As the global economy races toward an artificial intelligence-driven future, the conversation surrounding the infrastructure of this transformation has reached a fever pitch. Stakeholders are laser-focused on the “power crunch”—the staggering demand for electricity required to fuel the massive server farms housing the next generation of large language models. However, a far more precarious and often overlooked resource is at the heart of this dilemma: water.
While data center developers and local governments squabble over gigawatts, a critical, under-accounted risk is bubbling beneath the surface. The power plants that sustain the AI boom require vast amounts of water for cooling and steam production. If those sources run dry or become legally contested, the “always-on” promise of cloud computing may face unprecedented volatility.
The Water Behind the Power: An Indirect Liability
To understand the scale of the crisis, one must look past the immediate cooling systems of the data centers themselves. While modern facilities are increasingly designed to minimize direct water usage—often through high-efficiency "dry" or "closed-loop" cooling technologies—the vast majority of a data center’s water footprint is actually indirect.
Most of the grid’s electricity is generated through thermal processes: coal, natural gas, and nuclear power plants all rely on water to produce steam to drive turbines, followed by even more water to condense that steam back into liquid. As a result, when a data center draws power from the grid, it is effectively “consuming” the water used at the power plant.
Research from the Lawrence Berkeley National Laboratory, frequently cited by asset management firms like Janus Henderson, suggests that this indirect consumption can account for up to 75% of a data center’s total water footprint. Meta, a rare industry leader in transparency, reported in 2024 that its indirect water consumption was more than 20 times higher than its direct operational usage. Despite these staggering ratios, this consumption remains effectively invisible in the current financial and regulatory ecosystem.
Chronology of a Growing Crisis: From Neglect to Regulation
The trajectory of this issue has moved from a fringe sustainability concern to a boardroom-level risk over the last few years.
- 2020–2022: The Efficiency Focus. During this period, the industry focused almost exclusively on direct water usage. Developers touted closed-loop cooling as the gold standard for environmental, social, and governance (ESG) compliance, successfully lowering municipal water draw at the data center site.
- 2023: The Realization of Scope 3. Sustainability analysts began drawing parallels between carbon accounting and water accounting. Just as companies had to learn to track "Scope 3" (supply chain) carbon emissions, the realization grew that the power sector’s water usage was essentially a "Scope 3" water liability for tech firms.
- 2025–2026: Financial and Regulatory Friction. As AI deployment hit a massive surge, the strain on regional power grids became undeniable. Private credit firms began handling 70% to 90% of individual project-level borrowing for new builds, yet their due diligence remained tethered to traditional metrics: cost, reliability, and tenant credit.
- August 2026: The Texas Pivot. In a landmark move, Texas Governor Greg Abbott ordered a freeze on grid interconnection approvals for new data centers. The mandate required developers to verify their water use reporting and ensure long-term grid sustainability—a stark signal that the era of "unquestioned growth" is drawing to a close.
Supporting Data: The Financial and Environmental Mismatch
The disconnect between how lenders assess risk and the reality of water scarcity is significant. Currently, lenders evaluate data center projects based on standard utility assessments. Because water utilities are regulated entities, lenders often assume that the responsibility for water resource management lies solely with the utility provider, not the data center operator.
However, this ignores the temporal mismatch. A data center contract may lock in power terms for 15 to 20 years. Conversely, environmental conditions—river flows, groundwater tables, and regional drought patterns—can shift drastically within a single season.
The data suggests that the push for "water-neutral" data centers is inadvertently exacerbating the problem. To achieve low direct water usage, developers are increasingly turning to air-cooling and other water-efficient technologies. These systems, however, are significantly more energy-intensive. By increasing electricity demand to save water at the rack, the data center forces the grid to work harder, which in turn necessitates more water usage at the power plant. It is a classic "hydrological leakage" problem: the water is saved at the data center, only to be consumed at the power plant, often in a location where the lender has never conducted a site assessment.
Official Responses and the Quest for Standardization
The lack of industry-wide reporting standards acts as a shield for both developers and lenders. Currently, there is no settled methodology for calculating a data center’s indirect water footprint.
When asked about their transparency regarding this issue, the responses from major tech players have been guarded. Amazon acknowledges tracking indirect water consumption but has declined to publish figures, citing the absence of a verified industry standard. Other major cloud providers have remained largely silent on whether they have even begun the process of auditing their indirect water impacts.
The institutional response is beginning to take shape, however. A coalition comprising the World Resources Institute (WRI), the World Wildlife Fund (WWF), the CEO Water Mandate, and SCS Global Services is currently developing a "scopes-based" water accounting framework. The goal is to provide a standardized protocol similar to the Greenhouse Gas (GHG) Protocol, which has guided corporate carbon accounting for two decades.
Implications: The Future of Compute
The implications for companies leasing cloud capacity—and the enterprises relying on them—are profound. As water becomes a more constrained resource, the following risks are likely to manifest:
1. The "Water Premium" in Energy Costs
If power plants are forced to invest in advanced water-recycling technology or if they face water-related curtailments during drought years, those costs will inevitably be passed on to the largest energy consumers: the data centers. Enterprises should prepare for "water-adjusted" power pricing.
2. Operational Curtailment
As seen in Texas and in the case of Google’s halted campus in Chile, legal battles over groundwater permits can lead to indefinite construction delays or operational shutdowns. Lenders who ignore the "water-power nexus" are effectively underwriting projects that may lack the legal or physical right to exist in a decade.
3. The New Standard for Due Diligence
For sustainability teams and procurement officers, the "check-the-box" approach to data center sustainability is no longer sufficient. When negotiating hosting agreements, companies must start asking the hard questions:
- "What is the source of the water for the power plants supplying this facility?"
- "Has the developer assessed the long-term hydrological stability of the grid provider?"
- "Is there a backup plan for the facility if the local power plant faces water-use restrictions?"
4. A Shift in Financial Risk
The financial sector is slowly waking up to the fact that water risk is, in fact, credit risk. As standards from the WRI and others become more widely adopted, we can expect to see lenders incorporating water-stress scenarios into their loan covenants. Projects that cannot demonstrate a secure, long-term water and energy strategy will likely face higher interest rates or find it harder to secure capital altogether.
Conclusion: Bridging the Gap
The AI revolution is inextricably linked to the physical environment. The electricity that powers our digital tools is not a magical, infinite resource; it is a physical output of a complex, water-dependent system.
For too long, the industry has operated under the illusion that "sustainability" meant only what happened within the four walls of a data center. But the true test of AI’s environmental impact lies in the rivers, aquifers, and reservoirs that sustain the grid. By bringing the indirect water footprint into the light, stakeholders can begin to price the risk accurately, incentivize better utility management, and ensure that the digital future doesn’t come at the expense of our most fundamental natural resource.
The question for the next decade is no longer just "How much power do we need?" but rather, "Is there enough water to keep the lights on for the long haul?" It is a question that, if left unasked by the financiers and developers of today, will be answered with extreme volatility by the markets of tomorrow.
