Read the full analysis: US Output Is Up 10% As Hours Stay Similar. Where Does AI Fit? on ThorstenMeyerAI.com
TL;DR
US nonfarm business output increased 10.0% from the fourth quarter of 2022 to the second quarter of 2026, while hours worked rose 0.9%, according to calculations based on BLS indexes cited by Thorsten Meyer. Productivity has accelerated, but the figures alone cannot establish that AI caused the change; hiring, pay and the distribution of gains remain key questions.
US nonfarm business output rose 10.0% from the fourth quarter of 2022 to the second quarter of 2026, while hours worked increased 0.9%, according to calculations from Bureau of Labor Statistics indexes cited by Thorsten Meyer. The gap points to stronger productivity, but it does not establish that artificial intelligence caused the improvement or explain why employment and wage outcomes have been uneven.
The latest indicators present a mixed picture: real GDP grew at a 3.4% annual rate in the third quarter of 2026, according to the Bureau of Economic Analysis’ advance estimate released October 29, while the BLS reported that nonfarm business productivity increased 2.2% over the year to the second quarter. Over that same 12-month period, business output rose 2.5% and hours worked rose 0.2%.
Meyer’s longer comparison starts in late 2022, around the launch of ChatGPT, and ends in the second quarter of 2026. His calculations from BLS indexes show output up 10.0% and hours up 0.9%, equivalent to productivity growth of about 2.5% a year over that period. The BLS series also shows average annual productivity growth of 2.1% from the end of 2019 through the second quarter of 2026, compared with 1.5% from 2007 to 2019.
Employment measures do not mirror the output gains evenly. The source reports that payrolls grew by 496,000 in the 12 months to September, with health care and social assistance adding 520,000 jobs. Information industries lost 120,000 jobs and were 12% below their November 2022 peak. These figures describe different parts of the economy; they do not show that AI produced the job losses or gains.
US data
Infographic 01 · More output, same hours
Ten percent more output, almost the same hours
The US business sector produces much more than three years ago with almost no extra work. The new jobs are mostly in health care.
The last 12 months
Nonfarm business sector, Q2 2025 to Q2 2026
Outputreal, value added+2.5%Productivityoutput per hour+2.2%Hours workedall workers+0.2%
Where the jobs went
Payroll jobs, change over 12 months to Sep 2026 (preliminary)
Since Q4 2022: output +10.0%, hours +0.9%, productivity +9.1% (my arithmetic from the BLS indexes). Hiring rate in Aug 2026: 3.3% of jobs (2019 average 3.9%); layoff rate 1.0% (2019: 1.2%). Correlation, not proof of cause.
Sources: BLS, Productivity and Costs Q2 2026 (3 Sep 2026) · BLS, Employment Situation Sep 2026 · BLS, JOLTS Aug 2026.
Powered by Thorsten Meyer AI
Productivity Gains Meet Uneven Hiring
Productivity matters because it measures how much output businesses produce for each hour worked. Sustained gains can support higher incomes and economic growth, but the aggregate measure does not say who benefits, whether workers share in the gains, or whether fewer hours reflect efficiency, weak hiring or changes in the mix of industries.
The figures cited by Meyer raise distribution concerns. The labor share—the portion of business output paid as wages and benefits—fell to 52.8% in the second quarter, described in the source as its lowest level since records began in 1947. Real private-sector wages fell 0.6% over the year to September. Meanwhile, unemployment was 4.2% overall, compared with 8.0% for people aged 20 to 24 and 7.0% for Black workers. Those disparities matter to workers entering the labor market and to households whose pay may not keep pace with output.
For businesses and policymakers, the central issue is not simply whether productivity is rising. It is whether the increase persists, what is driving it, and whether it supports broadly shared wage and employment gains. The available measures make those questions more urgent but do not settle them.
What the Post-2022 Comparison Shows
The comparison uses the end of 2022 as a reference point because ChatGPT launched on November 30, 2022, making it a commonly used marker for the public arrival of generative AI tools. A before-and-after comparison can show that productivity and AI adoption developed over overlapping periods. It cannot by itself prove a causal link.
The BLS productivity measure cited here covers the nonfarm business sector: private businesses other than farms. It excludes government, nonprofits and households. Productivity is output divided by hours worked, so it can rise when output grows faster than hours, regardless of whether the source is AI, other technology, investment, organizational changes or shifts among industries.
Meyer says his analysis relies on public data from the BLS, BEA, Census Bureau, Federal Reserve and Eurostat. He also refers to an October 31 Financial Times opinion column asking whether the pattern is evidence of AI’s labor-market impact, and a Wall Street Journal headline about growth that is not reaching everyone. He says he could not access either full article and does not rely on their arguments. The reported figures and the question of causation should be kept separate.
““Numbers can show that two things happen at the same time. They rarely show that one causes the other.””
— Thorsten Meyer
AI’s Role Is Not Established
The data cited do not establish how much, if any, of the productivity increase came from AI. The source reports that about one in four US businesses said it used AI in early October, based on Census Bureau data, but does not provide a definition of use or show how adoption relates to each firm’s output or staffing. Technology-sector employment also began declining before generative AI was widely used in businesses, according to the source.
Other explanations remain possible, including investment in computers and software, changes in business practices, industry composition and broader economic conditions. The source says computers and software accounted for about a third of third-quarter growth, much of it linked to AI infrastructure investment, but that does not measure AI’s contribution to productivity or jobs. It also remains unclear whether the productivity gains will continue, how hiring will evolve, and whether wages will catch up with output.
Watch Adoption, Hiring and Pay
The next evidence will come from updated BLS productivity and employment releases, new GDP estimates from the BEA, and further Census Bureau data on business AI use. Together, those series can show whether output per hour continues to rise and whether hours, payrolls and wages change alongside it.
For a stronger assessment of AI’s effect, economy-wide trends will need to be matched with evidence about adoption and performance within firms and industries. Until then, the figures support a finding of faster productivity growth and uneven labor outcomes, not a definitive explanation of what caused them.
Key Questions
How much did US business output and hours worked change?
From the fourth quarter of 2022 to the second quarter of 2026, output in the nonfarm business sector rose 10.0% and hours worked rose 0.9%, based on calculations from BLS indexes cited by Thorsten Meyer.
Does the data prove AI caused the productivity gains?
No. The figures show productivity rising during a period of growing AI interest and investment, but they do not isolate AI’s effect from other technologies, investment or economic changes.
What does the latest productivity figure cover?
The BLS reported that nonfarm business productivity rose 2.2% over the year to the second quarter of 2026. The measure covers private businesses except farms, and excludes government, nonprofits and households.
Are jobs and wages rising along with output?
The indicators are mixed. Payrolls grew by 496,000 in the 12 months to September, but information industries lost 120,000 jobs. The source also reports that real private-sector wages fell 0.6% over the year to September.
Source: ThorstenMeyerAI.com