Artificial intelligence adoption and economic impact Report Interpretation

Overview

This macro tracker updates evidence through August 2026 on AI investment, business adoption, labor-market effects and productivity. Goldman Sachs finds continued strength in the investment and adoption cycle, while emphasizing that economy-wide labor and productivity transmission remains incomplete.

Core views

Goldman Sachs reports that AI-related investment remains strong. Equity analysts expect annualized global semiconductor revenue to reach $863bn by end-2026, while 2026 gross-profit forecast revisions for AI-exposed global public companies have reached $1.1tn since 2022Q3. In US national accounts, AI-related hardware investment is $485bn above its 2022 level on a three-month annualized basis, equal to 1.5% of GDP. Taiwan’s AI-related hardware shipments have grown 76% since 2022, and US AI hardware net imports reached $55.9bn in June. The report also expects $66bn of global AI-related net exports from AI exporters in August, supporting the view that physical AI investment remains substantial. Business adoption continued to advance: 22.4% of US establishments reported using AI in regular business functions, up 0.9 percentage points from July, and 25.9% expect to use it over the next six months. The information sector leads at nearly 50%, with publishing and computing remaining the leading subsectors. Finance and insurance recorded the largest adoption gains since the last update, while warehousing and storage and miscellaneous manufacturing reported the largest expected increases over the coming six months. Large establishments remain ahead: firms with more than 250 employees have a 40% adoption rate on the latest six-survey average, whereas adoption growth among smaller firms has slowed. Other surveys indicate that 40% of organizations have started deploying and scaling AI agents. The report relates adoption patterns to Goldman Sachs’ subsector AI automation-exposure scores, finding a strong correlation between the two. It cautions, however, that the Census Bureau changed the wording of its AI-use question beginning with the December 4, 2025 release, creating a level shift. Goldman Sachs therefore intends to track adoption under the new methodology going forward. Labor-market effects are present but narrow. The report continues to observe employment drags in activities with established AI use cases, including marketing, graphic design, customer service and some technology occupations. The relationship between adoption and broad labor-market slack remains limited, although it has strengthened recently; the recent negative relationship between adoption and employment growth is still not statistically significant. Corporate layoffs attributed to AI affected 11,000 employees in June, bringing the reported 2026 year-to-date total to 113,000, while 22% of Russell 3000 companies mentioned AI- and labor-related keywords on 2026Q2 earnings calls. The labor effects are not uniformly negative. Construction employment in data-center-exposed categories has risen sharply since 2022, and active AI-related US job postings stand at 89,000. The report also notes that the share of AI-related job postings continues to rise across major developed markets. This offsetting labor demand helps explain why Goldman Sachs characterizes AI’s aggregate labor effect as limited despite pressure in selected occupations. Productivity evidence is stronger in limited deployment settings than in aggregate data. Academic studies suggest an average labor-productivity uplift of about 23%, while company anecdotes suggest average gains of about 32%. Official US data now show a slight acceleration in productivity growth in industries with higher AI-adoption rates, but the report sees only modest evidence of economy-wide effects. It also finds no relationship between company-level AI exposure and margin growth, indicating that documented operating-efficiency gains have not yet translated into a broad cross-sectional margin effect.

Analysis framework

The tracker combines investment and trade indicators, Census Bureau business-adoption data, external organization surveys, labor-market statistics, layoff announcements, job-posting data, academic productivity studies and company anecdotes. Goldman Sachs compares adoption across sectors and firm sizes, relates it to AI automation-exposure scores, and distinguishes localized evidence from broad economy-wide effects.

Methodology notes

• Industry Analysis

  Subsector AI automation-exposure scores

  Goldman Sachs compares sector adoption rates with its estimate of the share of work exposed to AI automation to assess whether adoption is occurring where AI use cases are most applicable.

• Corporate Fundamentals and Finance

  Productivity evidence comparison

  The report compares academic estimates and company-reported anecdotes with official industry productivity data to separate localized deployment gains from broader economic effects.

Key data

• US establishment AI adoption22.4%Up 0.9 percentage points from July; 25.9% expect AI use within six months.
• Information-sector AI adoptionNearly 50%Highest sector adoption rate in the tracker.
• Global semiconductor revenue forecast$863bn annualizedExpected by end-2026.
• AI-exposed 2026 gross-profit forecast revisions$1.1tnCumulative revisions since 2022Q3.
• US AI-related hardware investment increase$485bnAbove the 2022 level on a three-month annualized basis; 1.5% of GDP.
• Taiwan AI-related hardware shipment growth76%Growth since 2022.
• Average productivity uplift23% academic studies; around 32% company anecdotesEvidence applies mainly to areas where generative AI has been deployed.

Impact & implications

The report’s evidence supports an ongoing AI investment and adoption cycle, especially in semiconductors, hardware and data-center-linked activity. It also indicates that realized productivity gains are currently more visible in specific deployed use cases than in economy-wide output, while labor displacement and labor demand are developing unevenly across sectors.

Risks

• The Census Bureau’s late-2025 change in AI-use question wording created a level shift, limiting direct comparability with earlier adoption readings.
• The observed negative relationship between adoption and employment growth is not yet statistically significant.
• Company-level AI exposure has not shown a relationship with margin growth.

What to watch

• Adoption rates under the Census Bureau’s revised survey methodology.
• Whether expected AI adoption in warehousing, storage and miscellaneous manufacturing materializes over the next six months.
• Whether localized productivity gains broaden into measurable economy-wide productivity and margin effects.
• Further changes in labor outcomes in AI-exposed occupations and data-center-linked construction.