Two days of training end with a diploma reception. Companies eager to hire the new graduates have come ready to hand out jobs that can pay more than $100,000 a year.
Welcome to the data center economy.
In the race to build out America’s AI infrastructure, tech companies are struggling to find technicians with sufficient skills to install the cooling, server racks, electrical equipment and fiberoptic cables necessary to meet the off-the-charts demand for the revolutionary technology. Even as AI politics have shifted – particularly after recent midterm election – the lack of available labor to build continuing data center projects would almost certainly trump any potential construction slowdown.
So the industry is going to extraordinary lengths to teach and hire people to fill those crucial jobs. Several nonprofits, online learning companies and all four of the big AI hyperscalers — Amazon, Google, Meta and Microsoft — offer various forms of free, accelerated training courses for the technical skills needed to build out data centers.
Some of those positions can pay six-figure salaries. And many Americans don’t even know these jobs exist.
Javonntae Hill studied video game design. He’s in his early 30s and he was looking for a career — but didn’t quite know where to start.
Hill figured he’d give it a shot. Over the course of a month, in a classroom of 30 people, he got thrown into a mix of lectures and hands-on training, learning about how to place and run fiberoptic and copper cables and what each cable insulator color specifies.
He was nervous: He hoped he could retain all that information for his new job.
“Graduation was on Friday,” Hill said. “I started orientation on Monday.”
Meta partnered with Associated Builders and Contractors, a construction industry trade group that has been training workers for decades. Unlike traditional training methods, including trade school or an apprenticeship, ABC’s accelerated courses aren’t designed to be comprehensive; they’re designed to get as many people working in data centers as possible.
“It’s like the military,” said Joel Thames, head of ABC’s safety and workforce group. “Does boot camp get them trained for any scenario? No. But it gives them a foundational education before they spend time on the job site.”
Hill said his nerves were unfounded: He showed up to the Ohio job site overprepared. And grateful.
“I could do this for the rest of my life,” Hill said. “I’ve seen postings of jobs available, but I have a lot of loyalty toward Meta.”
Despite growing public opposition and a wave of new lawmakers pledging to regulate their expansion, data center construction is virtually unstoppable.
AI infrastructure spending is expected to approach $1 trillion a year , annually for the foreseeable future, according to Columbia Business School economist Stijn Van Nieuwerburgh.
But construction faces many logistical hurdles , including building permits, power guarantees, materials and financing. Even the deepest-pocketed AI companies that can overcome those obstacles are encountering a consistent roadblock: America has far too few qualified people to build data centers — particularly in some of the remote or rural locations where they’re being built.
The United States has a shortage of 1 million engineers needed to meet data center construction estimates, PwC estimates. Many otherwise ready-to-go sites have been delayed for months while they assemble a qualified crew to build them. Only about half of the AI computing capacity scheduled to come online between now and 2028 is actually expected to meet its target date, according to Goldman Sachs.
That’s a costly wait: For large AI data centers, a month of delay can cost anywhere from the low tens of millions to well over $100 million, depending on location, scale and contract terms, said Mona Dajani, co-chair of Cooley’s infrastructure, energy and real estate practice.
So AI companies are addressing the challenge by ditching requirements for multi-year trade school courses or apprenticeships in favor of accelerated, high-intensity courses designed specifically for data center construction jobs. Many are offering the courses for free. A growing number will even pay learners to take the courses. Classes can run from 15 weeks to just two days, depending on the job.
Desperation for labor is so high that AI companies are placing some top graduates on the job the day after they gain their certifications.
Robert Burgess had been bouncing around construction jobs in and around Canton, Mississippi. He knew how to do some low-voltage electrical work and had heard that he could make a lot more money cutting fiber for a telecom or a data center site.
Burgess Googled “internet fiber training Mississippi” and a four-week Amazon training program called I2PA (Information Infrastructure Pre-Apprenticeship) surfaced at a local community college.
“I had nothing to lose,” Burgess said. “I put in my email, and I got accepted.”
Over the course of the training, Burgess learned how electricity is generated and delivered, how servers are installed and how they fit into a data center. Amazon trainers set up a mock data center so that Burgess and his classmates could get hands-on training.
At the end: a job fair. Burgess got a job from Yates Construction and Turner Construction, which were partnering on the tail end of a data center construction site in the area.
That was in 2024. Now he’s an assistant superintendent overseeing another data center job site — an opportunity he never knew was available to him.
“Without going through that program, I would have never been here,” Burgess said.
Burgess’ career path speaks to the biggest obstacle in overcoming the labor skills gap: the lack of knowledge that these high-paying jobs exist.
“People become forklift operators because they know what it is,” said Roger Wehner, Amazon’s vice president of economic development. “But there are incredibly high-demand jobs that people just don’t know about.”
By design, these intensive programs are not meant to be like trade schools.
“We’re not solving the world’s problems with 30 kids a class,” said Wehner. “We don’t duplicate community college. We know the cutting-edge technology. We bring the dollars. Community colleges have the top faculty.”
But some training companies believe their accelerated programs come with advantages over traditional classroom training methods. Doug Donovan, CEO of online training company Interplay Learning, noticed that the military, medical and aviation industries use virtual and augmented tools — but not trade workers. So Donovan started combining AR and IR with traditional teaching methods into his courses.
“We were surprised the trades weren’t leaning into that,” said Donovan.
Other programs — including the 15-week intensive training offered by nonprofit Per Scholas — try to replace on-the-job training in a classroom setting.
“We design immersive trainings in partnership with employers that try to replicate as much as possible the environments people will move into,” said Caitlyn Brazill, president of Per Scholas.
That hands-on approach helped Janay Ramsey, a former hair stylist in Ridgecrest, California, who now works as a Google data center technician outside Atlanta. The 15-week Per Scholas course was intense, Ramsey said.
“They don’t sugarcoat it when you’re there,” she recalled. “You’re doing bookwork and training, actual labs, learning the physical and the technical side.”
Despite the full-time 9-5 schedule — plus homework — Ramsey said she still took on clients for hair appointments to pay the bills while she was studying. Per Scholas’ courses were free, but she had no other way of bringing in money.
Working while learning isn’t recommended, says Brazill, but about a quarter of its students work nights and weekends to support themselves through their coursework. The organization offers learners free financial coaching and $3,000 interest-free loans funded by a $100 million investment from Google.
Not every training is so long or intense. Amazon offers two-day trainings for fiber splicers — the position Amazon struggles most to fill, Wehner said.
Fiber work requires marrying thousands of thin cable fibers with another piece of cable, finding the mate for each strand. They each need to be cleaned and melted 12 at a time, tested, resealed and placed correctly and safely. It’s not difficult work, per se, but it’s precise and ultra-repetitive work that requires a worker with a particular mental aptitude.
But after two days, the learners are ready to go to work. Graduates with perfect practicums get placed immediately in the top-tier jobs: in data centers, where entry-level workers make up to $32 an hour plus substantial overtime — take-home pay of around six figures.
“We bring technical acumen to the table; people get better jobs,” said Wehner. “That’s the dream.”