Ben Horowitz flags an electrician shortage that could slow the AI buildout

The AI race has plenty of chips, capital, and ambition. According to Ben Horowitz, it may be running short on people who know how to wire it all up.

The Andreessen Horowitz co-founder said that only 2% of US electricians are certified on DC power. That is a niche credential on paper. It is also one that next-generation AI data centers increasingly depend on.

Why DC power suddenly matters

Most buildings, including traditional data centers, run on alternating current, or AC. Direct current, or DC, flows in one direction. High-density AI facilities are moving toward high-voltage DC systems, including 800V DC setups, rather than conventional AC configurations.

AI data centers run at rack power densities of 120 to 140 kW per rack. Traditional setups run at 10 to 14 kW.

A shortage on top of a shortage

The US Bureau of Labor Statistics projects roughly 81,000 electrician openings each year through 2034. It also forecasts employment growth of 9% for the occupation, well ahead of the average across all jobs.

McKinsey estimates the US will need an additional 130,000 trained electricians between 2023 and 2030 to keep pace with AI-related infrastructure and other construction demand.

Big Tech is starting to pay for training

Google and Meta have pledged upwards of $165 million collectively toward training programs aimed at electricians and related skilled trades.

Background: the physical limits of the AI boom

Horowitz runs one of the most prominent venture firms in tech, and a16z has been a vocal backer of AI development. The concern echoes ongoing reports of labor shortages as AI workloads drive a surge in data center demand.

What this means

For investors and operators in AI infrastructure, the implication is that labor could become a gating factor on project timelines. Capital, land, and technology may all be lined up, and a project can still stall if the workforce is not there.

Companies that invest early in training, or lock in relationships with qualified contractors, may be better positioned to hit deadlines than rivals who assume the workforce will simply appear. The Google and Meta commitments look like an early version of that strategy.

Training an electrician takes time, and certification in specialized systems takes more. The BLS and McKinsey projections both stretch toward the end of the decade and beyond, which suggests the gap will not close quickly even with fresh funding.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.