How Much Are Firms Spending on AI (and What Will Happen to Headcounts)

How Much Are Firms Spending on AI (and What Will Happen to Headcounts)?

May 06, 2026

• Policy Hub: Macroblog
• Publication
• Economic Research
• Macroeconomy

We recently collaborated with other researchers and central banks to ask senior business executives about AI in their own firms. Our study, issued in February, revealed that AI adoption is already high-78 percent of US firms are using it-and set to grow. Back then, US business executives said AI adoption had little impact to date on productivity or employment in their firms. At the same time, these same executives think AI will raise productivity by 2.25 percent, on average, over the next three years and lower headcounts by 1.2 percent.

That's based on data that's now about six months old. Although we plan to update that paper very soon (and with the help of additional central banks surveying firms across the globe), a couple supplementary questions have come to mind. Specifically, how much are firms actually spending on AI? And how do firm managers anticipate the extra spending will change the composition of their workforce?

To this end, in March 2026, we asked, "For calendar year [2025/2026], how much [did/will] your firm spend on AI adoption and AI related technologies? Include software, subscription services, hardware, worker training, and IT staff support." And to ascertain the differential impact of AI on employment, we posed this question: "Looking forward over the next 12 months, how do you expect AI technologies to affect your firm's hiring levels for [workers with and without college degrees]?"

Spending on AI is rising swiftly

As the table shows, firms report a clear boost in planned investments in AI, signaling that adoption is moving beyond experimentation and toward broader deployment. Data show that per-employee spending on AI rose by 50 percent over 2025 levels: overall, firms spent $1,358 per employee in 2025 and are anticipated to increase spending this year to $2,068 per employee (on an employment-weighted basis). Multiplying this amount by the total number of private, nonfarm payroll employees, we get a ballpark estimate of $280 billion for the aggregate expected investment in AI by private firms in 2026. Though this estimate is simple in its approach, it's eerily close to this recent estimate ($285 billion) from the Stanford Institute for Human-Centered Artificial Intelligence (HAI).

Now, these data carry a whiff of "Bill Gates raising the average," as a small set of firms are deploying a huge amount of capital in the AI space, considerably skewing the underlying distribution. The averages we report in the table are winsorized at the 99th percentile, which pulls the average spending per employee down from $3,108 in 2026. More than half of respondents expect to spend no more than $200 per employee in 2026, while the top 10 percent of firms plan to invest at least $2,800 per employee over the same period. This 14-fold gap between leading adopters and the median firm highlights a highly skewed distribution of AI investment intensity. This widening dispersion in investment mirrors patterns documented in recent studies of digital technologies, where leading firms invest heavily but most firms don't.

Looking at the industry cross-section demonstrates ample divergence. Unsurprisingly, the largest investments are concentrated in larger firms and in knowledge-intensive sectors like professional and business services, where spending is expected to reach $3,470 per employee in 2026, a 74 percent increase from 2025. Conversely, AI spending per employee in the manufacturing sector was just $672 per employee in 2025 and is anticipated to increase to nearly $900 this year. Still, across all industries and across the firm-size distribution, AI spending is poised to increase markedly in 2026.

The impact of AI spending on employee headcounts is still modest and concentrated among workers without college degrees

As AI spending per employee ramps up, we also want to know if AI use will affect workers with college degrees differently than those without a degree.

Though the anticipated effects on hiring remain modest in magnitude, they are not negligible, and they differ notably by educational attainment. Firms increasingly expect AI to slightly dampen hiring needs, particularly for workers without college degrees. Overall, firms expect that the adoption of AI technologies will reduce the demand for workers with college degrees by 0.8 percent over the next 12 months (see the figure). Conversely, firms expect AI to reduce the hiring of workers without college degrees by 1.1 percent.

These numbers reflect the forecasts of senior business executives about how the role of AI will evolve in their own firms-obviously, they don't encompass new jobs at firms not yet formed. Furthermore, they don't capture the extra demand for goods, services, and labor that will flow from AI-driven gains in productivity and real incomes. For both reasons, we regard our survey-based estimates of AI effects on employment growth as tilted to the downside.

The implications of AI for workers with different levels of educational attainment vary considerably across industries. Firms in the professional and business services sector foresee the hiring of nondegreed workers decreasing by 1.8 percent because of AI. In contrast, firms in the manufacturing and retail and wholesale trade sectors expect their needs for college-educated workers to decrease by 1.0 percent and 1.2 percent, respectively. The fact that sectors such as retail, wholesale trade, and professional services report anticipate larger impacts on hiring also aligns with evidence that AI adoption tends to be scale- and capability-dependent, with early movers better positioned to integrate these technologies into their processes.

This pattern is consistent with broader evidence on the distributional effects of AI. Recent empirical findings from the ADP Research Institute show that employment growth has weakened for younger workers in highly AI-exposed roles, while remaining more stable for more experienced workers. Complementary analysis from the Federal Reserve Bank of Dallas highlights that jobs involving routine or easily codifiable tasks-often performed by less-experienced workers-are the most susceptible to automation or augmentation. Together, this evidence reinforces the idea that experience, in addition to education, is a key determinant of exposure to AI-driven changes in labor demand.

Implications and outlook

Taken together, this new evidence is consistent with our recent paper. Firms appear to be ramping up AI spending per employee significantly-by 50 percent, to roughly $2,000 per employee (or $280 billion in aggregate) this year. The distribution of spending shows great heterogeneity both across sectors and across firms within sectors. Some firms appear to have invested heavily in AI, while others are proceeding at a much more modest pace. And although AI's impact on hiring and overall employment levels remains modest-with US firms predicting a 1.2 percent reduction in employment over the next three years-the technology is altering hiring patterns, investment decisions, and the allocation of tasks within firms. The concentration of spending among larger firms and knowledge-intensive sectors, combined with less-educated workers' greater exposure to AI's impact, points to a future where AI's most significant effects could arise through dispersion-across firms, industries, sectors, and workers-rather than through large, aggregate shifts.