Investors are ignoring the biggest source of data-center water use
Lenders that finance data‑centre construction are focusing on the water drawn on site, even though most of a facility’s water footprint comes from the electricity it purchases, a new analysis shows.
Source: GreenBiz · October 3, 2026 at 10:00 PM · AI-assisted report
Single-sourceTEXAS, 4 OCTOBER 2026 —
Lenders that finance data‑centre construction are focusing on the water drawn on site, even though most of a facility’s water footprint comes from the electricity it purchases, a new analysis shows.
The finding highlights a gap in risk assessment for the power‑intensive AI and cloud‑computing boom, because indirect water use tied to power‑plant operations is rarely disclosed or priced into loan agreements.
The analysis, produced by the nonprofit Ceres and market‑research firm Bluefield Research, says the majority of a data centre’s water consumption is linked to electricity generation rather than to the cooling systems that lenders typically scrutinise. Janus Henderson, an asset‑management firm, cites research from Lawrence Berkeley National Laboratory indicating that indirect water use can account for up to 75 percent of a data centre’s total water footprint.
Meta, one of the few operators that has published an estimate, reported that its indirect water use for 2024 was more than 20 times the volume it consumed directly.
Power plants that run on coal, natural gas or nuclear fuel heat water into steam to drive turbines and then use additional water to condense the steam back into liquid. As a result, a data centre that purchases electricity from such plants consumes water indirectly, even if its own cooling system is dry‑cooled or closed‑loop and draws little or no municipal water.
That indirect consumption does not appear on an electricity bill, and lenders assess power primarily on cost, reliability and availability. The water intensity of the generating plant is therefore seldom a factor in financing decisions.
From a financial‑risk perspective, the exposure to water scarcity is often left to the regulated utility that supplies the power. Credit‑rating agencies already evaluate utilities’ water‑planning duties, providing a “reasonable defence” for lenders who delegate the risk. However, there is no settled methodology for measuring or reporting a data centre’s indirect water footprint. Amazon tracks indirect water use but has not published figures, citing the absence of industry standards.
Google and Microsoft declined to confirm whether they monitor such consumption.
The lack of a standard metric makes it difficult to embed water risk in loan covenants. While lenders may require dry‑cooling or closed‑loop systems and set water‑efficiency targets, they also check water rights and permits before construction to avoid delays caused by contested groundwater licences. Nevertheless, a dry‑cooled facility consumes more electricity, shifting water demand upstream to the power plant—a factor that may escape the lender’s diligence.
The time horizon of data‑centre contracts, which often span decades, amplifies the risk. River flows, drought conditions and groundwater levels can vary dramatically from year to year, potentially leading to curtailments or cost pass‑throughs later in the lease term. Some jurisdictions are beginning to address the issue.
In Texas, where data centres account for nearly 90 percent of new grid‑power requests, Governor Greg Abbott ordered a freeze on interconnection approvals until developers can verify water‑use reporting and demonstrate grid sustainability. Texas remains an outlier, however, as most other states have not introduced comparable requirements.
The mismatch between carbon accounting and water accounting stems from the maturity of the respective frameworks. Carbon emissions have benefited from two decades of protocols, scope definitions and assurance mechanisms, enabling sustainability teams to produce defensible Scope 3 figures for cloud use. By contrast, water consumption associated with electricity generation lacks an equivalent accounting apparatus, leaving it uncounted on the corporate side and unpriced on the lending side.
Efforts are under way to close the gap. An initiative led by sustainability‑standards organisations SCS Global Services, the World Resources Institute, the World Wildlife Fund and the CEO Water Mandate is developing a scopes‑based water‑accounting framework, described as the closest analogue to the carbon‑accounting system that has evolved over the past twenty years.
In the absence of formal standards, the onus falls on the buyers of compute power. Companies that lease colocation capacity, cloud providers or enterprises signing AI‑service agreements can ask whether water availability at the generating plant has been assessed. The response indicates whether the capacity they are contracting for rests on a resource that has been examined.
A dry‑cooled campus operating in a drought year still depends on electricity from a plant that may need river water, and the availability of that water in the later years of a 15‑year lease should not be left to chance.
Related: Texas
Malaysia Impact
2/10Potential indirect impact on Malaysian data centers and energy-intensive sectors (e.g., AI/cloud computing) if they rely on power plants with high water intensity, though no direct Malaysian entities or policies are mentioned.
energytechnologycommodities