The blue-collar AI job market is booming. Will data center backlash make it go bust?

When most people think of AI and jobs, the conversation often pivots to which occupations are most likely to be obliterated by its arrival. But, at least for some professions, the AI boom has been an employment one too.

AI is creating jobs in a variety of trades that prop up the sprawling ecosystem that powers giants like Anthropic, Meta, Amazon, OpenAI, and others.

The construction and maintaining of data centers and the entire AI ecosystem is requiring a new generation of HVAC technicians, construction workers, line workers, and pipefitters.

“These are occupations that are becoming increasingly important to the AI economy,” said Maria Flynn, president and CEO of national nonprofit Jobs for the Future. While data centers are the part that gets the publicity, they are only one piece of a much larger puzzle, requiring upgrades in transportation and modernization of the energy grid. “We are seeing major sources of demand that are converging at the same time,” Flynn said.

The skilled jobs often come with attractive pay. Flynn said an apprentice level technician can take home $40,000 to $60,000, with experienced electricians commanding north of $100,000.

Nicole Bachaud, labor economist at online employment marketplace ZipRecruiter, is seeing salaries even higher for more specialized, high skill roles. Bachaud said that the mean minimum salary for data center jobs spiked by 125.1% year-over-year to nearly $208,000.

“This is suggesting that highly specialized, top-tier engineering roles are pulling the overall average up drastically,” Bachaud said.

While many traditional blue-collar workers are seeing a boost, including truckers, the biggest blue collar beneficiaries, according to Bachaud’s data, are welders and pipefitters. Postings for those jobs are up 164% year over year. While that could be partly the result of a small sample size as more data centers are built, even if each one only needs a few pipefitters, the demand for these roles will increase, she said. The construction demand is also uneven with some places getting in on it more than others. Houston and Birmingham have seen particularly robust growth.

“Places with more land to expand and build, which are often places where building regulations make development easier, cheaper, and faster, will likely be where expansion continues, versus coastal hubs that have higher costs and more regulatory burdens,” Bachaud said.

That means an electrician or pipefitter in Birmingham will probably have more of a buyer’s market than one in Boston.

None of this surprises Justin Sinkovich, associate professor and associate director of the School of Business and Entrepreneurship at Columbia College Chicago. In his view, the usual AI jobs narrative of displacement is incomplete, with the data center counter-story as one example.

While AI may automate some work, its deployment depends on the buildout of infrastructure, including not only data centers but also substations, cooling systems, water, and piping infrastructure, which AI cannot build.

“Positive effects on the labor market are already visible, with an increased demand for cooling and HVAC engineers, industrial automation technicians, and traditional trade workers like construction workers and electricians. HVAC-engineering wages have risen, and many technical professionals entering data-center roles are seeing a pay premium,” Sinkovich said.

He pointed to Louisiana as a specific example, where both Amazon and Meta have massive projects.

“Amazon committed $12 billion to a new data center with 540 new on-site jobs, including 1,700 electricians, technicians, and security personnel. Meta’s Hyperion project in Louisiana is $27 billion,” Sinkovich said. “These projects create demand for on-site workers and therefore cannot be offshored or conducted remotely, unlike manufacturing and software,” he added.

He also said there is a longer-term story of AI enabling more decentralized and dynamic energy systems. This not only improves energy efficiency but also creates additional work to build smart controls, battery systems, microgrids, and distributed energy infrastructure.

Economic optimism vs. public skepticism

Recent research from global real estate services firm Cushman & Wakefield indicates that every 100MW of new data center development creates nearly 1,300 jobs in the local economy, generating approximately $110 million in annual wages, $344 million in gross output, and $187 million in gross regional product. And many of these jobs aren’t ones requiring a diploma. In fact, many of them are better suited to a path through trade school.

But the optimistic shovels on the ground analysis don’t match the national mood. Seventy percent of Americans oppose a data center being built in their local area, according to Gallup, and a New York Times/Siena poll conducted this month found roughly two-thirds opposed regardless of party — Democrats, independents and Republicans alike.

Even where data centers are proceeding at a robust clip and creating the most jobs public opinion has turned against them. Fifty-seven percent of voters oppose one in their own community, according to a Texas Politics Project poll, enough that some experts see Democrats gaining ground in the Lone Star State. According to Data Center Watch, at least 75 data center projects worth roughly $130 billion have already been blocked or delayed this year as the backlash spreads. 

Texas Gov. Greg Abbott, currently up for reelection, recently issued a data center grid approval moratorium and halt on new environmental permits. New York Governor Kathy Hochul recently implemented the first statewide ban on new data centers.   

The dynamics of the situation prompted Oracle this week to file a “force majeure” (a stipulation that a project may be delayed due to extraordinary circumstances) notice on its New Mexico data center project, part of the broader Stargate AI buildout, seeking to delay payment if the facility isn’t operational by 2028. The project has faced local opposition ahead of the midterms and environmental concerns, though Oracle says it remains on track.

Shenandoah University professor Michael Hicks, who has studied the economic impact of data centers, says that the public mood is often the result of the mismatch between what data centers promise and their reality.

“Most of the backlash is the result of developers being untruthful with local officials about potential effects, developers have woefully overpromised jobs and given tax breaks that are beyond obscene,” Hicks said.

He added that he believes studies like the ones produced by Cushman & Wakefield and others use modeling with the rosiest possible scenarios, whereas he has studied actual on the ground impact from data centers going back to the early 90s.

“Permanent labor market effects are muted. There are very few that have substantiative numbers beyond the early build out phase,” Hicks said.

