A number can look enormous before anyone asks what it is being measured against. That is increasingly true in debates over data centers, where water consumption, electricity demand, land requirements, tax incentives and even low-frequency sound can become powerful symbols long before the underlying measurement receives a proper frame. The problem is not that these figures are invented. Many are real, and some describe substantial resource requirements. The problem begins when a measurement becomes its own conclusion. A data center consuming millions of gallons of water sounds alarming when presented without the size of the facility, the cooling technology, the local water system, the period being measured or the source of that water. Electricity demand can look similarly stark when a project’s megawatt requirement appears without comparison to the surrounding grid, the region’s existing industrial load or the infrastructure required to serve other large consumers.
A large number is not automatically a large share
Data center debates often rely on the psychological force of absolute numbers. Millions of gallons sounds excessive. Hundreds of megawatts sounds enormous. Thousands of acres sounds transformative. Billions of dollars in tax exemptions sound like a direct public loss. Yet each figure answers only one question: how much? It does not answer how much relative to what. Industry research illustrates why this distinction matters. Its 2024 U.S. It estimates that data centers consumed about 4.4% of U.S. electricity in 2023, while projecting a range of 6.7% to 12% by 2028. The updated 2025 report projects a broader 2030 range of 9.5% to 15.3%. Those figures indicate a rapidly growing electricity load, but they also demonstrate why national percentages, regional grid conditions and individual project requirements must remain separate measurements.
The same principle applies to water. A facility’s annual water consumption cannot tell the full story without identifying whether the figure represents direct onsite consumption, indirect water associated with electricity generation, peak demand or an annual average. Cooling design can also materially change the result. Berkeley Lab’s research has explicitly treated direct and indirect water consumption as distinct components, reinforcing the point that a single water figure can conceal several different physical processes. That does not make the water requirement irrelevant. It makes the denominator essential.
Geography can turn the same statistic into a different problem
The most important denominator in infrastructure debates is often geographic. A water withdrawal that creates limited pressure within one large watershed can carry very different consequences in a drought-stressed basin. A major electricity load can have a different effect on a transmission-constrained region than on an area with substantial generation capacity and available interconnection infrastructure. The facility has not changed. The surrounding system has.
That is why national averages can become misleading when applied to individual projects. A percentage of U.S. electricity consumption cannot by itself establish whether a particular data center creates a local reliability problem. Likewise, an annual water figure cannot establish whether a facility materially competes with municipal supply without information about the relevant watershed, utility system, seasonal availability and competing demand. The correct comparison therefore may not be between a data center and an abstract national total. It may be between the facility and the marginal capacity available in the particular system where it operates. That is a much harder calculation, but it is also a more useful one.
Tax numbers need a similar baseline
Tax incentives create another version of the same problem. A headline may describe a data center receiving a large exemption and present the figure as revenue that governments have effectively surrendered. But an incentive cannot be evaluated from its headline value alone. The relevant comparison includes the tax treatment that would otherwise apply, the investment that qualifies, the duration of the exemption, the taxable property involved and the revenue generated under the resulting development.
Tax Foundation research published in 2025 found that data center tax burdens vary substantially across jurisdictions and that exemptions often depend on investment or economic-development requirements. It also found that data centers can remain significant contributors to local tax collections despite specific exemptions. None of this proves that every incentive represents good policy. It demonstrates something more basic: an exemption figure is not equivalent to a net fiscal impact. The calculation requires a counterfactual. Without that baseline, the number describes a policy mechanism rather than its economic result.
The missing number is often the denominator
This does not mean data center impacts should be minimized. Quite the opposite. If water consumption is significant relative to a stressed municipal supply, the comparison should make that clear. If a facility represents a meaningful share of local electricity demand, that deserves attention. If an incentive materially changes expected public revenue, the financial model should expose the difference. But those conclusions become stronger, not weaker, when they rest on comparable measurements. The industry also benefits from this discipline. Operators that publish only large absolute figures invite interpretation.
More useful disclosures would show consumption per megawatt of IT load, water intensity by cooling system, peak versus average demand, land use relative to developed capacity and tax benefits against the revenue and investment assumptions that support them. The debate would then move away from the spectacle of large numbers and toward the mechanics of infrastructure. That is where the real questions sit. A million gallons is a quantity. So is 500 megawatts. Neither one, by itself, is an argument. The increasingly important task for data center reporting is to ask what those numbers represent, what they are measured against and which part of the system actually absorbs the cost. Without those questions, opposition can mistake scale for significance. With them, the numbers become much more difficult to dismiss and much harder to misuse.
