Productivity rose, while labor's share fell to a record low

U.S. nonfarm productivity rose in the second quarter, but the headline number hides a change in distribution. Companies produced more output in the same amount of time, which can be read as an early sign that investment in artificial intelligence and automation is beginning to pay. A rise in productivity, however, does not automatically mean that workers received more.

U.S. productivity rises faster than expected in second quarter (Reuters, 2026-08-06)

출처: reuters.com

The Bureau of Labor Statistics preliminary release recorded a 1.4% annualized rise in nonfarm productivity, with output up 1.7% and hours worked up 0.3%. Unit labor costs rose 1.3%, so the productivity gain partly offset the increase in compensation costs. Reuters reported that economists had expected productivity to rise 0.6% and unit labor costs to rise 2.1%. Efficiency was better than expected, while the cost of producing one unit of output rose less than expected.

Productivity and Costs, Second Quarter 2026 Preliminary (U.S. Bureau of Labor Statistics, 2026-08-06)

출처: bls.gov

The original post also needs a small correction. It says inflation-adjusted compensation fell 3.2%, while the BLS table gives a 3.1% decline. The difference is small, yet a primary release has to control the number when the number carries the argument. The larger signal is labor's share. The BLS says compensation accounted for 52.9% of nonfarm output, the lowest reading in a series that begins in 1947. Productivity gains therefore arrived alongside a smaller share of output going to workers. The rule that creates more output and the rule that divides that output are moving in different directions.

This cannot be settled by asking only whether AI raised productivity. Automation may reduce the labor time needed for each unit, while pricing power, profit margins and weaker wage bargaining determine where the extra output ends up. The BLS release shows the distribution result, but it does not prove one single cause. Reuters' references to automation and AI should be read as a possible background, not as a demonstrated causal estimate. The calendar matters too. These are preliminary figures, and the revised release is scheduled for September 3, 2026. If productivity holds while labor's share rebounds, temporary price or compensation effects will explain more of the gap. If both measures stay in the same direction, the stronger reading is that the channel carrying productivity gains into wages has weakened.