Jensen Huang says electricians and plumbers are AI's real winners

Nvidia CEO Jensen Huang used his Carnegie Mellon commencement address to make a provocative claim: the biggest beneficiaries of the AI era won't be software engineers, but electricians, plumbers, welders, and construction workers. With US tech companies set to spend an estimated $700 billion on infrastructure this year alone, the physical buildout of AI — data centers, chip factories, power grids — is straining a labor market that was already running short on skilled tradespeople.

The numbers behind the pitch

Huang's remarks aren't just motivational rhetoric. A March 2026 analysis by staffing firm Randstad found demand for skilled trades workers up 27% over the past three years. Construction roles rose 30%, welding positions climbed 25%, and electrician openings grew 18%. According to CNBC (March 2026), robotic technician roles jumped 107% and HVAC engineer postings surged 67%. Randstad's CEO put it plainly: the talent shortage in skilled trades — not a lack of microchips — is now the real constraint on global AI growth.

McKinsey projections reinforce the gap. The US alone needs roughly 130,000 additional electricians, 240,000 construction workers, and 150,000 supervisors by 2030 just to support AI and data center expansion. BlackRock CEO Larry Fink has publicly warned that the US is already short 600,000 factory workers and 500,000 construction workers — with the electrical shortage emerging as the single biggest bottleneck.

The real infrastructure problem

Data center construction in the US slowed noticeably in 2025, held back by permitting delays, power supply constraints, and a tighter construction labor pool. The irony Huang is pointing to is real: the most advanced software in history can't run without concrete, copper wire, and the people qualified to install them.

White-collar jobs that once seemed automation-proof are increasingly exposed. Content writing, coding support, and data entry roles face significant displacement pressure from AI tools. Trades work is different — a server room flood requires a plumber on-site, not a ChatGPT subscription.

A two-decade training gap

The harder problem is time. Training pipelines for skilled trades take years to scale, and the industry has underinvested in apprenticeships for roughly two decades. Fortune (May 2026) notes that zoning delays and permit backlogs compounded the shortage in 2025 — but even if those cleared tomorrow, there simply aren't enough qualified workers in the pipeline to meet demand through 2030. Huang's message to graduates was direct: this infrastructure wave is "the largest in human history," and the people who build it will be well paid and in high demand for years to come.