Skip to content
AI.info

The Pulse

Fed researchers say AI's effect on productivity appears small so far

Federal Reserve researchers say AI's effects on aggregate productivity growth appear to be small so far, despite sharply rising data center investment and growing adoption. Fed Vice Chair Philip N. Jefferson separately said the timing and size of future gains remain uncertain.

Fed researchers say AI's effect on productivity appears small so far

The effects of AI on aggregate U.S. productivity growth appear to be small so far, even as data center investment has climbed sharply and adoption is growing, Federal Reserve researchers write. They add that the recent pickup in productivity likely reflects factors that predate the spread of generative AI.

Published on November 12 in the Fed's Finance and Economics Discussion Series, the paper assesses investment, adoption, prices, employment and productivity. The authors are Nick Carollo, Leland D. Crane, Daniel Garcia Molina and Joseph Nichols.

Investment and adoption

The authors found that investment in data centers has risen sharply, driven by current and expected future demand for AI services. That investment includes the buildings as well as the semiconductor chips and other information technology equipment inside them.

The paper also reports rapid growth among companies producing AI models, which now account for a noticeable fraction of software industry gross output. Prices have risen in sectors supplying direct inputs to data centers, including memory chips, the researchers say.

Although AI adoption is growing among consumers and firms, most firms use it little or not at all, according to the paper. Its authors describe the effects on the labor market to date as small and somewhat mixed.

How widely is AI used?

In a November 13 speech, Fed Vice Chair Philip N. Jefferson also described AI's contribution to aggregate productivity so far as limited, while saying it could grow substantially. He cautioned that the size and timing of future gains remain highly uncertain.

Surveys of workers have found that 40 to 50 percent of working adults use generative AI in their jobs, Jefferson said. Among workers who use it at work, only a modest fraction do so every day.

A Census Bureau survey indicated, as of October 2026, that roughly 25 percent of U.S. firms had used the technology in at least one business function as of the middle of the year. The share was closer to 40 percent in knowledge-intensive sectors such as professional services and information, and around 10 percent in mining and leisure and hospitality, Jefferson said.

Mixed evidence on gains

Jefferson noted that U.S. labor productivity growth averaged 2.2 percent between 2020 and 2025, compared with an average of 1.5 percent over the previous business cycle. He said the increase began before popular AI chatbots debuted in 2022 and likely reflects several factors, including a pickup in new business formation.

Some studies have found positive associations between firm-level productivity growth and AI adoption or exposure, while surveys attempting to measure causality more directly have found modest effects so far, he said. Jefferson pointed to a study of senior managers that estimated AI had increased sales per employee by 1/2 percent over the past three years.

The Fed notes that the discussion-series research represents the authors' views and does not indicate agreement by other Board staff or the Board of Governors. Its research and conclusions are often preliminary and circulated for discussion and critical comment.

Sources

Related stories

Explore

More articles