Meta recruiting electricians at $1M+ annual salaries and establishing vocational school for power infrastructure skills.

The latest bottleneck in AI competition has arrived at construction sites.

On early mornings in Saline, Michigan, hundreds of electricians and construction workers report for duty, then work consecutively for ten hours, repeating this routine seven days a week. What they are building is OpenAI's "Stargate" project—a $16 billion ultra-large-scale data center.

According to McKinsey's workforce forecast, AI infrastructure expansion alone will require the United States to train an additional 130,000 electricians, 240,000 construction workers, and 150,000 construction supervisors between 2023 and 2030. The U.S. Bureau of Labor Statistics estimates there will be 80,000 unfilled electrician positions every year through 2034.

AI companies are willing to pay substantial premiums compared to traditional sectors like housing, healthcare, and energy. Short-term maintenance positions in AI data centers command salaries 42% higher than traditional sectors, and skilled electricians can earn $240,000 to $280,000 annually. Yet even these salaries pale against the costs of delay. Construction of electrical systems accounts for 45% to 70% of a data center's total cost. A single 60-megawatt project delayed by labor shortages loses $14.2 million in revenue each month.

Microsoft President Brad Smith stated publicly: "The electrician shortage has become the number one barrier for us to expand data centers in the United States."

The construction complexity of AI data centers far exceeds traditional precedent. This complexity manifests in three critical areas.

First is the enormous power demand. GPU racks consume 120 to 140 kilowatts each—ten times the standard server racks of a decade ago. A typical ultra-large-scale AI data center may deploy tens of thousands or even 100,000 GPUs, with facilities consuming hundreds of megawatts—equivalent to supplying power to hundreds of thousands of households simultaneously.

Second is an extraordinarily complex power distribution system. The entire system must be designed from scratch, including switchgear, transformers, uninterruptible power supply systems, bus ducts, and cooling loops. Installation and commissioning must be completed by certified professionals; no software can replace this work.

Third is the heat dissipation challenge. The heat density of AI facilities has exceeded the limits of air cooling, requiring direct liquid cooling and immersion cooling. Design and installation depend on certified plumbers and HVAC engineers. HVAC engineer job vacancies increased 78% between 2022 and 2026.

Skilled electricians have become prime targets in a competitive poaching campaign. Some have been recruited three times in eighteen months—like "drafting in Major League Baseball." In 2026, Alphabet and Meta announced combined investment of $265 million in skilled worker training, despite capital expenditures reaching $335 billion.

Rather than wait for the labor market to adjust, AI companies are training workers directly. Meta allocated $115 million to launch a construction worker training school with 5,000 trainees in its first cohort. Tuition, airfare, and accommodation are fully covered, with living stipends provided. After four weeks of training, graduates begin work immediately. Meta has saturated Facebook and Instagram with recruitment ads.

Meanwhile, in March 2026, OpenAI partnered with North American building trade unions to hire union workers, gaining access to personnel with deeper training experience.

This competition for skilled workers has reached high schools. AI companies actively recruit high school graduates, promising lucrative careers. The results have been striking: between 2017 and 2025, Generation Z enrollment in technical schools increased 1,421%. A 2025 survey found 42% of Generation Z willing to skip college and enter skilled worker training directly, with 60% already planning careers in technical fields.

The electricity question looms over tech executives' desks.

Global data center power consumption reached 565 terawatt-hours in June 2026, a 26.4% year-over-year increase—nearly all driven by AI servers. Though AI servers represent only a small share of installed capacity, they consumed 31% of total electricity, a figure expanding at 84.2% annually. The International Energy Agency calculated that in 2025, AI data center power consumption grew 16 times faster than global overall power consumption. Some institutions project that by 2028, data center power consumption will account for 6.7% to 12% of total U.S. power consumption.

These costs are passed to American residents. Electricity prices in areas with dense data centers have risen up to 267% over five years. PJM, the nation's largest power grid operator, calculated that between June 2024 and June 2025, data centers added $9.3 billion in additional electricity costs for consumers across Illinois through Washington, D.C.

Yet data center construction positions are project-based. The Stargate facility in Texas, consuming 1,000 megawatts—equivalent to a nuclear reactor's output—requires 6,400 workers during construction but only 100 to 1,000 permanent workers for daily operations. After completion, hundreds of thousands of trained workers will disperse to other industries, potentially depressing wages across the sector.

The best-case scenario unfolds when these data centers come online just as a real estate boom accelerates.