Christopher’s reading: labor is the fourth constraint

The Workforce Economics of Data Center Development

Why labor capacity is becoming a critical execution risk — a transparent, public-data workforce exposure index for pre-commitment site screening at the labor-shed level.

Power, land, and capital get the diligence. A Lab working paper argues regional skilled-labor capacity belongs beside them — and shows how to screen it from public data before capital is committed.

CN monogram
Christopher Nelson
Workforce Intelligence
Key finding

In the most exposed labor sheds, skilled-trade capacity behaves like infrastructure, not a variable cost: no wage clears the market on the project’s timeline, and the warning signs are visible in public data before the schedule slips.

Christopher Nelson’s executive commentary on The Workforce Economics of Data Center Development (WIL-WP-2026.2-DCX), published by the Workforce Intelligence Lab in July 2026. The underlying research, its data, and its methodology are the Lab’s; the reading below is his.

Every data-center site-selection deck I have seen scores three constraints: power, land, and capital. The working paper I prepared for the Workforce Intelligence Lab, The Workforce Economics of Data Center Development, argues for a fourth — regional skilled-construction labor capacity — and, more usefully, specifies a way to screen it from public data before capital is committed. It is a working paper, and it says so on every page. What it claims, and what it does not, are both worth being precise about.

The problem national numbers hide

In aggregate, the electrician picture looks manageable. The Bureau of Labor Statistics projects about 81,000 average annual openings for electricians over 2024–2034 and a net employment gain of 77,400 over the same decade (BLS Occupational Outlook Handbook). Those figures collapse at the level of a single campus. Peak crews on gigawatt-scale sites have grown from roughly 750 workers five years ago to 4,000–5,000 today, with documented sites reaching 6,000–8,000 (IBEW project disclosures and regional press, 2025–26). For electrical work alone, one campus can require two to four times a local union’s entire membership (IBEW, The Electrical Worker, June 2025).

That is the mechanism the paper formalizes as the union-multiple: demand from one project measured against the auditable membership of the local that has to supply it. It is why a labor shed — the metro area containing the project cluster — is the right unit of analysis, and why national aggregates are the wrong one.

Making the constraint screenable

The paper’s contribution is a workforce exposure index: a transparent 0–100 composite of five weighted, public-data components — availability deficit (30%), hiring competition (25%), compensation pressure (15%), project activity (15%), and a pipeline constraint built from workforce age distribution and apprenticeship throughput (15%). The composite maps to fixed bands: Low, Moderate, Elevated, High, and Severe. On the paper’s Interim-Mode computation from public proxies, Northern Virginia reads 89.2 (High), Dallas–Fort Worth 81.1 (High), and Phoenix 67.4 (Elevated), each with the arithmetic disclosed in full.

Two features matter more than the numbers. First, the shed rule is stated and applied uniformly; the paper explicitly retires an earlier "Central Texas" score that combined Abilene and Dallas–Fort Worth — roughly 180 miles apart — because a score computed on an incoherent geography is worse than no score. Second, every value is presented as a band and a direction, never as a spot forecast.

What the evidence shows, and what it does not

The paper’s strongest finding is qualitative and retrospective: on public proxies, Northern Virginia’s exposure was already moving into the High band well before 2026, as IBEW Local 26’s membership doubled to about 14,700 over seven years and apprentice intake rose to 615 in 2025 from 510 (Northern Virginia Technology Council, February 2026). A market does not double a skilled local in seven years unless demand is outrunning native supply. Compensation pressure is likewise observable: data-center construction pay runs up to roughly 30% above typical construction work (ITIF, January 2026, reporting a Wall Street Journal finding), and travel packages and per diems are standard practice.

