Brian Deese, an Innovation Fellow at MIT, was director of the National Economic Council from 2021 to 2023. Anna Pasnau, a Stanford law student, worked at the Council of Economic Advisers from 2021 to 2024.
Everywhere you look, there are escalating concerns that artificial intelligence will cost American jobs. Sen. Bernie Sanders (I-Vermont) has suggested that AI could eliminate nearly 100 million U.S. jobs, while Sen. Josh Hawley (R-Missouri) fears the technology could drive unemployment as high as 20 percent in five years. In its May cover story, the Economist warned of an AI-induced “jobs apocalypse,” where “humans could, like horses in the age of the car, become uneconomical.”
This concern is so overarching and existential that it is obscuring an urgent job risk: AI is exacerbating labor shortages in essential industries. For years, the United States has faced workforce shortages in the skilled trades, and AI is making many of them worse. The reason is physical. Realizing AI’s economic potential requires an enormous build-out of physical capacity — not just data centers, but power plants, electrical transmission lines, grid upgrades and semiconductor factories.
All of this must be built, wired and maintained by skilled tradespeople — electricians, welders, plumbers, HVAC technicians — and there aren’t nearly enough of them. Eighty percent of general contractors already report difficulty filling positions. One in five construction workers is 55 or older. An estimated 30 percent of union electricians will reach retirement age within a decade.
That demand is about to surge. The five largest U.S. cloud and AI infrastructure providers alone have committed at least $660 billion in capital expenditure in 2026, the vast majority directed at data centers and networking. The announced pipeline for data center construction will require about 1.2 million person-years of skilled labor — and data centers are only a subset of the full infrastructure need. Skilled tradespeople are also essential for building adequate energy generation and transmission capacity. National electricity demand is expected to grow more than 30 percent over the next five years, driven by AI, electrification and domestic manufacturing. Meeting that demand will require a massive mobilization of skilled labor.
The crunch is already visible at the local level. In Northern Virginia, the country’s most concentrated data center market, the local electrician’s union has doubled its membership in seven years — and still can’t meet demand. Across metro areas nationwide, inflation-adjusted electrician wages have swung wildly over the past decade, rising more than 60 percent in some places and falling by half in others, suggesting acute local mismatches between supply and demand.
Rising pay alone won’t fix this problem. Skilled technical workers require years of training, and hiring is constrained by inadequate workforce pipelines. As a result, labor shortages are slowing data center construction; Microsoft president Brad Smith has called them the single greatest challenge to building new U.S. capacity.
The consequences extend well beyond tech: If the electrical generation and grid infrastructure that the country needs can’t be built, it will lead to higher costs for American families and businesses, delayed industrial projects and an economy that cannot keep pace with its own ambitions.
Solving this problem won’t be easy — but it presents perhaps our best opportunity to demonstrate one place where the AI boom can help create high-quality jobs for American workers.
Here’s what it will take:
First, make it easier for workers to move to where they’re needed. A licensed electrician in Ohio can’t relocate to Atlanta — now the second-largest data center market — without extensive paperwork and retesting. There is no national electrician’s license, and state reciprocity agreements are a fragmented patchwork. In 2023, Virginia enacted universal licensing recognition, allowing electricians with three years of good standing elsewhere to work without retesting. Every state should follow Virginia’s lead. National apprenticeship credential standards — codeveloped by government, employers and unions — would accelerate the shift.
Second, leverage AI itself. New tools offering real-time on-the-job guidance, immersive training simulations, and streamlined administration can increase both the productivity and supply of skilled workers. As MIT’s Daron Acemoglu, David Autor and Simon Johnson recently argued, AI’s potential as a collaborator — “extending human judgment, enabling new tasks, and accelerating skill acquisition” — is as significant as its capacity to automate.
Third — and most important — we need a radical national apprenticeship effort. Tweaking around the edges of our training system will solve this problem a decade too late. America has made this kind of bet before: In the 1960s, the federal government and states opened roughly one new community college per week, more than doubling enrollment in five years. That kind of ambition is needed again.
The place to start is paying for performance: reimbursing apprentice sponsors when trainees hit milestones. Australia, Finland and Britain do this. California adopted the model in 2022 and saw apprenticeship enrollments rise nearly 10 percent the following year. Both the Biden and Trump administrations have supported piloting this approach, but the current $145 million federal investment is far too small. At $4,000 per apprentice, a national program could cost roughly $4 billion annually — nine times current federal spending, but only about 20 percent of total workforce spending. Beyond pay-for-performance, a federal apprenticeship fund should invest in training instructors, intermediary organizations and program start-up costs where capacity doesn’t yet exist.
The tech companies building AI infrastructure have the most to lose from this labor shortage — and the most to gain from solving it. They should eagerly step in to underwrite a national apprenticeship effort. Already, companies are making promising individual commitments, like Google’s commitment to training 100,000 electrical workers and 30,000 new apprentices.
But this moment calls for even greater ambition. If hyper-scalers don’t step up, government should ensure they pull their weight through fees on data center construction. The cost of building this workforce should fall on the businesses that benefit, not on taxpayers.
The power plants, transmission lines, data centers and semiconductor fabs that will define the 21st-century economy cannot be conjured up by algorithms — they must be built, wired and maintained by human hands. The skilled trades shortage is urgent, but it is solvable. Meeting it head-on would be one of the best things AI could do for American workers — and for every American who pays an electricity bill.
The post AI is sparking a boom in blue-collar jobs. Here’s how to fill them. appeared first on Washington Post.
