The newsroom · Source-backed insight

Data centers are good. Here is what host communities are getting.

America should build more compute and the power, water and connectivity systems to support it. Loudoun County, The Dalles, Prineville and Henrico County show how host communities can convert that investment into public revenue, infrastructure, work and household value.

By Singularity Farmer · Published September 24, 2026

Data centers are good infrastructure. America should build more of them—and build the generation, transmission, water systems and connectivity needed to support them.

The strongest case begins with what host communities have already received.

In Loudoun County, Virginia, data centers generated $1.2 billion in FY 2026, equal to 39% of the county budget. The county says the revenue helped fund schools, public safety, libraries, parks, social services and capital improvements while the homeowner real-property tax rate was lowered. (Loudoun County)

In The Dalles, Oregon, agreements required the data-center owner to fund pipeline, well, reservoir, pump-station and distribution upgrades rather than shifting those infrastructure costs to existing customers. City officials reported that data-center revenue held a planned 2025 water-rate increase to 7% rather than 35%. (Oregon Data Center Advisory Committee)

Henrico County, Virginia, created a $60 million affordable-housing trust backed by data-center tax revenue. By August 2026, the county reported 87 first-time home sales through the program. (Henrico County)

These are not promises about every future project. They are delivered examples showing what communities can pursue when they welcome investment and put local value into the deal.

Electricity is capacity, not a moral verdict

Berkeley Lab estimates that U.S. data centers used 176 terawatt-hours in 2023, about 4.4% of total U.S. electricity use. (Berkeley Lab)

The comparison class matters. EIA reports 2023 retail electricity sales of 1,450 TWh residential, 1,408 TWh commercial, 1,009 TWh industrial and 6.9 TWh transportation, totaling 3,874 TWh. Those sector figures provide context for the data-center estimate; they should not be added to it as a separate load. (EIA)

The buildout is accelerating. Berkeley Lab’s 2026 update models a 2030 reference case of 649 TWh, or 11.8% of U.S. electricity use, with compounded-uncertainty bounds of 521–843 TWh and 9.5%–15.3%. These are modeled scenarios, not measured demand. (Berkeley Lab)

That growth is a mandate to build capacity, not ration computation. Communities should ask who will fund generation and transmission, how rate and curtailment risks will be allocated, and whether storage or private and off-grid systems make technical and economic sense.

Put water numbers in proportion

Berkeley Lab’s 2024 report depicts roughly 65 billion liters per year, or about 47 million gallons per day, of direct U.S. data-center water consumption in 2023. (Berkeley Lab)

That number needs measurement labels and comparison classes. USGS reports that thermoelectric power, irrigation and public supply accounted for 90% of U.S. water withdrawals in 2015. EPA reports that commercial and institutional facilities account for about 17% of public-water withdrawals, with offices, schools, hospitality and healthcare identified as the largest subsectors in that category. (USGS; EPA)

These figures use different years and labels: the data-center estimate is direct consumption, while the USGS and EPA figures concern withdrawals. They should not be treated as interchangeable quantities. A serious local proposal should state its operating assumptions and put cooling, reuse, supply upgrades and customer protections in writing.

The Dalles demonstrates the better approach. Do not treat water infrastructure as an afterthought. Negotiate who builds it, who pays for it and how existing customers are protected.

Work and local capability

Prineville, Oregon, offers another observed record. The city reported unemployment near 20% in 2008, before Facebook selected it for a data center in 2009 and Apple followed in 2012. (City of Prineville)

By the city’s 2019 application period, its industrial area supported 436 full-time jobs and more than 1,500 sustained construction positions, with average compensation of $73,000. The city also reported that information-industry compensation increased from $1.6 million in 2009 to $34 million by the application period. (City of Prineville)

Those are official before-and-after observations, not proof that one project caused every change. They still show why communities should compete for compute investment with a clear plan for training, contracting, infrastructure and long-term local capability.

Welcome the infrastructure. Write the terms.

The practical question is not whether a community should reluctantly tolerate a data center. It is how much durable value the community can build around one.

A strong proposal should address:

• who owns the land, facility and supporting infrastructure;
• who funds power, transmission, storage and water upgrades;
• which tax payments, community funds and infrastructure obligations are binding;
• how hiring, training and local contracting commitments will be measured;
• enforceable terms for noise, setbacks, lighting, traffic and restoration; and
• who pays for modification, decommissioning or remediation.

More compute is coming. Communities should not wait to be discovered.

Singularity Farmer can help assess readiness, package the opportunity, reach credible developers, compare proposals, support negotiations and track delivery. We do not promise that a community will secure a project. Legal, engineering, utility, public-finance and other regulated decisions must remain with the community’s qualified advisers and responsible authorities.