BofA finds no AI job collapse, 127,000 capex jobs offset

AI adoption has not triggered US job destruction, with high-exposure industries holding employment flat versus 2 percent growth elsewhere, yet college graduates aged 22 to 27 face unemployment above 2019 levels, Bank of America Securities finds.

"Technology shocks primarily replace tasks, not entire occupations or labor demand," Stephen Juneau, economist at Bank of America Securities, said in the Aug. 11 report.

The analysis, covering 206 industries using the Felten, Raj and Seamans AI exposure index, found no statistical correlation between AI exposure and job growth since ChatGPT 3.5 launched in November 2022. Information-sector firms, which report the highest AI adoption at 42.1 percent, saw labor demand fall 1.9 percent in the first half of 2026, while finance and insurance — at 34.8 percent adoption — recorded a 1.1 percent decline. Professional, scientific and technical services, by contrast, grew labor demand 1.2 percent despite 37.7 percent AI adoption.

The offset comes from AI-driven capital expenditure. Non-residential construction added 95,000 jobs year-to-date and AI-related manufacturing added 32,000, together accounting for roughly 127,000 positions — about 25 percent of all private-sector job gains in 2026. The data center buildout is converting AI capex from a technology-company line item into construction and manufacturing payrolls.

The flat employment picture in high-exposure industries may partly reflect post-2019 overexpansion rather than AI displacement alone, Juneau noted. Total hours worked also show no clear deterioration tied to AI exposure, suggesting companies have not yet cut both headcount and hours simultaneously. The pattern echoes earlier technology transitions: the PC productivity boom of the 1990s and the offshoring wave of the 2000s both took years to show up in aggregate employment data before reshaping specific occupations.

The Census Bureau's Business Trends and Outlook Survey data complicates a simple negative correlation. Education services, at 34.6 percent AI adoption, held labor demand roughly flat, while construction — the lowest adopter at 12.3 percent — grew demand 0.9 percent. Manufacturing, at 17.9 percent adoption, was unchanged. Services account for nearly 84 percent of US private-sector employment, which is why the white-collar focus of AI displacement carries outsized economic significance compared with earlier automation waves that targeted factory floors.

Young Graduates Face the Steepest Entry Barrier

The most visible pressure point sits at the bottom of the white-collar ladder. Unemployment among college graduates aged 22 to 27 has risen since its 2023 low and remains above 2019 levels, even as the overall graduate unemployment rate improved after tariff-related uncertainty faded. Juneau attributed part of the increase to last year's tariff policy uncertainty, but noted little improvement even as that uncertainty receded. Entry-level roles concentrate on standardized, decomposable tasks — document preparation, data compilation, routine analysis — precisely the work AI handles most readily. If firms continue substituting AI for these functions, the barrier to a first white-collar job rises, with consequences extending beyond near-term unemployment into skill accumulation and career trajectory.

AI Capex Creates a Construction and Manufacturing Counterweight

The data center buildout has become a meaningful employment engine. Non-residential construction added 95,000 jobs year-to-date, with data center construction a primary driver, while AI-related manufacturing added 32,000. Together these represent roughly a quarter of 2026 private-sector job gains. Manufacturing firms serving AI infrastructure have outperformed non-AI manufacturers in employment growth, and capital expenditure plans continue to be revised upward, suggesting the offset will persist in the near term.

The report's central implication reframes the AI employment debate from aggregate collapse to structural redistribution. High-exposure industries are not shedding workers en masse, but they are not hiring either. The gap between the speed at which AI replaces tasks and the speed at which workers adapt to new ones will not show up in headline payroll numbers — it will show up in first jobs, promotion paths and job content. For investors, the evidence points to sector and demographic differentiation rather than a broad directional bet on the labor market.

This article is for informational purposes only and does not constitute investment advice.