Every data center proposal arrives with a jobs number attached, and that number is almost always the largest, most impressive figure available — typically a construction employment estimate, presented in a way that lets the audience assume it represents permanent, ongoing employment. This essay separates what’s actually true about data center employment from what’s marketing, and it’s worth saying upfront: the honest picture is more nuanced than either side’s talking points usually suggest.
The construction-phase jobs cited in nearly every developer presentation are genuinely real and often well-paying. Microsoft hired more than 3,000 workers to build its two Wisconsin data centers. The first Stargate facility in Abilene, Texas required 6,400 workers to construct its 1-gigawatt capacity. A large-scale data center can employ as many as 1,500 construction workers for up to three years, generating on average $140,000 in total wages and benefits per worker annually — genuinely strong compensation for skilled trades work.
The honest counterpoint, made directly by defenders of the industry, deserves to be stated fairly: construction work is not supposed to be permanent in any industry. Tradespeople move from project to project by the nature of the work, and dismissing construction employment as not “real” jobs misunderstands how the building trades function across the entire economy, not just data centers specifically.
Where the jobs pitch becomes genuinely misleading is in how permanent operational employment gets presented relative to that much larger construction figure. Investigative reporting has found many data centers employ fewer than 125 permanent workers, with some facilities employing as few as 25 — and one developer application in North Carolina showed that out of 69 total permanent jobs, 25 were security and janitorial staff, not the high-skill technical positions typically implied in public messaging.
The reason for this gap is structural, not incidental. Automation and remote monitoring now handle most of what would have required on-site staff a decade ago — the headcount is low by design, not by underinvestment. A facility representing a billion dollars or more in capital investment is simply built to run with a small permanent crew, because that is the entire economic logic of the technology, as covered in Essay #1.
One of the most consistently underreported aspects of data center jobs promises is the timeline attached to them, which frequently gets lost entirely in public discussion. Amazon promised 1,000 positions in Virginia tied to a $35 billion investment that will take roughly 17 years to fully complete. In Indiana, 400 promised jobs are specifically conditioned on reaching “full development” — meaning after an entire 16-facility data center complex is built out, not after the first building opens.
This matters enormously for the tax incentive negotiations covered in Essay #9: a community that grants a tax abatement today based on a jobs number that won’t fully materialize for nearly two decades, if ever, has accepted significant near-term cost in exchange for a benefit that may arrive on a dramatically longer timeline than the abatement itself.
Not every permanent data center job pays the high wages implied by industry messaging. One former data center technician reported earnings as low as $15 per hour — a figure that complicates the narrative of data centers as uniformly high-paying technical employment, even as other roles within the same facility, such as electricians and engineers, genuinely do command six-figure compensation.
Industry-aligned research presents a considerably rosier national picture than the project-level investigations above suggest. A 2025 report from the American Edge Project — a policy advocacy group formed by Meta — projected U.S. data centers would generate 4.7 million temporary construction jobs and roughly 697,000 permanent operational jobs nationally. It’s worth noting plainly that this estimate comes from an organization founded by one of the industry’s largest players, which doesn’t make the number wrong, but does mean it should be read alongside the project-level, independently reported figures above rather than in place of them.
“While developers often highlight construction employment, those jobs disappear once the facility is built, typically within a couple of years.” — Good Jobs First analysis of data center employment claims, March 2026
Even the genuine construction-phase employment doesn’t always benefit the host community the way it’s implied. Due to a nationwide shortage of skilled tradespeople, many construction positions are filled by workers from outside the local community entirely, traveling in for the duration of a build rather than representing local hiring — a pattern one workforce expert described as a “whack-a-mole style spike in job growth, rather than heightened and sustained employment growth” for any single region.
The honest version of this story is neither “data centers create no jobs” nor “data centers are a jobs bonanza” — it’s that construction employment is real, temporary, and often filled by outside workers, while permanent operational employment is genuinely modest by design across virtually every facility, regardless of its size or cost. Communities evaluating a proposed project are best served by demanding the specific, time-bound numbers — not the combined, undifferentiated figure that makes a 50-person permanent operation sound like 2,000 ongoing jobs.