Do Data Centers Actually Create Jobs? Real Numbers

The Short Answer

Short answer: Data centers create many temporary construction jobs but very few permanent ones. A single hyperscale campus can take 2,400 to 5,000 workers to build, yet often runs with only about 25 to 50 permanent staff once it opens.

So the honest picture has two parts. There is a big, short building boom, then a small, quiet workforce that stays.

That gap matters when a town weighs the tax breaks it gives away. The costs last for decades. Most of the jobs do not.

Construction Jobs vs. Permanent Jobs

The biggest job numbers you hear about data centers come from construction, not operations. These are two very different things, and they last for very different amounts of time.

Construction jobs are real and can pay well. But they are temporary by design.

The building boom

Peak crews on a large project run from about 2,400 to 5,000 workers. The Stargate site in Abilene, Texas needed roughly 6,400 workers to build one giant campus.

That surge usually lasts 6 to 18 months at full size. When the building is done, the crews move on to the next site.

Many of these workers travel from other regions. So a share of their pay leaves town when they do.

The operations reality

Once the building is finished, the campus needs only a small crew to run. These are the permanent jobs a town actually keeps.

This is the number that matters most for long-term local benefit. It is also the number boosters mention least. Our guide to data center impact breaks down the full local picture.

How Many Permanent Jobs, Really?

A hyperscale data center typically runs with only a few dozen permanent workers. Industry workforce data points to roughly 25 to 50 permanent staff per 100-megawatt campus.

That is a tiny number for a project that can cost several billion dollars to build. Data centers are among the least labor-intensive structures in the economy.

The reason is simple. The machines do the work, and software handles most of the load.

Why the buildings run so lean

A data center is rows of servers in a locked, cooled building. It does not need a factory floor full of people.

The steady jobs are things like security, facilities, and technicians. A campus can hum along around the clock with a small team on shift.

Some larger sites do reach a few hundred workers. But the biggest, most automated campuses can run with as few as 20 to 30 people per 100 megawatts.

The Brookings Institution puts it plainly. Large projects often promise only dozens to a few hundred permanent workers, while the construction jobs are temporary.

Pay and Who Gets Hired

The permanent jobs that data centers create can pay well, but there are not many of them. So the total payroll a town keeps is small.

Good pay for 40 people does not lift a whole local economy the way a factory of 400 might. The math is about count, not just wages.

Who fills the roles

Some permanent roles need special training in IT or electrical systems. Those hires may come from outside the town.

Local workers often get the security and facilities jobs. Those matter, but they are a small slice of the promise. Meanwhile, the AI layoffs hitting other industries can offset new hiring.

Brookings found that counties with data centers saw wages rise about 3% to 4%. That is a modest bump, not a boom.

The indirect-jobs claim

Boosters often add "indirect" and "induced" jobs to make totals look bigger. These count spending that ripples out to shops and services.

Those effects are real but easy to inflate. Brookings warns that naive estimates can overstate the true gain by a factor of about three.

The Tax-Break Tradeoff: Cost Per Job

To land a data center, states and towns often give up huge amounts of tax revenue. When you divide that cost by the permanent jobs created, the price per job is steep.

Good Jobs First, a subsidy watchdog, found an average cost of about $2 million per job across major data center deals. Some single deals ran far higher.

Real subsidy-per-job figures

In one New York deal, a data center promised 125 jobs for $1.4 billion in support. That works out to about $11 million per job.

An Apple deal in North Carolina reached roughly $6.4 million per job. These are not small rounding errors.

Good Jobs First argues officials should cap data center subsidies at $50,000 per job. Many real deals blow past that limit many times over.

Public officials should be ready to walk away from bidding wars that guarantee net losses for taxpayers, Good Jobs First urges.

Property-tax abatements often last 10, 20, or even 30 years. So the lost revenue keeps growing long after the ribbon-cutting.

Promised vs. Actual: A Side-by-Side Look

The gap between what data centers are pitched and what they deliver is easiest to see in a table. Here is how the promise compares to the reality.

The left column is the sales pitch. The right column is what the data and state audits actually show.

Case Examples: Virginia, Georgia, and More

Real state numbers show the cost-per-job problem clearly. Virginia and Georgia both track their data center tax breaks, and the figures are striking.

Virginia

Virginia is the world's largest data center market. Its own legislative watchdog, JLARC, studied the industry in 2024.

In FY2025, the state added about 1,610 new data center jobs. For that, it gave up roughly $1.9 billion in tax breaks.

That comes to about $1.2 million in lost revenue per new job. The sales-tax exemption alone topped $1 billion in a single year.

Georgia

Georgia expects to lose about $2.5 billion in FY2026 to its data center sales-tax exemption. That is many times higher than an earlier state estimate.

Georgia's own audit found a key catch. About 70% of data center projects would have located there even without the tax break.

Texas tells a similar story, spending roughly $1 billion subsidizing data centers in FY2025. The pattern repeats state to state.

Not all of this is loss. Data centers do pay property taxes and fund construction spending, and Virginia's study found real value in some deals.

But the permanent-jobs return stays small. That is why the fight over these projects keeps growing, as we cover in our look at the AI backlash.

How to Weigh a Data Center Proposal

When a data center is pitched to your town, focus on the permanent jobs, not the construction headline. That single number tells you what stays after the crews leave.

Ask hard questions before any vote. The answers reveal whether the deal is a fair trade.

Questions worth asking

• How many permanent jobs, in writing, not construction jobs?
• What is the total tax break, and the cost per permanent job?
• How many years does the abatement last?
• Would the project come without the incentive?
• Are the job promises legally binding, with clawbacks if they fall short?

Compare the cost per job to the $50,000 cap that watchdogs suggest. If it is far higher, the math likely does not work for residents.

Also weigh other local costs, like power, water, and land. Our guides on data centers on farmland and how to stop a data center can help you dig deeper.

Data centers are not villains, and some deals are fair. But the jobs promise is often the weakest part of the pitch, so treat it with care.