A Data Center’s Tax Windfall Can Leave Neighbors Paying

New Albany, Ohio, offers a revealing test of the data-center bargain. The city reports roughly $10 million in community development charges from data-center projects for tax year 2024, collected in 2025. Meanwhile, neighbors of Google’s campus have been reporting cooling-fan noise since September 2025. The city’s current update schedules a permanent solution for completion in December 2026. Revenue has arrived; relief for those neighbors is still being promised. (datacenters.newalbanyohio.org¹)

That gap is where I would start judging a proposed development. A project should have to pass two separate tests: whether its financial arrangement benefits existing residents, and whether its location protects the people closest to it. Municipal receipts cannot excuse an unresolved neighborhood burden. Ordinary taxation and developer-backed payment for additional infrastructure should be the baseline; concessions deserve approval only when officials can demonstrate that residents receive something better in return.

The fiscal case for data centers is real. Loudoun County, Virginia, taxes their computer equipment at its general business personal-property rate, currently $4.15 per $100 of assessed value. The county credits data-center receipts with funding services and capital projects while allowing lower residential property-tax rates. Its adopted fiscal 2027 budget also maintains vehicle-tax relief. This is a meaningful route from server investment to household benefit, although a lower tax rate does not by itself establish a lower bill for every property owner. (loudoun.gov³)

New Albany uses a different mechanism. Its tax abatements, temporary reductions in taxes otherwise owed, come with minimum annual revenue requirements. Companies can satisfy those requirements through taxes and other designated payments, then make a cash payment to cover a shortfall. The city benchmarks the obligation against payroll from comparable office or manufacturing developments and reviews performance annually. Some community-charge revenue services infrastructure debt; the remainder supports public amenities. This is a more serious bargain than offering a discount against vague employment promises. (datacenters.newalbanyohio.org¹)

Employment still needs honest labeling. New Albany’s 2024 financial report described Meta’s campus as having 200 employees, with expansion expected to bring the total to about 300. It separately projected construction staffing rising from 1,200 to 1,800. Those are different benefits with different lifetimes, and the report did not establish that the future operating jobs had materialized. (ohioauditor.gov⁴)

Nor do the available financial disclosures settle whether the incentive was necessary. New Albany’s audited statements report approximately $1.45 million in city property-tax revenue forgone through its abatement programs in 2024, but that figure covers participating properties generally, not data centers alone, and does not represent every affected government’s losses. It cannot simply be subtracted from data-center community charges to produce a net-benefit figure. Virginia’s legislative auditors likewise documented $928 million in state sales-tax savings for the industry in fiscal 2023, a cost falling beyond any single host county’s accounts. (ohioauditor.gov⁴)

I would require a project-level reconciliation: promised and collected receipts, ongoing jobs, public-service spending, infrastructure debt and concessions across governments. Receipts already reduced by an abatement should not have that same concession subtracted twice. Then comes the harder comparison: whether the project would proceed without a discount, and whether another feasible use of the land would serve residents better. Keeping a subsidized promise cannot establish that the subsidy bought additional investment.

Electricity introduces another boundary mismatch: the host government collects revenue, while customers across a utility’s territory can bear investment risk. Ohio’s response provides evidence that financial commitments change the development pipeline. On February 13, 2026, AEP Ohio reported that preliminary data-center interest exceeding 30,000 megawatts had narrowed to 13,022.7 megawatts requesting paid engineering studies, then to 5,642 megawatts accepting binding agreements. That is the utility’s account, not an independent finding of household savings. Still, the falloff shows why expressions of interest should not be treated as bankable demand. (aepohio.com⁶)

AEP Ohio’s special rate schedule requires reimbursement of all buildout costs for specified cancellations or long delays before energization. It also imposes minimum payments under an initial contract lasting eight years beyond a ramp-up period of up to four years. But the pre-service reimbursement obligation ends when the project is completed. Financially qualified customers can avoid collateral; otherwise, the required guarantee or collateral equals half the contract’s minimum charges. These protections are substantial, but the formula alone does not demonstrate recovery of every unrecovered infrastructure dollar. (aepohio.com⁷)

I would require regulators to test cancellation, default and sustained underuse before approving construction. Security should track the remaining project-caused exposure, including investment whose useful life exceeds the contract. Households should not become the fallback payer for capacity built on an investor’s forecast.

There is a credible objection: large customers can spread an existing grid’s fixed costs over more electricity sales, lowering average rates. An October 2025 Berkeley Lab and Brattle analysis found that 10% statewide demand growth during 2019–2024 was associated with roughly 0.6 cents per kilowatt-hour lower average retail prices. But the researchers warned that expensive, constrained expansion could reverse that relationship. (eta-publications.lbl.gov⁸)

That evidence supports welcoming useful demand, not automatically charging residents for its expansion. Virginia’s December 2024 legislative review modeled an additional $14–$37 in monthly generation-and-transmission costs for a typical Dominion residential customer by 2040 because of data-center growth, in constant 2024 dollars. This was a scenario-based forecast, not an observed increase. My rule would allow shared investment only for independently demonstrated benefits to other customers, measured against the least-cost plan without the project. (jlarc.virginia.gov⁵)

Housing evidence makes the neighborhood test more urgent, while resisting a universal verdict. Realtor.com’s August 2026 analysis compared 43 ZIP codes receiving large data centers with matched locations and found no meaningful initial sale-price advantage. Its follow-up was short, its geography broad, and it excluded states without public sale-price disclosure. (realtor.com⁹)

A September 2026 Berkeley working paper used a different comparison: transactions near newly announced large facilities versus similar areas where facilities were announced later. It estimated average housing-value losses of 3% within three miles over five years, rising to 7.2% within one mile. As a working paper, it warrants scrutiny rather than treatment as settled fact. Nevertheless, its distance-specific design offers evidence that broader neighborhood averages can miss concentrated losses. Neither study supplies a universal setback distance or substitutes for examining local zoning changes and preexisting price trends. (haas.berkeley.edu¹⁰)

The practical response should begin before operation. I would require buffered industrial siting, independent sound measurements under full operating conditions, public water-use records, generator operating and emissions records, and funded remedies tied to enforceable deadlines. Virginia’s auditors found that persistent low-frequency noise could disturb residents without violating conventional noise ordinances. New Albany says it has needed no additional Columbus water or sewer capacity during its hyperscale expansion, but that system-level statement cannot certify each facility’s consumption or environmental compliance. (jlarc.virginia.gov⁵)

These records do not provide a complete, independently audited comparison of net household outcomes across the two locations. They establish valuable revenue mechanisms, meaningful utility safeguards and unresolved neighborhood exposure. I would approve projects that close those gaps, not presume every data center fails. Back in New Albany, the promised Google remedy has a December 2026 completion date. A sound development bargain would make residents’ protection an enforceable condition of operation, rather than something they must wait for after the public money begins arriving. (datacenters.newalbanyohio.org²)