AI Infrastructure Boom Drives 126,000 Non-Residential Construction Jobs in Past Year
Summary
- • US shed 23K jobs overall in July while construction added 22K, nearly all in non-residential and infrastructure — AI's physical build-out is the key driver.
- • Non-residential construction and infrastructure sectors added 126,000 jobs over the past year, even as residential construction lost 44,000.
- • RSM chief economist calls it a 'historic cap-ex super cycle': $1.6T invested in AI infrastructure in two years, with $4.5–$5T projected over the next five.
- • Apollo data shows AI model/app companies lose $0.59 per dollar of revenue, while energy and grid companies profit $0.41 — physical AI infrastructure is the clearest financial winner.
Details
Construction gains 22K while US economy sheds 23K jobs in July
US employers shed 23,000 jobs overall in July, while construction firms added 22,000 — almost entirely from non-residential and infrastructure categories tied to the AI build-out.
Non-residential construction added 126K jobs over the past year
Over the past year, non-residential construction and infrastructure sectors added 126,000 jobs, while residential construction shed 44,000 — a clear divergence driven by AI infrastructure investment.
Nonresidential specialty trades added nearly 78K jobs in a year
Nonresidential specialty trades — electricians, HVAC, site prep, concrete — added nearly 78,000 jobs over the past year, the single largest category of AI-linked labor growth.
Engineering and building construction each added 20K+ jobs
Nonresidential building construction added 28,000 jobs and heavy/civil engineering added 21,000 over the past year, tracking closely with where hyperscaler capital is flowing.
AI model companies lose $0.59 per dollar; energy firms profit $0.41
Apollo chief economist Torsten Slok finds energy/grid companies keep $0.41 of operating profit per $1 revenue; silicon/equipment $0.34; compute/cloud $0.24. Model and application companies lose an average $0.59 per dollar of revenue.
RSM cites $1.6T two-year AI cap-ex run rate expanding to $5T
RSM chief economist Joe Brusuelas describes a 'historic cap-ex super cycle' — $1.6T invested in AI infrastructure over the past two years, with $4.5–$5T expected over the next five years.
AI's physical world expansion requires copper, power, concrete, and labor
Hyperscalers are writing enormous checks for chips, data centers, and power infrastructure. Soaring electricity demand is turning the power grid into a binding constraint on the AI build-out, further driving demand for electrical and engineering workers.
How the AI infrastructure investment cycle is reshaping the U.S. labor market — creating jobs far from Silicon Valley, in hard hats rather than hoodies.
What This Means
The AI investment cycle is reshaping the U.S. labor market in a way that most economic narratives miss: the fastest-growing jobs tied to artificial intelligence are in construction and trades, not technology. As hyperscalers pour trillions into data centers and power infrastructure, electricians, HVAC technicians, and specialty contractors are the immediate beneficiaries — and are working far from Silicon Valley. Apollo's profit margin data underscores a striking paradox: companies building AI models are currently losing money, while companies building the cables, cooling systems, and grid capacity to power those models are highly profitable. The build-out's sustainability ultimately depends on continued financing of model and application companies, even as they remain unprofitable at scale.