Recent research from Brookings Institution found that there is a local economic impact from data center construction, but it can be muted, and it can greatly depend upon the type of data center being built — a massive hyperscaler facility versus more typical colocation data centers owned by real estate developers and where data services are being shared across customers. Brookings stressed that while the data center debate is moving fast, on the economics, the “evidence base has not kept pace.”

Hicks is a fan of data centers, but not because of jobs. He said if data centers are taxed at the same rate as households they are generally going to cause an enormous increase in taxable property for a community. He said that surplus can then be used to cut residents property taxes, citing Loudon County, Virginia where average property taxes have dropped by $3,400 since 2020 due to the huge data center build out. Or instead of cutting taxes, the additional revenue can be used to increase community amenities and services that improve quality of life that outweigh the noise and grid strain.

At the same time, some states are also lavishing tax breaks on data centers that some say are both undeserved and ultimately, bad economics.

How a small town and family business in Georgia have benefited

But observations on the ground do bear out the current labor boom.

Trey Travis, vice president of operations at industrial hose and expansion joint manufacturer Southeastern Hose in Bremen, Georgia (population 7,000), is the third generation at the helm of this family-owned company. AI was science fiction when the company was started by his grandfather in the 1960s. Now, AI is their booming reality.

Southeastern Hose’s expertise is in fluid conveyance and thermal expansion. Travis said their products are essential for managing temperature changes and preventing catastrophic failures in rigid piping systems. In simpler terms, his company produces the plumbing for the plumbing, the obscure pieces of the data center.

“If there is fluid moving anywhere or a change in temps, there will be a hose or expansion joint in the piping system somewhere,” Travis said.

He compared his products to the “missing screw” that might come up in a home DIY project, something obscure – but crucial – to a home project that will send you back to the hardware store if you don’t have it because the project can’t proceed without it.

The market for the piping and joints made by Southeastern Hose is booming now, giving Travis’s tiny Georgia town an economic lifeline.

“We’ve grown pretty significantly over the last 18 months,” he said.

The company’s 150-person workforce has ballooned by 40 percent during that time.

“We’ve got a pretty vast relationship with some of our local trade schools and high schools for work-based learning programs,” Travis said of their ability to hire workers. He said workers can get all the overtime they want, make plenty of money, and still have a family life.

Some of the skills needed to make data center components are even outrunning the trade schools, like the specialized welding skills needed.

“If I get a top notch welder from the local trade school, I still have on-site training because they’ve never seen this type of welding on this type of product before,” Travis said.

None of this surprises labor economist Michael Rosenbaum, who said the popular narrative is that AI is replacing workers, but that isn’t the whole story.

“The more important labor-market story is that AI is increasing the pace at which jobs are created, transformed, and displaced,” Rosenbaum said, adding that as demand surges for electrical engineers, HVAC technicians, cable engineers, and other skilled trades supporting data center construction, the question isn’t whether enough workers exist in the economy.

“It’s whether we can identify them, train them, and connect them to opportunity quickly enough. AI may ultimately prove just as valuable in helping workers navigate career transitions as it is in automating work itself,” Rosenbaum said.

Microsoft’s Boydton data center in Southern Virginia has driven workforce development, with Southside Virginia Community College and Southern Virginia Higher Education Center’s Data Center Academy supporting traditional college-aged and adult learners to earn credentials for data center work.

Rachel Lipson, founder of Harvard’s Project on Workforce and an advisor in the Biden administration’s investments in chips and semiconductor manufacturing, said AI will change the American workforce in ways that previous waves of technology didn’t.

“This wave of technological change looks very different from what we saw in the early days of computers and the internet and the outsourcing of manufacturing in the U.S.,” Lipson said. During those periods, almost all of the job growth went to workers with the highest level of education, but this time around things are different. Welders, she said, can graduate from high school and be making $65,000 to $75,000 a year in short order with almost unlimited overtime.

“A lot of job opportunities today don’t require advanced degrees. We are seeing a blue collar resurgence,” Lipson said. “But it will require making a different set of choices of where we invest in education, training, and what we support,” she added.

Even with the public and, along with them, politicians swinging hard against AI in recent weeks, Lipson doesn’t expect an immediate retreat from the jobs boom.

“In the short term, the moratoria and public backlash are unlikely to have a big impact on demand for these roles, because there are so many AI and data center projects already underway or approved that will continue moving forward despite the recent shifts,” Lipson said.

If opposition to data centers continues to grow, that could muddy the market.

“The longer-term picture could be different. If moratoria spread and significantly fewer new projects are approved, eventually that would flow through to labor demand — but with a lag, as the existing pipeline gets built out,” Lipson said.

ZipRecruiters Bachaud agrees that any shift in sentiment towards data centers will take awhile to manifest itself on the ground because of the time involved in building out AI infrastructure.

“The planning, permits, and actual build can take months, or even years in some parts of the country. As public sentiment shifts, the future of data center locations and the speed at which they develop will likely change, but the rollout of these changes will be slow,” Bachaud said.

Larger macroeconomic forces will also help determine the future of data center workers in construction trades.

“While data centers are boosting the local labor dynamics in places that are building, future demand for many of these workers will depend on general economic growth, like demand for residential and multi-family housing and commercial buildings,” Bauchaud said. “The need for additional housing is high, so many of these workers could transition into other roles if economic conditions improve, but waning affordability for home buyers threatens building progress and thus job availability.”

—CNBC’s Hugh Leask contributed to this report.