As the global economy races to satisfy an insatiable appetite for artificial intelligence, the narrative surrounding the sector has been dominated by a singular, urgent question: How do we power the next generation of data centers? While the challenge of meeting these massive electricity demands is widely understood, a critical, under-addressed risk is flowing silently through the infrastructure: the massive, indirect water footprint of the power plants feeding these facilities.
As investors, policymakers, and tech giants scramble to secure grid capacity, they are overlooking a fundamental dependency. Most of the water consumed by a data center does not occur at the facility itself—it happens at the power plant that generates the electricity. This "hidden water liability" is largely absent from current financing models, creating a systemic risk that threatens both environmental stability and long-term economic growth.
The Hidden Water-Energy Nexus
Understanding Indirect Consumption
To grasp the scale of the issue, one must look at how electricity is produced. The vast majority of baseload power—the steady flow of electricity required to keep AI clusters operational—comes from coal, natural gas, or nuclear power plants. These facilities operate on a thermal cycle: they heat water into high-pressure steam to drive turbines, then require vast quantities of additional water to cool that steam back into a liquid state.
According to research from the Lawrence Berkeley National Laboratory and analysis by asset management firm Janus Henderson, this indirect consumption accounts for up to 75% of a data center’s total water footprint. For context, Meta—one of the few industry leaders to provide transparent reporting—revealed in 2024 that its indirect water consumption was more than 20 times greater than the water consumed directly by its own cooling systems.
Despite these staggering figures, water usage at the point of generation remains a blind spot in project financing. When private credit firms—which now handle 70% to 90% of project-level borrowing for new data center construction—assess a deal, they focus on traditional metrics: tenant credit, lease terms, grid interconnection, and power delivery. Water at the source is treated as an externality, ignored under the assumption that utility providers will manage their own resource planning.
Chronology of a Growing Crisis
From Infrastructure Needs to Regulatory Freezes
The trajectory of this issue reflects the rapid acceleration of the AI industry.
- 2010–2020: The Efficiency Era: Data centers focused primarily on Power Usage Effectiveness (PUE). The industry successfully transitioned to more efficient cooling methods, including "dry cooling" and closed-loop systems, which significantly reduced direct water use at the facility level.
- 2022: The Generative AI Inflection Point: With the launch of ChatGPT and the subsequent massive scaling of Large Language Models (LLMs), power demand shifted from "moderate growth" to "exponential demand."
- 2023–2024: The Resource Collision: Major tech companies began facing public pushback regarding their water usage in drought-prone regions. Notably, Google’s expansion plans in Chile faced significant legal hurdles after local courts halted development over concerns regarding groundwater depletion.
- 2026: The Texas Precedent: In a watershed moment for the industry, Texas Governor Greg Abbott issued a freeze on interconnection approvals for new data centers. The order forced developers to verify their total water impact and prove grid sustainability before further development could proceed. This served as a stark warning to the sector: the status quo of unchecked resource consumption is no longer politically viable.
Supporting Data: The Disconnect in Disclosure
The divide between direct and indirect water accounting is not just a technical oversight; it is a structural failure. Currently, no industry standard exists for measuring indirect water footprints, making it difficult for lenders to incorporate these risks into loan covenants.
The Accounting Gap
| Metric | Status |
|---|---|
| Carbon (Scope 1, 2, 3) | Highly standardized, audit-ready, integrated into ESG reporting. |
| Direct Water | Commonly tracked, though standards vary by region. |
| Indirect Water (Grid) | Largely unmeasured, unpriced, and excluded from risk models. |
While companies like Amazon track their indirect water consumption, they have yet to publish the figures, citing a lack of industry consensus. Google and Microsoft have been similarly opaque regarding whether they currently account for the water intensity of their grid-supplied power. This lack of transparency creates a false sense of security for financiers who believe they are funding "water-efficient" projects, when in reality, they are simply moving the water burden upstream.
Official Responses and Industry Defense
The financial sector generally defends its current approach by pointing to the "Utility Responsibility Model." Under this framework, lenders argue that water scarcity is a localized issue managed by regulated utilities. Because utility companies are subject to oversight by public utility commissions (PUCs) and credit rating agencies, the risk of a "water-induced blackout" is considered to be baked into the utility’s long-term planning.
However, critics argue this defense is flawed. Utilities plan based on historical data, which is becoming increasingly irrelevant in an era of climate volatility. A 15-year lease contract might assume steady water availability, but if a drought triggers a state-mandated curtailment of industrial water use, the data center operator—and by extension, the lender—has no protection.
The "dry cooling" paradox further complicates this. By implementing dry cooling at the data center, developers reduce local municipal water demand but increase electricity usage to run the cooling fans. This increased power draw requires the grid to generate more electricity, which in turn consumes more water at the power plant. By attempting to solve a local water problem, the industry may be inadvertently exacerbating a regional water-energy crisis.
Implications for the Future
Toward a Scopes-Based Framework
The current mismatch between carbon scrutiny and water negligence cannot hold indefinitely. As climate-related risks become more frequent, the financial cost of this "hidden liability" will eventually manifest as project delays, regulatory fines, or higher utility costs passed down to tenants.
The Move Toward Accountability
A coalition of organizations—including the World Resources Institute (WRI), the World Wildlife Fund (WWF), and the CEO Water Mandate—is currently developing a scopes-based water accounting framework. This initiative aims to provide the same level of rigour for water as the Greenhouse Gas (GHG) Protocol has provided for carbon over the last two decades. Once this framework is established, "indirect water consumption" will likely become a mandatory reporting metric for any large-scale infrastructure project.
A Strategic Imperative for Enterprise Users
For companies buying cloud capacity or AI services, the implications are profound. If a data center’s sustainability claims are based only on direct water use, they are ignoring 75% of the total footprint. Sustainability teams must start asking harder questions during procurement:
- Was water availability assessed at the generation source?
- Does the power supply agreement account for drought-related grid curtailment?
- Is the data center’s cooling strategy inadvertently increasing the facility’s indirect water consumption?
Conclusion
The era of treating water as an infinite, free utility for power generation is ending. As states like Texas begin to use their regulatory power to curb data center expansion, the industry is entering a new phase of accountability. For lenders, developers, and tech giants, the risk is no longer just environmental—it is financial. Those who fail to account for the water behind their power will eventually find their operations stranded by the very climate reality they have ignored for so long. The transition to a water-conscious grid is not just a sustainability goal; it is a requirement for long-term operational survival in the age of AI.
