Datacenter construction recently surpassed office construction for the first time, and the employment numbers are following.
New data shared by venture firm a16z — sourced from BLS and Citadel Securities — shows that nonresidential specialty contractors, the category that captures most datacenter construction work, have been the dominant driver of net new construction payrolls since mid-2024. By January 2026, the sector added roughly 75,000 net jobs in a single month, dwarfing contributions from heavy and civil engineering, residential specialty contractors, and construction of buildings combined.
The residential and general commercial segments tell the opposite story. Residential specialty contractors have been shedding jobs steadily since mid-2024, with losses deepening through the back half of the year. Construction of buildings — the category covering offices, schools, and retail — has similarly dragged negative. The aggregate “Total” line on the chart, shown as a dotted orange trend, flipped sharply positive in November 2025, driven almost entirely by the datacenter subcategory.
The underlying reason is well-documented. The launch of ChatGPT in late 2022 triggered an arms race among hyperscalers — Microsoft, Google, Amazon, and Meta — that has since redirected enormous capital flows into purpose-built AI infrastructure. Training and running large language models requires specialized hardware that must be housed in facilities with industrial-scale power and cooling, and spending on datacenter construction crossed above office construction for the first time in late 2025. The employment chart from a16z shows that labor markets are now reflecting the same shift.
Jensen Huang has framed this transition as something more fundamental than a construction cycle. He argues that datacenters are no longer storage facilities but “AI factories” — active generators of tokens and intelligence — and that the world’s enterprise IT infrastructure needs to be rebuilt from the ground up to support that model.
Office construction, meanwhile, faces headwinds that are structural rather than cyclical. Remote and hybrid work has permanently reduced the floor space companies require, and AI-driven automation is putting additional pressure on white-collar headcounts. New office supply is likely to stay subdued for years.
The a16z data puts a labor-market face on a capital story that has been building since 2023. The physical infrastructure of the AI era is being built now, and it is pulling construction employment with it — away from homes and offices, and toward the server farms that increasingly underpin the economy.