What a data center really does
to a local economy

“How many jobs did the data center bring?” is the first question every county board asks and every operator answers with a press release. Here is what the official statistics actually say — measured across nine host counties with the employment surface we just widened, every figure one call from a SHA-256-verified BLS row.

The number you’d ask for is blank

BLS publishes quarterly employment for the data-center industry (NAICS 518210). But when one operator dominates a small county, confidentiality rules withhold the cell — and that is precisely the situation in the rural counties where hyperscale campuses land. In Mecklenburg County, VA (Microsoft’s Boydton campus, building since 2010) the operator’s headcount has been withheld in all 48 quarters since 2014. Same for Amazon’s two Oregon counties. Loudoun — urban, many operators — shows all 48.

source BLS QCEW county files · as_of 2025-10 (2025Q4 vintage) · sha256-verified raw · bls.gov ↗

The public argument runs on a number that, where it matters most, does not exist. An absence, to be clear, that we serve as an absence — a withheld cell is never a zero.

So measure the place, not the industry code

A data-center buildout hits a county as two different shocks. Construction is big and temporary — a hyperscale campus runs on the order of a thousand workers for years. Operations is small and permanent — tens to a few hundred staff. The two have opposite signatures, and conflating them is why the debate talks past itself: the industry cites construction, critics count operations, both are right about different phases.

So we measured everything instead: total covered employment, every wage dollar, construction, and the local-service sectors, for nine host counties against their own states, 2014→2025 — the place-based series the employment API now serves for every US county.

Jobs: there is no consistent boom

Total county employment growth, 2014→2025 annual averages, county vs its own state:

source BLS QCEW county files · as_of 2025-10 (2025Q4 vintage) · sha256-verified raw · bls.gov ↗

Four counties beat their state, two track it, two lag it. Mecklenburg — fifteen years of Microsoft buildout — grew +4.5% while Virginia grew +13.3%. Berkeley County’s +62% is real but confounded (Volvo’s plant and metro-Charleston growth arrived in the same window); Loudoun’s +38% is all of Data Center Alley’s tech economy, not a single campus. Read across all nine: hosting a hyperscale campus does not reliably make a county out-grow its state in headcount.

Pay: the signal is consistent

Weekly wages tell a different story — computed from total quarterly payroll over average monthly employment, the same way BLS does:

source BLS QCEW county files · as_of 2025-10 (2025Q4 vintage) · sha256-verified raw · bls.gov ↗

Seven of the eight rural host counties beat their own state’s wage growth. Mecklenburg is the cleanest case: jobs up 4.5%, but the average weekly wage went from $590 — 10% below its Southside-Virginia neighbors — to $1,044, crossing above them: +77% against Virginia’s +51%. A county of 30,000 absorbed a multi-billion-dollar campus, and the result was not more jobs. It was the same number of jobs paying more. The one urban contrast inverts: Loudoun’s wage growth trails Virginia — it already sat at the top of the pay scale ($1,674/week today).

Why construction barely shows up

The thousand-worker construction phase is mostly invisible in the host county’s statistics — not because it didn’t happen, but because of how the data counts: QCEW books a job at the employer’s establishment, not the work site. A crew building in Boydton for a Richmond-based contractor counts in Richmond. Mecklenburg’s construction sector moved by roughly +150 at its best through billion-dollar build years. The exception proves the rule: in tiny Morrow County, OR (592 construction jobs at the 2015 peak, 57 by 2020 — a pulse and collapse), the county was small enough, and the hiring local enough, for the shock to be legible.

source BLS QCEW county files · as_of 2025-10 (2025Q4 vintage) · sha256-verified raw · bls.gov ↗

The question the data poses

Not house opinion — just what these series ask next: if the durable local effect of a hyperscale campus is wage level rather than job count, then the right fiscal question for the next county board isn’t “how many jobs?” — it’s “whose pay rises, and does the tax base capture it?” The data to answer that is public, quarterly, and now one API call away for any county in America.

Method. Counties chosen for hosting hyperscale campuses with public operator histories; each compared to its own state as the naive baseline. This is descriptive — no matched-control causal claim — and two structural caveats apply everywhere: jobs count at the employer’s establishment (construction understated on-site), and NAICS was revised at 2022Q1 (the 518210 industry definition changed; our API flags this). All series: BLS QCEW, quarterly, 2014Q1–2025Q4 vintage, national/state/county.