The Electrician Making $250,000 Building the Servers That Are Replacing Office Workers

While 80,000 tech workers lost their jobs in Q1 2026, most of them through a terse email citing AI productivity gains, something else has been happening in the construction industry that receives almost no coverage.

Electricians in Texas are earning $250,000 a year. Not the top 1% of electricians — young electricians under 30, three to five years into their careers. TV host Mike Rowe, who spent years documenting the dignity of manual work, said recently he had met three electricians under 30 making between $240,000 and $280,000. The work they are doing? Building the data centers that power the AI that is replacing the office workers.

The irony is almost architectural.

The numbers behind this are not anecdotal. A Randstad analysis of more than 50 million job postings found that demand for robotics technicians has jumped 107% since 2022. HVAC engineers are up 67%. Construction roles have grown 30%. Electricians and welders are up 18 to 25%. These are not marginal movements. They represent a structural reorientation of where physical labour is needed, and it is being driven almost entirely by one thing: the AI infrastructure buildout.

Jensen Huang, the CEO of Nvidia, put it plainly at Davos in January. “This is the largest infrastructure buildout in human history that is going to create a lot of jobs,” he said. “We are going to have plumbers, electricians, construction and steel workers.” Salaries, he added, had nearly doubled in some areas.

Fortune’s analysis of data from Skillit, an AI-powered hiring platform for construction workers, found that workers on data center projects earn an average of $81,800 annually — 32% more than those working on standard construction. Electrical work alone accounts for an estimated 45% to 70% of total data center construction costs, according to the International Brotherhood of Electrical Workers. The US will need roughly 300,000 new electricians over the next decade just to meet demand, on top of replacing the 200,000 expected to retire.

The labour market data makes this even more striking. New figures from Randstad show that it now takes an average of 56 days to hire a skilled tradesperson — plumber, electrician, HVAC technician — compared to 54 days for a desk-based professional. That is the first time in at least fifty years that hiring a tradesperson has taken longer than hiring a knowledge worker. The labour market has, in this specific sense, flipped.

Trade unemployment is now below that of college graduates for the first time in recent memory. Construction employment sits near an all-time high at roughly 8.2 million workers, and 92% of construction firms still report difficulty finding workers to hire. The Associated Builders and Contractors trade group estimates the construction industry needs 349,000 net new workers in 2026 alone, rising to 456,000 in 2027. Oracle reportedly pushed data center project timelines back a full year because they could not find enough qualified tradespeople. Microsoft’s president has called the electrician shortage “the number one problem” slowing their data center expansion.

This is worth sitting with for a moment, because it complicates the dominant narrative about AI and work in a way that is rarely acknowledged.

The standard story runs like this: AI is taking knowledge work and the future belongs to people who can use and direct AI systems. The implicit corollary, rarely stated, is that manual and physical work is a fallback — something you do when you cannot compete in the knowledge economy.

The data center boom is dismantling that corollary. The AI economy is not purely digital. It requires an enormous physical substrate — buildings, cooling systems, power infrastructure, fibre, steel — and that substrate requires human beings with hands-on skills that AI cannot replicate and robots cannot yet replace. The people building that substrate are being paid accordingly.

The deeper story here connects to something that runs through several pieces on this site. Anna, the life coach whose clients started using ChatGPT instead of hiring her, chose to retrain as a dog trainer. Her reasoning was instinctive: physical presence, physical skill, physical trust. She was not wrong. She was slightly ahead of a trend that the data is now confirming.

The trades were never actually inferior to knowledge work. They were undervalued — partly because of cultural bias toward desk-based careers, partly because wages did not reflect the actual difficulty and scarcity of the skills involved. The AI buildout is correcting that mispricing faster than any policy intervention could.

There is a caveat worth stating, and the cynical reader will have already thought of it.

The data center construction boom will not last forever. Infrastructure gets built, and then it does not need to be built again in the same way. The question of whether the maintenance and upgrade cycle sustains employment at current levels — or whether the construction peak subsides, leaving fewer but still well-paid operations roles — is genuinely uncertain.

What is not uncertain is that the electrician in Texas earning $250,000 this year made a better career decision than the computer science graduate who cannot get a junior coding role because the entry-level work has been automated. One of them followed the conventional wisdom about which careers were safe in the age of AI. The other one did not.