Data Center Water, Land and Jobs Statistics: The Local Footprint

Data center water, land and jobs statistics sit at the heart of the local debate over whether communities should welcome new campuses. This article in the Data Center Economy hub gathers the best available evidence on water consumption, land use, employment and tax revenue, drawing on Virginia’s legislative auditor JLARC, Lawrence Berkeley National Laboratory and the hyperscalers’ own sustainability reporting. It complements our data center electricity consumption statistics and our deeper look at data center sustainability, carbon and water.

How much water does a data center use?

Water use varies enormously by cooling design, climate and size. Lawrence Berkeley National Laboratory estimates the average US data center consumed just over 0.36 litres of water on site per kWh through 2023, and expects that to rise to 0.45–0.48 litres per kWh by 2028.

Most data center water is used for evaporative cooling, where water absorbs heat as it evaporates. Facilities in hot, dry climates that rely on evaporative towers use far more than those using air-cooled chillers or closed-loop liquid cooling. LBNL’s 2024 report for the US Department of Energy attributes the expected rise in average water intensity to the shift toward hyperscale and colocation facilities, many of which use evaporative systems to save electricity.

At the facility level, published estimates span a huge range. The Environmental and Energy Study Institute (EESI), summarising press reporting, cites estimates that a large data center can use up to 5 million gallons a day and a medium-sized facility around 110 million gallons a year for cooling. Virginia’s JLARC took a more measured view in its 2024 study: most data centers use about as much water as a large office building, a few use substantially more, and some use less than a typical household. Our explainer on air, liquid and immersion cooling covers the engineering trade-offs.

Direct versus indirect water

On-site cooling is only part of the picture. Power plants also consume water, so data centers have an indirect water footprint through their electricity use. EESI cites LBNL’s estimate that US data centers’ indirect water footprint was about 211 billion gallons in 2023 — far larger than typical on-site figures. That means a facility that switches from evaporative cooling to electricity-hungry chillers may reduce local water use while increasing water consumed at distant power plants, unless its power comes from wind or solar.

How much water do Google, Microsoft and Meta use?

The hyperscalers report water differently, which makes comparisons difficult. Google says its water projects replenished about 7.7 billion gallons in 2025, roughly 78% of its freshwater consumption, while Microsoft says it replenished more than 14.2 million cubic metres in fiscal 2025.

Google’s sustainability site states that 87% of its 2025 freshwater withdrawals came from sources at low or medium risk of depletion or scarcity. Since 7.7 billion gallons equals about 78% of consumption, the implied total is around 9.9 billion gallons, though Google directs readers to its 2026 Environmental Report for exact figures. Google also reports a fleet-wide power usage effectiveness of 1.09 for 2025, one of the lowest in the industry.

Microsoft’s 2026 Environmental Sustainability Report says its fiscal 2025 replenishment exceeded the water it withdrew. It highlights a 23% year-on-year improvement in water usage effectiveness at its Phoenix data centers after cooling upgrades, and rainwater harvesting at sites in Ontario and Québec expected to collect about 1.5 million litres a year.

Meta says data centers are the largest water users in its operations and has set a goal to be water positive in 2030, restoring 200% of consumption in high water-stress watersheds and 100% in medium-stress ones. Its restoration projects returned more than 1.6 billion gallons to stressed regions in 2024.

Source: Google sustainability and data center efficiency pages; Microsoft 2026 Environmental Sustainability Report; Meta sustainability (water); LBNL 2024 US Data Center Energy Usage Report; EESI.

How much water do data centers use in Northern Virginia?

Northern Virginia, the world’s largest data center market, offers the best local data. EESI cites Financial Times reporting that data centers in the region used close to 2 billion gallons in 2023, up 63% from 2019, with Loudoun County facilities accounting for roughly 900 million gallons.

Despite that growth, JLARC concluded in December 2024 that data center water use in Virginia was growing but currently sustainable. Operators can further reduce pressure on drinking-water supplies by using recycled or non-potable water for cooling where local utilities offer it. The bigger concern in Virginia is electricity, not water: JLARC found the state’s data center sites used about 5,050 MW and that unconstrained statewide demand could double within a decade.

Water stress is far more acute in desert markets such as Phoenix, where hyperscalers have invested in lower-water cooling, and in parts of Texas and Spain where droughts recur. The local context — not national averages — determines whether a project’s water use is a problem.

How much land do data centers take up?

Land use is growing quickly as campuses scale up. JLARC found Virginia’s roughly 150 data center sites cover about 7,200 acres and contain more than 63 million square feet of data center space across about 340 buildings.

