The Case For and the Case Against

People on both sides of a data center fight usually have a real point. The job of a resident, advocate or official trying to decide is to find each side's best argument, check it, and say where it holds and where it does not. What follows gives each case in its strongest form, with the evidence each one actually rests on.

The case for

A large tax base that asks little back

A campus with a billion dollars of buildings and equipment adds a large block of taxable value. It sends no children to school, generates little traffic after construction, and makes few calls on police or social services. JLARC found that in five mature Virginia localities, data center revenue ran from under 1 percent to 31 percent of all local revenue. Counties at the top end have used that money to hold down residential tax rates or fund schools and roads. Even after an abatement, the remaining property tax can exceed what the land paid as farmland many times over.

Grid investment paid for by someone else

New substations and lines built for a data center can strengthen the grid for everyone nearby. Under a full-cost contract, the customer pays. Idaho Power's 2022 application for Meta's affiliate said the company would pay all costs of a 200 MW solar purchase agreement so other customers were not harmed. Ohio's 2025 tariff makes large data centers pay for 85 percent of subscribed capacity for up to 12 years whether they use it or not. A steady customer that pays full cost spreads fixed costs over more sales, which can lower rates for others.

Construction spending and well-paid jobs

JLARC estimated the Virginia industry supports about 74,000 jobs and $9.1 billion in GDP a year, mostly through construction. The 2025 incentive review found 84 jobs per $1 million of exemption, better than the average state incentive. Permanent jobs are few but pay well, and construction brings months of spending at local hotels, restaurants and suppliers. An independent Brookings analysis by Dany Bahar and Greg Wright (updated August 2026), comparing counties that got data centers with counties whose announced projects were canceled, found data processing employment up about 56 percent over the first decade of operation.

The demand is real

The IEA projects global data center electricity more than doubling to about 945 TWh by 2030. Someone will host that load. A community that sets clear terms has more say over how it is built than one that refuses and watches it land in the next county anyway, along with the tax base.

The case against

Residents and farmers may pay for the grid

JLARC estimated a typical Dominion residential customer's generation and transmission costs could rise $14 to $37 a month by 2040 under strong data center growth. Ari Peskoe and Eliza Martin of Harvard's Environmental and Energy Law Program, after reviewing nearly 50 proceedings, argue that utilities have often offered discounted terms and pushed costs onto other ratepayers. In Idaho, Adam Young, a Bingham County farmer writing a February 2026 guest column in the Idaho Capital Sun, reported Idaho Power's farm demand charge rising from $6.94 to $16.50 per kW between 2021 and 2026 and argued that large new loads drive it. Tariffs can protect against some of this, but shared transmission and new generation are hard to fence off entirely.

Water and power are local and finite

A single evaporative-cooled campus can use as much water as thousands of households, and it peaks on the hottest days when cities also peak. Han, Li, Wierman and Ren (2026) estimate national new water capacity needs of hundreds of millions of gallons a day through 2030. A large campus can also equal a sizable share of a utility's current peak, forcing new plants whose costs and emissions last decades.

Incentives cost more than they return

Virginia's exemption returned about 48 cents in state revenue per dollar given up. A University of Georgia study cited by researcher Tommy Turner in an October 2026 Cardinal News opinion piece found only about 30 percent of data centers there could be attributed to the state's exemption, meaning most of the subsidy went to projects that would have come anyway.

Few permanent jobs and long commitments

About 50 operating jobs per building, half contractors, per JLARC. Idaho's exemption requires 30. The same Brookings study that found job gains found no change in county wages and home prices up 2 to 5 percent. Meanwhile the community commits land, water rights and grid capacity for decades, and a PILOT can lock in low rates for 20 years.

The process is often closed

NDAs, code names and shell companies can mean key decisions get made before residents know who the customer is. Tucson's Project Blue became a case study in how that erodes trust.

Question	Factual or values?	What could settle it
Will residential rates rise because of this project?	Factual, with uncertainty	Cost-of-service testimony, contract terms, commission order
Will the aquifer support it on a peak day?	Factual	Hydrology, water right review, peak-day demand figures
Would the project come without incentives?	Factual, hard to know	But-for studies, site-selection records
Is an industrial campus right for this valley?	Values	Public process and elections
Is supporting AI growth a good use of local resources?	Values	Public debate

Where the evidence points

Some things are clear. National data center electricity use is a single-digit share today and growing fast. Most individual sites use modest water, and a few use a lot. Permanent jobs are few relative to investment. Construction drives most economic benefit. In Virginia, the best-studied case, the sales tax exemption returned less in state revenue than it cost, though more than most other incentives. And whether other ratepayers pay depends on contract terms that are public in a commission docket, if you go read them.

Other things are unsettled: how many proposed projects will actually be built, how much load growth the high forecasts overstate, and whether large-load tariffs will hold up in practice over a decade. Treat those as open.