Water Abundance Is Not Water Availability

Wisconsin sits within one of the largest freshwater systems on the continent. That regional abundance does not determine what any single facility can withdraw, consume, or sustain — and the distinction is becoming central to large-load planning.

Discussions about siting water-intensive infrastructure in Wisconsin often begin from the same premise: the state has abundant fresh water, so water is unlikely to be the limiting factor. The premise is accurate as a description of the region. It is not a reliable guide to what any particular site can do. Abundance describes a system. Availability is a property of a location, a permit, and a watershed. The two are related, but they are not the same measure, and treating them as interchangeable introduces error into siting and capacity planning.

Abundance describes the region, not the site

Wisconsin sits within the Great Lakes basin and, secondarily, the Mississippi basin. By any regional accounting, the state holds significant fresh water. That position is real, and it is durable. It is also governed at the scale of the basin. The Great Lakes Compact, for example, restricts diversion of water outside the basin, a framework built around the integrity of the regional system as a whole.

None of that regional structure speaks to the question a specific developer or planner actually faces: how much water can this parcel draw, and at what rate, without exceeding what the local supply and the permitting process allow. Regional abundance sets the backdrop. It does not set the site-level number. A facility does not withdraw from “the Great Lakes” in any practical sense; it withdraws from a municipal system, a river reach, or an aquifer, each with its own capacity and its own constraints.

Availability is defined at the point of withdrawal

Site-level availability is shaped by a permitting structure that operates well below the scale of the basin. Any new water withdrawal above 100,000 gallons per day within the Great Lakes basin requires registration and permitting through the state. That threshold is a site-level gate. It applies regardless of how much water the region holds, and it is one of the first places where abundance and availability diverge in practice.

The picture is complicated further by how a facility sources its water. A site that draws directly from surface water or groundwater generally holds its own withdrawal registration or permit. A facility served by a municipal system may not need its own withdrawal permit at all, because the utility holds it. That arrangement does not make the water less finite; it shifts where the withdrawal is accounted for. The site’s demand still lands on a real watershed, even when it is aggregated inside a municipal total. Understanding availability therefore means looking at both the permitting threshold and the path the water actually takes to the site.

Consumptive use changes the accounting

Volume withdrawn is only part of the question. What matters for a watershed over time is how much of that water is consumed — lost to the system — rather than returned. Here, cooling introduces a specific and often underappreciated effect. More than half of the water used in evaporative cooling is consumed to evaporation and does not return to the watershed. This is a meaningful departure from much historic industrial cooling, which was largely non-consumptive: water was drawn, used, and returned to rivers at close to the volume taken.

A withdrawal that is mostly consumptive draws down local supply in a way that a mostly-returned withdrawal does not, even at the same headline volume. Scale sharpens the point. A single large evaporative-cooled data center can require on the order of one to five million gallons per day. Consumptive use above five million gallons per day within the basin draws additional scrutiny under the Great Lakes Compact Council. The threshold exists precisely because consumptive draw at scale is a different category of demand from historic industrial water use.

Watersheds do not average out

Regional abundance can create the impression that local supply is effectively unlimited, but water availability is not fungible across the basin. A watershed or aquifer under one site does not benefit from the abundance of a lake fifty miles away. Localized overuse can stress smaller aquifers and watersheds even where the broader system remains abundant. Wisconsin’s Central Sands region is an existing example of this dynamic, where irrigation demand already pressures local groundwater. Drought narrows the available margin further and does so unevenly across the state.

The practical consequence is that two sites a modest distance apart can face very different answers. One may sit on a municipal system with ample headroom; another may sit above an aquifer already carrying substantial existing demand. Regional data cannot resolve this. Only a parcel-level assessment of the specific supply, its current load, and the consumptive profile of the proposed use can.

Why the distinction is becoming more consequential

For most of the past three decades, aggregate water demand in Wisconsin has trended downward. Statewide water-utility sales fell from roughly 175 billion gallons in 1998 to about 138 billion in 2023, an approximately 18 percent decline, and industrial water use fell by roughly a third since the late 1990s. That trajectory reflected efficiency gains and a shifting industrial base, and it shaped an intuition that water demand is a receding concern. Large-scale, water-intensive compute introduces a new demand profile that does not follow that trend. A single planned corridor of Midwest data-center development has been estimated to require on the order of 150 billion gallons over five years — a figure that has to be located against specific supplies, not against a regional total.

The point is not that the region cannot support additional demand. It is that a declining aggregate trend at the state level tells a planner very little about the capacity of any single watershed asked to absorb a new, concentrated, partly consumptive load. Abundance measured across decades and across a basin can coexist with genuine constraint at a particular site in a particular year. The two readings are not in tension; they are answers to different questions.

What this means for planning

The distinction between abundance and availability is not academic. It determines whether a site’s water demand is genuinely supportable, and it should be evaluated at the level of the individual parcel and its watershed rather than inferred from regional totals. Cooling architecture is a significant variable in that assessment, because it governs how much of a site’s withdrawal is consumptive. One announced first-phase campus in the state reported an intended draw of roughly eight million gallons per year from the local system — a relatively low figure, consistent with heavy reliance on air economization or closed-loop cooling. That a facility of significant scale can operate at a modest annual draw underscores the point: site-level water demand is a design and permitting outcome, not a fixed consequence of the load.

For planners, utilities, and public decision-makers, the useful frame is site-specific: what a particular parcel can withdraw, consume, and sustain under existing thresholds, evaluated against the real capacity of its local supply. Regional abundance remains an asset. It is simply not a substitute for that parcel-level analysis.

The open question for the field is whether current planning and permitting frameworks are set up to evaluate water at the parcel and watershed level, at the pace that large-load siting now demands. Where does the responsibility sit for ensuring that a site’s assumed abundance and its actual availability are the same number?

Source: the Wisconsin AI Infrastructure Readiness Brief, cooling and water — basin abundance against local availability, the 100,000 gallon-per-day withdrawal registration threshold, municipal-supply accounting, the consumptive share of evaporative cooling, and Great Lakes Compact Council scrutiny above five million gallons per day of consumptive use.