New AI campuses are an order of magnitude larger than traditional facilities. Meta’s Hyperion data center in Richland Parish, Louisiana occupies a site of about 2,250 acres with a planned 4 million square feet of floor area and power demand of up to 5 GW. Epoch AI’s tracking of OpenAI’s US Stargate sites shows multiple campuses projected at 1.2–2.2 GW each, many of them on former farmland or industrial land in rural counties.

The land itself is rarely the binding constraint — power is. JLARC noted data centers need large, flat parcels, and developers increasingly choose sites based on access to transmission lines and substations rather than proximity to cities. That is why JLL reports that 77% of North American capacity under construction is in frontier markets such as West Texas, Ohio, Louisiana and Indiana. Our data center site selection guide explains the full checklist, and our pipeline statistics track where the big campuses are rising.

How many jobs do data centers create?

Data centers create many construction jobs but few permanent ones. JLARC found about 1,500 workers can be on site at peak for a single building over a 12–18 month build, while a typical 250,000 sq ft facility then employs about 50 full-time workers, roughly half of them contractors.

At the state level, the numbers are still substantial because so much building is underway. JLARC estimated Virginia’s data center industry supports about 74,000 jobs, $5.5 billion in labor income and $9.1 billion in GDP annually, but stressed that most of this comes from construction rather than ongoing operations. If construction slowed, much of that economic activity would fade.

Large AI campuses come with headline job claims that are hard to compare. Meta estimates Hyperion will support about 1,500 jobs, without separating construction from permanent roles in public statements. The Stargate venture has said it will create more than 100,000 US jobs, a figure that includes construction and indirect employment. For a sense of how operators staff their facilities, see our guide to data center operating costs and staffing.

Construction versus operations

Economists sometimes describe data centers as capital-intensive rather than labour-intensive: hundreds of millions or billions of dollars of equipment are run by small teams of technicians, security staff and engineers. The construction phase supports electricians, pipefitters and equipment suppliers, often brought in from outside the county. Communities expecting a factory-style employment base are often disappointed; those focused on property tax revenue are usually satisfied.

How much tax revenue do data centers generate?

Data centers can transform local budgets. JLARC found data center revenue made up 31% of Loudoun County’s total local tax revenue in FY23, and ranged from under 1% to 31% across five mature Virginia localities. Most comes from business personal property tax on servers and equipment.

Because servers are replaced every few years and taxed as business equipment, a data center can generate far more local revenue per acre than housing or offices, while demanding few school places or social services. That fiscal surplus — the ability to fund schools and services while easing pressure on residential tax rates — is one of the strongest arguments made by supporters in counties such as Loudoun.

State-level economics are less favourable. JLARC calculated Virginia’s sales and use tax exemption for data center equipment cost $928 million in FY23, with about 90% of the industry using it. The exemption returns about 48 cents in state revenue per dollar of tax forgone — better than the 41-cent average for Virginia economic development incentives, but still a net cost to the state. The exemption is scheduled to expire in 2035. Our guide to US data center investment compares state incentive regimes.

Why is local opposition to data centers growing?

Opposition is growing because residents see the costs — noise, land, transmission lines, water and electricity bills — more clearly than the benefits. JLL’s 2026 survey found only 14% of Americans would support a data center in their community, even though 79% support US leadership in AI.

Sightline Climate reported in February 2026 that moratorium proposals on new data centers had been introduced in at least ten US states, including Louisiana, Michigan, New York, Ohio and Virginia, and that a proposed $1 billion project in Michigan was cancelled after local opposition. In Ireland, where data centers used 23% of metered electricity in 2025, grid constraints have halted most new connections around Dublin.

Supporters counter that data centers deliver large tax surpluses with little demand on public services, fund grid upgrades through large-load tariffs and anchor investments in renewable and nuclear power. The balance of these arguments differs by location, which is why the same project can be welcomed in one county and rejected in the next. Our analysis of who pays for grid upgrades examines the electricity-bill debate in detail.

What do these statistics mean for communities and investors?

For communities, the evidence suggests data centers are strong tax generators but weak permanent employers, with water impacts that depend heavily on cooling design and local supply. For investors, local acceptance has become a material project risk.

Local officials evaluating a proposal should ask for site-specific water consumption and supply source, the expected power draw and who funds grid upgrades, a realistic split between construction and permanent jobs, and projected property tax revenue over the equipment replacement cycle. Comparing those figures with benchmarks from JLARC and the hyperscalers’ disclosures gives a clearer picture than headline investment totals.

For investors and developers, community support now affects timelines as much as permitting or power. Projects that use reclaimed water or closed-loop cooling, commit to community benefit agreements and fund their own grid upgrades are likely to move faster. For wider context, see how many data centers there are in the world and our overview of how to invest in data centers.

Frequently Asked Questions