A company announcing a multibillion-dollar data centre investment is making a statement about capital expenditure, not automatically about the benefit received by the host community. A portion may reach local construction firms and workers, while another portion pays for imported servers, electrical equipment and specialist contractors. The total investment figure can be accurate and still be a poor measure of what stays in the local economy.
Employment is where this gets particularly confusing. Construction may involve hundreds or thousands of people over a period of months or years. Once the site is running, a much smaller permanent team may be required, although those positions can be highly skilled and well paid. Both types of employment matter, but they should appear on separate lines with dates, rather than being combined into one impressive number. Claims about indirect jobs also need a methodology and an explanation of where those jobs are expected to occur.
Data centres can deliver other benefits: tax revenues, purchases from local suppliers, better digital infrastructure and training partnerships. They can also bring expenses or local effects, including utility upgrades, pressure on water supplies, construction disruption and sound from mechanical equipment. The people gaining from the investment may not be the same people who live nearest the buildings. That isn't a verdict against the project. It's a reason to make the comparison more precise.
Our research should pair the economic forecasts used during approval with measured results after opening. Permanent local payroll, local procurement, tax receipts and promised training programmes belong beside public incentives, utility cost allocation and environmental monitoring. We also need to keep track of which claims remain estimates. Only then can a community decide whether the arrangement delivered what was promised.
Loudoun County and Alberta: what can be measured?
Loudoun County, Virginia is a mature comparison market, not a forecast for Alberta. The county reports approximately 233 existing data-centre buildings, built or at various stages of construction, as of March 1, 2026. This is a building count and must not be presented as 233 fully operational facilities. For FY2026, the county reports US$1.2 billion in real and personal property tax revenue from data centres, equivalent to 39% of its total budget. Its reported 17,000-plus data-centre-related jobs include both direct and indirect employment, rather than permanent onsite jobs alone.
In reporting published November 5, 2026, CTV News Edmonton interviewed Loudoun supervisor Juli Briskman about local noise, air quality, land-use and end-of-life obligations. These are attributed concerns, not a finding that every centre has the same impact. Alberta officials said, as CTV reported, that existing legally binding conditions cover air quality, noise and light pollution, while the province generally does not require a standalone environmental impact assessment for a data centre; its power and water components can face separate regulation. The article also reported the Pembina Institute's estimate of potential electricity-sector emissions associated with Meta's proposed Sturgeon County power plant, not measured operational emissions.
A 2026 analysis of 1,584 US data centres by Simon Buechler and Lyndsey Rolheiser found approximately 1,000 to 1,300 additional nonconstruction jobs in host counties relative to comparison counties three to five years after initial openings. After accounting for later capacity growth, the researchers estimate roughly 2.7 additional nonconstruction jobs per eventual megawatt of capacity. These are county-level associations, not direct hires or a guaranteed effect at an individual project. The sample ends in 2023 and cannot be assumed to represent new AI megacampuses or sites in Alberta.
What the local-benefit scorecard must disclose
For each project, distinguish announced, approved, under construction and operating status. Record permanent onsite employment separately from construction jobs and modelled indirect jobs; report tax receipts actually collected separately from projected revenue, and incentives separately from private investment. Track the utility interconnection agreement, contracted versus metered MW/MWh, who funds grid upgrades, water withdrawal limits versus actual water use, and measured noise and air pollutants. Identify the presence or absence of an independent environmental and public-health assessment, including underlying assumptions and uncertainty ranges.
Finally, record zoning setbacks, complaint and enforcement history, and decommissioning responsibility: closure/remediation plan, estimated cost, financial security or bond, and the entity liable if an operator fails. Missing information must be marked not disclosed, never scored as proof of harm or compliance. Comparable reporting requires a measurement date, jurisdiction, source, status and the distinction between projected and observed values.
Sources: Loudoun County official data-centre overview; CTV News Edmonton, November 5, 2026; The Conversation, October 27, 2026.
Local consent: 2026 US ballot measures
On November 3, 2026, voters in ten US states decided 41 local ballot questions affecting data-centre development. Ballotpedia reported on November 6 that 40 of the 41 results favored tighter development conditions or rejected a question that would facilitate a project. Ohio accounted for 21 measures. Results were unofficial pending certification. The GridSight's individual-measure tracker provides jurisdiction-level descriptions.
These are not 40 enacted bans. The total includes binding full or partial bans, temporary moratoriums, zoning or voter-approval provisions, requirements for financial guarantees, and nonbinding advisory questions. Of the 41 November votes, 13 were questions that would have helped a project; all 13 failed. Harwood, North Dakota rejected an added industrial review requirement, the lone result counted against limits. Six earlier 2026 local ballot measures also favored limits, making 46 of 47 for the year through November 3. This is a selected set of communities holding votes, not a representative national opinion poll and not evidence that every project was rejected.
For project research, track the exact question text, legal effect, result certification date and jurisdiction; classify binding versus advisory, outright ban versus threshold-based restriction, and permitting versus financial assurance. Note the development stage and any existing approvals. In Perry, Ohio, voters approved a financial-guarantee requirement tied to special tax incentives, a useful example of community cost protections that does not itself prohibit all data centres. Compare these policy choices with tax receipts, local jobs, grid costs and environmental findings rather than attributing voter motivation without survey evidence.
These explainers separate announced plans, recorded outcomes and unresolved questions. Find individual facility sources and submit corrections in the Observatory.
General background: IEA, Energy and AI. For site-specific claims, consult each facility's source records.