builds—they seldom consider the water implications linked to energy production. Their focus typically rests on factors like cost, reliability, and availability, with water consumption at generation plants frequently left out of the equation. Consequently, unaccounted risks may manifest later as unexpected costs or operational delays related to water shortages.

The Financial Perspective on Water Risks

The rationale behind letting water utilities absorb the risk is compelling from a financial perspective. Credit rating agencies already analyze regulated utilities, which bear responsibility for their own water planning and risk management. However, the absence of a standardized method for measuring a data center's indirect water footprint complicates this situation.

While Amazon tracks its indirect water use, it has refrained from disclosing figures due to a lack of established industry benchmarks. Google and Microsoft have been similarly non-committal on the issue. Without a clear means to quantify these water risks, incorporating them into loan documents becomes difficult, effectively leaving this critical aspect unaccounted for in financing deals.

Over time, this oversight poses a risk as conditions affecting water availability—like droughts and changes in river flows—can shift dramatically, impacting long-term agreements that may not adapt to such variations.

State Responses to Water Concerns

Some regions are beginning to acknowledge this issue. In Texas, for example, where data centers account for nearly 90 percent of new power requests, Governor Greg Abbott has initiated a freeze on interconnection approvals until developers can verify their water use reporting and ensure the sustainability of the grid. Still, Texas stands as an exception rather than a prevailing trend across the country.

Lender Perspectives on Water Liabilities

The distinction between the water liabilities lenders evaluate and those they defer to utilities becomes clearer when assessing individual facilities. Lenders often mandate the usage of closed-loop or dry cooling systems in data centers, which minimize or eliminate reliance on municipal water supply. They impose water efficiency targets as part of loan covenants and conduct thorough checks on water rights and permits before construction begins, as contested permits can cause delays—something lenders and developers want to avoid at all costs.

Data center projects that have faced scrutiny over water use failures typically represent situations that this diligence aims to prevent, such as the recent halt of Google’s data center in Chile due to a court ruling against its groundwater rights.

However, it’s vital to recognize that using less water-intensive cooling methods like dry cooling can lead to increased electricity demand, which, in turn, impacts water consumption at the generating facility. This water transfer creates a situation where consumption at the power plant goes unchecked during due diligence processes.

Implications for the Industry

The discrepancy in scrutiny between carbon emissions and water usage reflects a structural issue in the industry. Carbon accounting boasts decades of protocol, enabling companies to produce reliable Scope 3 emissions data for their cloud services. In contrast, there’s currently no comparable framework for water usage associated with electricity generation, leaving this critical metric unmonitored and unpriced in financial contexts.

Efforts are underway to rectify this gap. The SCS Global Services, along with organizations like the World Resources Institute and the CEO Water Mandate, is working on developing a scopes-based water accounting framework—an essential step toward recognizing the importance of these metrics in corporate sustainability efforts.

Ultimately, stakeholders in the compute procurement process—ranging from cloud providers to enterprises signing service agreements—must take a proactive approach. Even without formal standards, sustainability teams negotiating power agreements should inquire about assessments of water availability at power generation sites. Understanding the potential risks tied to water resources can significantly influence long-term operational viability when entering contracts.

The escalating demand for AI and cloud computing requires more than just a focus on efficiency; a holistic approach, mindful of water resource management, will become increasingly critical.