On April 24, Maine’s Gov. Janet Mills signed two bills that pulled in opposite directions. The Democrat vetoed legislation that would have made Maine the first state to impose a moratorium on large new data centers, but also signed a law that will exclude data centers from the state’s business development tax incentives.
The veto preserved a $550-million data center at the former Androscoggin paper mill in Jay, a Franklin County town that lost roughly 22% of its property tax base when the mill closed in 2023. Mills cited 800 construction jobs, 100 permanent jobs, and substantial property tax revenue for Jay as her rationale. The legislature’s attempt to override her veto failed five days later by a margin of 72 to 65.
The Maine fight is one version of a debate now underway in 30 other state legislatures, where much of the public argument over data centers comes down to whether the jobs they promise are real. However, this is a question that is rooted in a misunderstanding of the true job opportunities presented by a new data center.
The actual economic development that data centers generate does not live inside the data center itself. It lives in the energy infrastructure built to power it.
Measuring jobs per dollar
To compare investments across sectors that pay people very differently, the standard frame is direct full-time-equivalent (or FTE) positions per dollar of capital deployed. FTE is a labor accounting unit, not a headcount. One permanent FTE could be one person working full time, two people splitting a position across shifts, or any pattern of contract and salaried work that nets to a year of full-time labor. The distinction matters — even though press releases, political announcements, and even this article often use “jobs” to mean FTE.
Each bar on the chart represents a hypothetical 1-gigawatt U.S. project at industry-benchmark costs and staffing assumptions. Any single project will land above or below the figures shown depending on automation, contracting, and region, and the data center capital cost reflects fully fitted-out construction excluding IT equipment, which depreciates on a five-to-seven-year refresh cycle and is not directly comparable to long-lived generation infrastructure. The chart is built to show the size of the gaps between technologies.
As compared with energy projects, data centers yield significantly fewer jobs because hyperscale facilities are designed to run with very few people, and most of the capital cost is hardware that gets replaced every five to seven years rather than long-lived infrastructure that requires operating crews.
Industry benchmarks put permanent staffing at the most automated data center campuses at roughly 25 to 40 operators per 100 megawatts. A 1-GW data center at $11 billion in capex generates fewer permanent FTE per dollar than any form of large-scale energy generation project.
Offshore wind is the inverse. More of an offshore wind project’s capex pays for human labor than almost any other energy investment in the U.S. Vessels need crews. Port logistics require longshoremen, riggers, and dockside fabrication. Foundation installation, blade assembly, and cable-laying are labor-heavy at scale and have to happen in U.S. waters with domestic content requirements that pull more of the supply chain onshore than land-based wind.
Dominion Energy’s 2.6-GW Coastal Virginia Offshore Wind project, the largest offshore wind farm under construction in U.S. waters, is on track for completion at the end of 2026 at a total cost of $11.5 billion. Dominion cites 2,000 direct and indirect American jobs and $2 billion in economic activity. The labor density of those projects makes the Trump administration’s decision to pay developers nearly $2 billion to walk away from U.S. offshore wind leases all the more striking.
Onshore wind paired with storage leads the permanent FTE figures because wind projects run for 25 to 30 years, and the maintenance regime supports steady regional employment over that lifetime through turbine inspections, blade repair, gearbox servicing, and balance-of-plant work.
Combined-cycle gas, by contrast, requires roughly one-third the permanent crew for comparable capacity, which is why the political story that gas is a pragmatic economic development workhorse does not survive the comparison to renewables, at least in terms of jobs created.
Data center jobs are energy jobs
Mills cited 100 permanent positions at a $550-million project, which works out to a higher permanent-FTE figure per dollar of capex than any other technology on the chart. No other U.S. data center has come close to that number.
A Vantage Data Centers facility outside Reno disclosed 73 permanent positions for a 1.1-million-square-foot site in 2024 business records. Meta’s $10-billion Lebanon, Indiana, campus will employ roughly 300 people once complete. Amazon’s $35-billion Virginia announcement promised 1,000 permanent positions phased across 17 years.
The Jay project may genuinely deliver its claim, however, because the developer has bundled 150 MW of on-site solar and existing on-site hydroelectric power into the project, with a maximum 25-MW draw from Central Maine Power on cloudy days. The 100 permanent positions Mills cited would include the operations and maintenance jobs the bundled solar and hydro assets create.
Strip out the on-site solar and the hydroelectric integration, however, and what remains is a roughly 250-MW data center load that will likely yield 25 to 40 permanent operators, consistent with hyperscale averages and dramatically below the figure Mills cited.
When generation is bundled into a data center announcement, the labor figures reflect the full picture. When generation is procured separately through utility power purchase agreements that surface in integrated resource plans months or years later, the FTEs that the data center load is actually creating get attributed to other projects, other entities, and other press conferences. Meta’s Richland Parish campus is being served by 2.3 GW of new gas generation that Entergy is building in Louisiana explicitly for the project. The construction and permanent FTEs from that buildout do not appear in Meta’s announcement; they appear in Entergy’s.
Ultimately, the economic development case for a data center is not the facility itself; it’s the energy buildout — and what kind of resource it chooses matters. The data center itself contributes the same small number whether it draws on renewables or fossil gas, but the former creates three times as many jobs as the latter. The resource choice determines the labor outcome.
This reframing changes which deals are worth doing and on what terms. Communities offering tax credits to attract data centers without conditioning those credits on the generation that will serve them are giving away the case for the deal. The data center alone, plugged into existing grid capacity, delivers the thinnest possible version of economic development: construction trades, property tax revenue, and a small operating staff.
The most significant benefits require that the data center bring its energy with it — and that the energy be renewable to produce real long-run labor.
