US labor's share of income just hit its lowest level since 1947
The AI-and-jobs argument is moving off headcount and onto paychecks — and the government's own accounts now show workers losing ground on the biggest possible scale.
Labor's share of US nonfarm business income fell to 52.8% in the second quarter of 2026, the lowest reading in a series that begins in the first quarter of 1947, according to the Bureau of Labor Statistics productivity report that CNBC examined this morning. That is not an AI finding; it is an accounting identity that says a shrinking slice of what the economy produces is reaching workers as compensation. The companion numbers are just as uncomfortable: the BLS Employment Cost Index shows inflation-adjusted wages and salaries fell 0.4% year over year through June, and August job growth beat expectations while pay growth lagged inflation. Some researchers attribute the long slide to decades of automation, which AI may accelerate.
The temptation is to read that as evidence. It isn't — not yet. CNBC notes that high-paying tech and professional services are shedding jobs while lower-paying hospitality and health care drive hiring, which drags the average down by arithmetic alone, and that pandemic-era wage growth was the anomaly, not this.
What has changed is the quality of the attempts to measure it. Torsten Slok and Sania Edlich of Apollo Global Management found that workers in occupations classified as highly exposed to AI saw real-wage growth run 6.7 percentage points slower after 2023 than workers in less-exposed occupations, with no statistically significant effect on employment — the reading being that firms pocket AI's productivity gains as wage compression instead of cutting staff. Ben Zipperer of the Economic Policy Institute pushed back on exactly that framing: the sample is thin (321 of roughly 800 BLS occupations, only 11 clearing the high-exposure bar), and money not spent on cheaper software reappears as hiring somewhere else, which makes exposed jobs look worse by comparison. Daron Acemoglu expects the same asymmetry for different reasons, arguing that in a flexible labor market with a weak safety net the wage impact will land harder than the employment impact.
David Autor's objection is the one that should stick: knowing an occupation is "exposed to AI" tells you nothing about whether its wages rise or fall. His work with Neil Thompson tracks two clerical jobs that faced the same computerization and diverged — accounting clerks up 39% in pay on 32% fewer jobs, inventory clerks down 13% in pay on 175% more hiring.
We've been watching the demand side of this — one in five US workers now hands tasks to AI, not colleagues — and the pattern there is task delegation, not layoffs. The Dallas Fed reached a similar conclusion in February from the other direction: no overall correlation between AI exposure and wages, except in occupations with a low "experience premium," where AI replaces newcomers and veterans alike and the entry rung of the white-collar ladder stops paying for itself.
What to watch: the third-quarter productivity and labor-share release, and whether the exposed-occupation wage gap survives a bigger sample.
Is lower pay the real shape of AI job displacement, or are we reading a composition effect and calling it a revolution? Tell us in the comments.