Now the limits, which the paper states more bluntly than most research I read. It is a method-and-protocol contribution. It demonstrates that the index is internally consistent, arithmetically reproducible from its published sub-scores, and robust in banding to plausible re-weightings. It does not establish predictive validity, and none of its hypotheses is statistically tested. The component sub-scores are structured analyst assessments anchored to public indicators, not yet computable from them, because the transfer functions are unpublished. And one claim from the June 2026 release — a developer cutting data-hall count because of labor — was retired in the July verification audit when no attributable public source could be found. The absence is itself a finding: owners do not publicly attribute scope changes to labor, so the outcome this field most needs to observe is largely hidden in public data.

Why I think it still changes the decision

The value of a screen is not that it is a forecast; it is that it triggers diligence in the right place at the right time. Read that way, the paper supports four practical moves.

  • Screen the labor shed before commitment, with a stated shed definition, alongside the power and interconnection studies.
  • Engage contractors and local trade organizations early. IBEW job-call pages are a public, current signal of clearance conditions and offered packages.
  • Monitor direction, not just level: negotiated scale changes with their effective dates, backlog months, union-multiple ratios on auditable membership, and cross-market migration.
  • Treat every published labor figure as a dated observation. The paper’s own audit found the most common defect in this literature is not error but staleness and wrong-edition attribution.

The reframing

In the most exposed sheds, no wage clears the local market on the project’s timeline. The operative limit is the number of qualified, licensed workers a shed can supply on schedule — a quantity that behaves like infrastructure capacity, not like a commodity price. Where labor behaves like infrastructure, it can be assessed, secured, and built ahead of demand rather than priced at the margin. That is a modest reframing with a large consequence for how capital gets allocated, and the paper proposes it for testing rather than asserting it as proven. I would rather publish the protocol and be held to it than publish a number and ask to be trusted.

Find the record in the Lab library →

Sources & methodology

Workforce Intelligence Lab. “The Workforce Economics of Data Center Development.” WIL-WP-2026.2-DCX, v2.1. Published July 2026. https://wilinstitute.org/library

Figures cited in this commentary

U.S. Bureau of Labor Statistics. “Occupational Outlook Handbook — Electricians.” 2024–34 projection round. Projection ~81,000 average annual openings; +77,400 net employment over the decade.
IBEW project disclosures; regional press. “Peak crew sizes on gigawatt-scale campuses.” 2025–26. Reported figures ~750 five years ago → 4,000–5,000 today; documented sites 6,000–8,000. Vendor figures excluded.
IBEW. “The Electrical Worker.” June 2025. Union estimate Single-campus electrical demand at two to four times a local’s membership.
Workforce Intelligence Lab. “WIL-WP-2026.2-DCX, Rev. 2.1 — workforce exposure index.” July 2026. Interim-Mode index · illustrative, directional Northern Virginia 89.2 (High); Dallas–Fort Worth 81.1 (High); Phoenix 67.4 (Elevated). Not a validated forecast.
Northern Virginia Technology Council. “IBEW Local 26 membership and apprentice intake.” February 2026. Reported figures Membership ~14,700 (doubled over seven years); apprentice intake 615 in 2025 vs 510 prior.
ITIF (reporting a Wall Street Journal finding). “Data-center construction pay premium.” January 2026. Secondary report Up to ~30% above typical construction work; original paywalled.

How to cite this commentary

APA Nelson, C. (2026, September). The Workforce Economics of Data Center Development. Christopher Nelson — Workforce Intelligence. https://www.cnels.com/research/labor-is-the-fourth-constraint/
Chicago Nelson, Christopher. “The Workforce Economics of Data Center Development.” Christopher Nelson — Workforce Intelligence, September 2026. https://www.cnels.com/research/labor-is-the-fourth-constraint/.

All formats, plus the Lab record

This commentary interprets The Workforce Economics of Data Center Development, published by the Workforce Intelligence Lab in July 2026. The underlying research, its data sources, and its methodology are the Lab's; this piece is Christopher Nelson's reading of what it means for executive and investment decisions. The original report is available at https://wilinstitute.org/library.