Data Center Power and Water: Which Public Numbers Can You Trust?

For data-center power and water, trust public numbers only when they identify the site or reporting boundary, the period measured, and the method used. Government and utility records, regulatory filings, and facility-level disclosures can establish specific facts; industry estimates and company-wide averages are useful context, not precise measures of a particular building’s impact.

Why the numbers differ

“Data center use” can mean several different things: electricity consumed inside a facility, power drawn from the grid, water used for cooling, or the upstream resources associated with generating electricity. Reports also vary in geography, time period, and whether they cover one building, a company’s entire fleet, or a forecast of future demand. Two figures may both be correct while measuring different things.

Power is usually expressed as a rate, in watts or megawatts; energy is power accumulated over time, in kilowatt-hours or megawatt-hours. A facility’s 100-megawatt capacity is not the same as using 100 megawatts continuously. Likewise, “water withdrawal” is water taken from a source, while “consumption” generally means water not returned to that source in the same condition or place. Always check which quantity a headline reports.

Which public sources are strongest?

Start with sources closest to the measurement, then use broader datasets to check context. No single public source answers every question, and a disclosure’s existence does not by itself make its numbers independently verified.

• Utility and grid records: Public utility filings, resource plans, interconnection documents, and regional grid-operator reports can show planned capacity, connection requests, or system-level electricity demand. A request or forecast is not proof that a data center has been built or is consuming that amount.
• Environmental permits and agency records: Water-use permits, discharge permits, and environmental assessments may identify authorized withdrawals, sources, or limits for a named project. Permitted use is not necessarily actual use. Records may also be incomplete, delayed, or difficult to compare across jurisdictions.
• Company sustainability reports: These may disclose annual electricity, water, or efficiency metrics. Check whether the figures cover all operations, only owned facilities, or a subset; whether they are measured or estimated; and whether an auditor provided assurance.
• Government energy statistics: National and regional statistics are useful for totals and trends. They commonly combine data centers with other commercial computing or communications uses, so they may not support claims about one operator or site.
• Planning documents and public statements: These help identify proposed projects and stated design capacity. Treat them as plans or claims unless operating records confirm actual demand.

For U.S. national context, the Energy Information Administration publishes electricity data, while the Department of Energy and Lawrence Berkeley National Laboratory have published analyses of data-center energy use. Such studies can be valuable, but read the publication date, scope, and estimation method: a national estimate is not a current meter reading for an individual facility.

How to judge a power figure

For a particular site, the most meaningful public figure is a dated measurement of electricity actually consumed, with a clear boundary. Ask whether it includes servers only or the whole facility, including cooling, lighting, power conversion, and other support systems. A grid connection’s maximum capacity, a project’s requested load, and a facility’s annual energy consumption are different measures.

Useful checks include:

• Capacity versus consumption: “100 MW” may describe a planned connection or maximum load. It does not establish average demand. Annual consumption in MWh or GWh is more informative about energy used over a year.
• Site versus campus: A campus may contain several buildings and phases. Confirm whether a reported number covers one building, the whole site, or a proposed build-out.
• Actual versus forecast: Utility forecasts and developer projections depend on assumptions about construction, occupancy, and computing demand. Look for actual operating data before treating a forecast as current use.
• Electricity source: A renewable-energy purchase or certificate does not mean a facility physically receives only renewable electricity at every hour. It describes a procurement or accounting claim whose details matter.

Power usage effectiveness, or PUE, is a ratio of total facility energy to IT-equipment energy over a defined period. A lower PUE indicates less overhead energy relative to IT energy, but it is not a measure of total electricity demand or environmental impact by itself. A very efficient, very large site can use more electricity than a smaller site with a higher PUE.

How to judge a water figure

Water data require an equally clear boundary. A facility may use water directly for cooling, receive water from a municipal system, or use little on-site water while relying on electricity whose generation consumes water elsewhere. Those are distinct impacts. A claim about “water use” that omits the source, metric, and reporting period is hard to compare.

Water usage effectiveness, or WUE, is commonly reported as water used divided by IT energy, often in liters per kilowatt-hour. It can help compare operational water intensity when organizations use consistent definitions. It does not reveal the local scarcity of the water source, the facility’s total volume, or indirect water associated with electricity. A low WUE is not automatically a low-impact result if the site is much larger or located in a water-stressed basin.

Check whether the reported figure is potable water, reclaimed water, or a mixture; whether it measures withdrawal or consumption; and whether it includes only cooling or all facility operations. Seasonal variation matters, too. Annual totals can conceal periods when water demand coincides with drought or peak community demand.

Common traps in comparisons

One frequent mistake is comparing a proposed campus’s maximum power capacity with another operator’s annual energy use. Another is treating a company-wide average as evidence for a named site. Both comparisons mix different scopes or units.

Watch for these warning signs:

• A number has no date, source, geographic boundary, or explanation of whether it is measured, modeled, permitted, or projected.
• Power is described as “energy,” or a peak load is presented as typical continuous use.
• Water withdrawal, water consumption, and recycled-water use are treated as interchangeable.
• A global efficiency average is applied to a local facility without evidence that its cooling design, climate, and operating conditions match.
• A headline repeats an estimate without linking to the underlying report or explaining uncertainty.

For a quick check, trace the claim to its original document and record the operator, location, period, unit, boundary, and method. If one of those is missing, label the number as incomplete rather than silently filling in the gap.

Frequently Asked Questions

What is the most reliable public source for a data center’s electricity use?
There is no universal source. A dated, facility-level consumption disclosure with a stated boundary and independent assurance is strong evidence. Utility connection records usually show capacity or plans, not necessarily actual consumption.

Does a 100 MW data center use 100 MW of electricity all the time?
No. That may describe a maximum or planned load. Actual demand varies, and annual energy use must be reported in units such as MWh or GWh. Do not infer annual consumption without a measured load profile or a clearly stated assumption.

Does a data center use more water than a household?
It depends on the facility, cooling system, location, and what is counted. Compare dated volumes using the same definition and period; a broad household comparison may be misleading if it mixes direct and indirect water use.

Is WUE enough to compare data-center water impacts?
No. WUE describes water intensity relative to IT energy, but not total volume, local water scarcity, or indirect water used in electricity generation. Use it alongside site-level water data and location context.

Are company sustainability figures independently verified?
Some are subject to external assurance and some are not. Read the report’s assurance statement to see which metrics were checked, by whom, and to what level. An assurance of selected data does not necessarily cover every claim.

A practical standard for trust

Trust a public number in proportion to its specificity and documentation. A facility-level, time-bounded, clearly defined measurement is stronger than a forecast; a transparent, independently assured disclosure is stronger than an unexplained headline. For broader questions, combine those records with government and grid data, while keeping capacity, consumption, direct water, and indirect impacts separate. When no source provides the needed detail, the accurate conclusion is that the public evidence is insufficient—not that a convenient estimate is a measured fact.

Related

• Best HIIT YouTubers for Beginners to Burn Calories
• Micro1’s $1 Billion Data-Buying Plan Explained
• Robots.txt and AI Use: What the Two Choices Mean