Abstract / About
The rapid rise of artificial intelligence and hyperscale computing has turned freshwater from a background utility into one of the central constraints on technological growth, yet the institutions that govern water have not kept pace. AI data centers withdrew on the order of 264 billion gallons in 2025, and direct water use by United States data centers is projected to climb from roughly 17 billion gallons a year toward 38 to 73 billion gallons by 2028. The water embedded in electricity generation and chip manufacturing pushes the real total higher still.
This paper argues that the leading industry response — the voluntary “water positive” pledges that major hyperscalers have made for 2030 — rests on an accounting foundation too fragile to bear the weight now placed on it. Estimates of how much water AI actually consumes differ by one to two orders of magnitude depending on what is counted. Replenishment credits are issued against inconsistent baselines, checked by a patchwork of registries with uneven oversight, and remain open to double counting and to projects that would have happened anyway.
The right to water is itself divided across a tangle of state rules, interstate compacts, and ongoing lawsuits that treat surface water and groundwater as separate worlds. The deeper problem is that water, unlike carbon, cannot be freely traded across places and times: a gallon restored in one basin does not cancel out a gallon drained from a stressed basin somewhere else — a fact that current offset frameworks tend to obscure.
We therefore make the case for a single, independently verifiable water-credit system that works across the country and is anchored to the hydrology of individual basins, so that every withdrawal, return, and act of replenishment can be measured, audited, and traced back to the watershed where it occurs. Such a system, we conclude, is not only an environmental safeguard but a precondition for the legitimacy and staying power of the computer-driven economy.
Who it’s for
Hyperscale operators and their sustainability leaders, infrastructure investors, water utilities and basin regulators, and policymakers who need a defensible, verifiable picture of AI’s water footprint — not a marketing one.
Need this scoped to a specific portfolio or region? Commission a custom briefing →