Artificial intelligence has rapidly woven itself into everyday life. Much like the internet reshaped nearly every industry in the early 2000s, AI is now being added to everything from search engines and office software to healthcare, finance, manufacturing, and customer service.

Large language models such as ChatGPT, Claude, and Grok have changed how millions of people find information, solve problems, and complete everyday tasks. At the same time, companies such as Nvidia and Anthropic have seen their revenues and valuations soar as investors race to capture a share of the AI boom.

But behind the software, models, and massive valuations lies a far more physical reality: AI requires an extraordinary amount of land, electricity, equipment, infrastructure, and skilled labor.

McKinsey estimates cumulative global data center investment could hit $6.7 trillion by 2030. Dell'Oro projects data center capex alone will blow past $400 billion in 2026. That capital has to flow through job sites — and the job sites don't have enough people.

For years, Mike Rowe has warned that America is suffering from a widening disconnect between the jobs available and the workers trained to perform them.

Through the mikeroweWORKS Foundation, Rowe argues that millions of skilled positions remain unfilled while vocational education has disappeared from many schools and four-year college has been promoted as the default path for nearly every student. His message has generally focused on the people needed to repair roads, maintain utilities, manufacture equipment and keep the country’s physical infrastructure functioning.

Artificial intelligence has made that warning more urgent.

The AI boom is usually described as a competition for chips, computing power and software talent. But none of it works without electricians, HVAC technicians, pipefitters, welders, lineworkers, equipment operators and construction supervisors. These are the people who must build the data centers, connect them to the electrical grid, install their cooling systems and maintain the physical infrastructure that keeps them online.

The problem is that technology companies are entering this construction boom after America has spent decades underinvesting in the skilled-trades pipeline.

The Bureau of Labor Statistics projects approximately 649,300 openings annually across construction and extraction occupations through 2034. Electricians alone are expected to have roughly 81,000 openings each year, partly because experienced workers are retiring or leaving the occupation. Becoming a qualified electrician generally requires a multiyear apprenticeship, and most states require licensing. This is not a workforce that can be produced as quickly as a technology company can order more servers.

Data-center developers are already feeling the consequences. A 2026 Deloitte analysis found that job postings by data centers for electrical technicians increased by more than 180% between 2023 and 2025. Sixty-three percent of surveyed data-center executives identified a shortage of skilled data-center labor as their greatest obstacle to securing talent. Power companies and data-center operators are also competing for many of the same technicians, engineers, plant operators and lineworkers.

The labor shortage is no longer merely a workforce issue. It is becoming a constraint on how quickly AI infrastructure can be built.

Developers are facing higher wages, more expensive subcontractor bids, scheduling delays and greater competition for qualified workers. Electrical work can represent between 45% and 70% of a data center’s construction cost, making the shortage of electricians particularly consequential. Industry analysis now describes skilled labor—not financing or technological capability—as one of the most serious bottlenecks facing data-center development.

Money is showing up somewhere — just not in the form of a faster pipeline. It's showing up in wages. Data center construction jobs now pay roughly 25% to 30% above standard commercial rates, and in hot markets that premium gets extreme: specialized electricians in Northern Virginia and Texas are reportedly commanding salaries up to $280,000. Union data center electricians typically earn another 10% to 20% on top of that versus non-union counterparts. The DMV region's local electricians' union has roughly doubled its membership since 2018, now sitting around 14,700 members — almost entirely on the back of data center demand.

Tech companies are responding by trying to widen the pipeline itself rather than skip it. Google.org has pledged funding to expand the NECA/IBEW Electrical Training ALLIANCE to 70,000 apprentices. Microsoft and the IBEW are running parallel investments into apprenticeship expansion. None of it collapses the five-year clock — it just means more people enter the pipeline at the front end, for a shortage that won't resolve until well into the 2030s.

There's a genuinely strange data point buried in a recent labor-risk analysis worth sitting with: one AI-automation-risk model rated electricians a 9 out of 100 for replacement risk — the lowest exposure to AI automation of any major trade tracked. The industry building the machines that are supposedly coming for every job has, by its own risk modeling, decided the electrician's job is nearly un-automatable. That's not an accident. Physical, code-governed, high-voltage field work bolted to a specific building doesn't compress into a model. It's the one part of the AI supply chain that can't be trained on a GPU.

Silicon Valley’s story has always contained a romantic strain of pure intellect conquering the material world. The current chapter is a reminder that the material world still has the final say. Algorithms do not install switchgear. Models do not pull wire. Large language models do not commission chilled-water systems rated for the heat density of modern AI racks.

The future of artificial intelligence will be shaped by the people who show up every morning with tools, experience, and the willingness to do difficult, precise work under deadline. Their numbers, their training, and their willingness to take these jobs will determine how quickly—and how broadly—the AI boom actually arrives.

The next phase of the technology story is not only about smarter software. It is about whether we can produce enough skilled tradespeople to give that software a physical home. For once, the bottleneck is not in the code. It is in the hands that build the data centers.

The irony is difficult to miss. Some workers are being told that artificial intelligence could eventually replace them. Meanwhile, the companies building that technology cannot grow without hundreds of thousands of workers whose jobs require them to show up in person and work with their hands.

Silicon Valley may be inventing the future. But it still needs blue-collar America to build it.