For years, the AI infrastructure conversation has focused heavily on carbon. Data center operators publish efficiency targets, renewable energy commitments and power usage metrics. Investors and regulators have built entire reporting frameworks around emissions. Water has received far less attention, even though it is becoming one of the most important constraints on where and how new AI infrastructure gets built.
That gap is getting harder for the industry to defend. Earlier this year, more than a dozen institutional investors pushed Amazon, Microsoft and Alphabet for more detailed site-level water and power disclosure ahead of their annual investor meetings. Among the three largest cloud providers, Google has gone further than its peers in publishing that kind of data. The pressure on the rest of the market is not going away because investors are starting to understand that water exposure is not just an environmental issue. It is a financial, permitting and reputational risk.
Communities are applying the same pressure from a different direction. When residents push back on a large data center project because they do not understand how much water it will use, what the community receives in return or how the operator plans to manage local resource strain, the issue is not simply opposition to development. It is often a lack of trust. Without clear site-level data, communities are left to assume the worst, and that makes resistance more likely before a project even reaches the point of construction.
The industry’s current measurement standard does not fully solve that problem. Water Usage Effectiveness, the metric most operators use, normalizes water consumption against IT load. On its face, that sounds reasonable. In practice, it can make a facility look more efficient while still using a significant amount of water in absolute terms. A company can show progress in a ratio while the real water footprint remains difficult for communities, regulators and investors to evaluate.
That distinction matters because water risk is local in a way carbon reporting often is not. A portfolio-wide sustainability number does not tell a community whether a facility will strain its watershed, compete with residential or agricultural demand, or require additional infrastructure that local residents will ultimately have to live with. Aggregated reporting may be useful for a corporate sustainability report, but it is not enough for permitting decisions or community trust.
There is also a tradeoff emerging that deserves more scrutiny. Some operators are responding to water concerns by moving toward dry cooling systems, which can reduce a facility’s direct water use. But dry cooling can require more energy, and more energy demand can create additional water use elsewhere in the power generation lifecycle. The water cost may not disappear. It may simply move to a part of the system that is harder for communities and regulators to see.
That is the problem with treating energy and water as separate reporting categories. In infrastructure, the two are deeply connected. Every megawatt of power has a water footprint somewhere in its lifecycle, and every gallon of water treatment, transport or desalination has an energy footprint. Reporting frameworks that separate the two can create a cleaner story on paper than the one that exists on the ground.
I work in hydrogen-based power generation, where the relationship between energy and water is impossible to ignore. Power production, water treatment and infrastructure sit in the same system, even when companies account for them in different spreadsheets. That is why AI infrastructure needs a more complete disclosure model, not one that allows companies to show improvement in one category while pushing costs into another.
The fix is not especially complicated, even if the politics around it are. Water reporting should include absolute consumption figures, not only efficiency ratios. It should account for the energy-water tradeoff when operators change cooling methods. It should also provide site-level data rather than relying on portfolio averages that can hide the impact of individual facilities.
Regulators are already starting to move in this direction. States are increasingly factoring water impact into permitting decisions for new energy and data center infrastructure, much as they have long evaluated air, land use and grid impact. The timeline for those approvals can vary significantly by state, and water is becoming another variable that can slow or reshape a project. Operators that have not built water transparency into their planning may find themselves facing delays they did not anticipate.
This is not an argument against AI infrastructure or against building the data center capacity the market clearly needs. Demand for compute is not going away, and pretending otherwise does not help communities, regulators or companies make better decisions. The argument is that the industry has a chance to create a water disclosure framework before it is forced to do so through a patchwork of state rules, investor mandates and community opposition.
That proactive approach would also be less expensive. It is cheaper to build transparency into site selection, permitting and community engagement from the start than to retrofit compliance after a project has been challenged or delayed. Water disclosure should not be treated as a threat to competitive advantage. Communities and investors are not asking operators to reveal proprietary cooling designs or server architecture. They are asking how much water a facility uses, where that water comes from and what happens to it afterward.
Those are infrastructure facts, not trade secrets. Operators that recognize that distinction will be in a stronger position as the AI buildout accelerates. They will have a clearer path through permitting, a stronger case with local communities and fewer surprises when investors ask for more detailed disclosure.
The AI industry has spent years learning how to count carbon. Water is now becoming the next test of whether the industry can grow responsibly in the places where infrastructure is actually being built. The companies that treat water transparency as part of the planning process, rather than a disclosure problem to manage later, will have a real advantage as communities, regulators and investors raise the bar.
