AI lifts U.S. growth, but productivity gains lag

Growth is running hot—productivity, less so

Artificial intelligence is energizing the U.S. economy, largely through a wave of corporate spending and soaring stock valuations. But the technology hasn’t yet delivered broad, measurable gains in how much the average American worker produces per hour.

Productivity typically improves when workers can accomplish more in the same amount of time. AI could help in two ways: by augmenting people—automating routine tasks so employees focus on higher-value work—or by replacing some roles entirely, raising efficiency among those who remain. For now, evidence suggests these effects are limited and uneven.

Economists split on current AI impact

Wall Street economists are split on whether AI is currently lifting productivity. Goldman Sachs researchers say output per worker has accelerated in tech and related fields such as scientific research over the past five years, and they attribute part of that improvement to AI. JPMorgan Chase economists, by contrast, have not found a strong link between AI usage and industry-level productivity so far and see no clear relationship—outside tech—between AI adoption and slower employment growth.

Early labor-market signals are modest

Martha Gimbel of Yale’s Budget Lab says the palpable excitement around AI hasn’t yet translated into equally dramatic economic results. Her team’s analysis of Labor Department data suggests some early-career workers have been displaced since the launch of ChatGPT in late 2022, but the impact appears modest.

Occupational mix shows little change

If AI were reshaping the labor market at scale, the mix of occupations would be shifting markedly—toward jobs considered more insulated from AI, like home health aides, or those that can be enhanced by AI, such as computer-systems managers. Instead, the Budget Lab found little change: the share of U.S. workers in jobs highly exposed to ChatGPT was about 18.2% in the three months ended November 2022 and roughly 18.3% in the three months ended August 2024.

Young workers feel it first

The effects may be more concentrated among new entrants to the workforce. Recent graduates are seeing faster shifts in which occupations they enter compared with older workers, a sign AI could be altering early-career pathways. That aligns with a recent study by Stanford economists, which finds job prospects deteriorating for younger workers in fields where generative AI can easily automate tasks—software development being a prime example. Still, the group is small in the context of the broader economy: only about one-quarter of the roughly two million U.S. software developers are under 30, versus about 163 million people employed overall.

Investment is the clear channel—for now

Where AI’s effect is unmistakable is in spending. In the first half of the year, roughly two-thirds of U.S. GDP growth came from business investment in software and information-processing equipment, reflecting a rush to build AI infrastructure and deploy new tools. That investment, alongside a tech-led market rally, has also supported consumer spending via wealth effects.

Adoption remains in early innings

Adoption remains in early innings. A recent Census Bureau survey shows about 10% of firms report using AI in some capacity, up from about 6% a year earlier and roughly 5% when ChatGPT debuted. As more companies climb the learning curve, productivity benefits could accumulate.

Lessons from past tech waves

History suggests patience is warranted. Economist Joshua Gans of the University of Toronto notes that with past technologies—like desktop computers—productivity gains arrived only after workers and managers rethought processes and developed new skills. Many users today are still experimenting with AI, he says, often trying small tasks, encountering friction, and stepping back. Gans, who founded an AI education-technology company, remains optimistic that significant productivity dividends will emerge as organizations redesign workflows around the tools.