US tech workers
U.S. productivity has been rising, but economists are still trying to determine how much of the improvement can be attributed to artificial intelligence. Getty Images

American workers are producing more per hour than they have in years, with productivity growing at an annualized rate of 2.5% since early 2023. But despite hundreds of billions of dollars being invested in artificial intelligence, economists are finding that much of the improvement may have little to do with AI, at least for now.

Research from the Federal Reserve Bank of San Francisco shows that increased business investment and more intensive use of labor and equipment have contributed substantially to recent productivity growth.

Between the first quarter of 2023 and the second quarter of 2026, business-sector output per hour grew at an average annualized rate of 2.5%, compared with 1.5% from 2005 to 2022. The latest pace is still below the 3.3% recorded during the technology-driven productivity boom between 1996 and 2004.

Of the recent improvement, increased capital investment and more intensive use of productive resources contributed 1.3 percentage points annually, nearly double their contribution over the preceding 18 years, the San Francisco Fed found.

Read more

A broader measure of efficiency, known as total factor productivity, which accounts for improvements not explained by additional labor or capital inputs, grew at just 0.7% annually. That was only slightly above the 0.5% average recorded between 2005 and 2022.

The figures do not mean AI has contributed nothing. Some investments in software, technology and equipment may be directly related to artificial intelligence. However, the research could not establish how much of the recent improvement was specifically driven by AI.

The latest Bureau of Labor Statistics figures show that nonfarm business productivity increased at an annualized rate of 1.4% in the second quarter of 2026, as output rose 1.7% and hours worked increased 0.3%. Compared with the same quarter a year earlier, productivity was up 2.2%.

For businesses spending heavily on AI tools, turning faster individual work into measurable company-wide improvements has become a central challenge.

S&P Global has a target of roughly 2% annual productivity growth, yet even the financial information company cannot clearly separate the benefits of AI from other improvements to its operations.

The picture is considerably different inside companies developing the technology. AI startup Anthropic disclosed in its initial public offering filing with the Securities and Exchange Commission on Monday that its Claude AI models wrote more than 90% of contributions to the company's codebase in September.

The average daily volume of code contributed per engineer during the three months ending September 2026 was more than 45 times the average recorded in 2024.

Anthropic also reported that its revenue over the 12 months ending September 2026 increased approximately 1,200% from a year earlier, while headcount grew about 170%. Revenue per employee rose from $1.4 million to more than $6.6 million over the same period.

The company attributes part of its increased operating efficiency to extensive internal use of Claude across engineering, research and other departments. However, the growth in code contributions is a measure of output volume rather than a direct measure of software quality or the financial return generated by AI.

The contrast becomes more apparent when comparing America's largest technology companies with the rest of the stock market. An analysis published by Apollo Global Management on Oct. 24 found that revenue per employee and profit margins were rising among the so-called Magnificent Seven technology companies but remained broadly flat across the other 493 companies in the S&P 500.

The analysis, written by Apollo Chief Economist Torsten Slok, found little evidence so far of a measurable financial payoff from AI across corporate America outside the largest technology businesses.

The Magnificent Seven include Apple, Microsoft, Alphabet, Amazon, Nvidia, Meta and Tesla, which have been among the biggest beneficiaries of investor enthusiasm for artificial intelligence.

While stronger revenue and margins among these companies do not establish how much of their performance is caused by AI, the difference raises questions about how widely the technology's economic benefits have spread.

Research by the San Francisco Fed also found that industries with higher AI adoption, including information, finance and education, had not consistently increased their contributions to overall productivity growth compared with the years before the pandemic.

Anthropic's own economic research offers another perspective. Its analysis estimated that widespread use of current-generation AI could add roughly 1.8 percentage points to annual U.S. labor productivity growth over the next decade under certain assumptions.

Once researchers accounted for the reliability of AI-generated work and the possibility that some tasks would still require human intervention, the estimated contribution fell to approximately 1 percentage point annually. That figure represents potential future productivity gains, not improvements already recorded in the U.S. economy.