AI adoption rate rises to 20.6%, with continued strengthening in investment and productivity evidence

Overview

This report updates Goldman Sachs' AI adoption tracker through June 2026. The core conclusions are that AI-related investment remains strong, enterprise adoption continues to rise, labor-market effects are visible but still narrow in scope, and evidence of productivity gains is fairly clear in areas where generative AI has already been deployed. The report focuses on tracking U.S. business surveys, semiconductor and AI hardware investment, employment and hours worked, AI-related layoffs, hiring demand, and industry productivity performance.

Core views

First, the AI investment cycle remains supported by semiconductors and hardware, with global semiconductor revenue expectations, South Korean AI-related manufacturing shipments, and U.S. AI hardware net imports all indicating strong demand. Second, U.S. enterprise AI adoption continues to increase, with large firms still leading, though the pace of adoption among mid-sized firms with 100 to 250 employees has recently accelerated. Third, AI's impact on employment remains concentrated in occupations and industries with already well-defined use cases and has not led to broad labor-market slack, though early negative-correlation signals have emerged for some job growth and hours-worked growth. Fourth, AI has already shown substantial productivity gains in limited deployment scenarios, and official productivity data for high-adoption industries has begun to accelerate slightly.

Analysis framework

The report uses a multi-source data-tracking framework that cross-validates official surveys, national accounts, trade and manufacturing data, company earnings-call text, hiring data, layoff announcements, academic research, and corporate case studies. On the investment side, it focuses on semiconductor revenue expectations, AI hardware investment, South Korean shipments, and U.S. net imports; on the adoption side, it mainly uses the U.S. Census Bureau's BTOS business survey; on the labor side, it tracks employment, unemployment, hours worked, job postings, layoffs, and earnings-call mentions; and on the productivity side, it combines academic research, corporate case studies, and official industry productivity data.

Methodology notes

• Enterprise adoption trackingU.S. Census Bureau BTOS Business Trends and Outlook Survey

  AI usage rate in routine business functions

  Starting on December 4, 2025, the BTOS question changed from asking about AI use in producing goods and services to asking about AI use in any business function, so the change in definition creates a level shift and subsequent adoption rates need to be tracked consistently under the new definition.

• Industry automation exposureGoldman Sachs AI Automation Exposure Score

  Share of work in industry subsectors exposed to AI automation

  The report compares industry AI adoption rates with Goldman Sachs' prior estimates of AI automation exposure by subsector to assess whether adoption rates are in line with the share of work that could potentially be automated.

• Labor-market impactIntegrated tracking of employment, hours worked, layoffs, and earnings-call text

  Marginal impact of AI adoption on job growth, hours worked, and corporate labor commentary

  The report compares AI adoption rates with job growth, hours-worked growth, unemployment rates, AI-related layoff announcements, and keywords from Russell 3000 earnings calls, emphasizing that broad labor-market evidence remains limited at present.

• Productivity validationCross-validation using academic research, corporate case studies, and official productivity data

  Efficiency gains after generative AI deployment

  The report uses academic research and corporate case studies to estimate labor-productivity gains from AI and observes whether productivity growth in high-AI-adoption industries is beginning to improve in official data.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Semiconductors

  One of the core beneficiary assets of the AI investment cycle.

  Strengths

  Global revenue expectations are projected to reach an annualized $826bn by the end of 2026, and revenue forecasts for memory producers have been revised up significantly since ChatGPT's public release.

  Weaknesses

  Revenue expectations have already been revised up substantially, making the sector more sensitive to future demand and capex durability.

  Comparison

  Compared with labor and productivity data, semiconductor revenue expectations and hardware shipments are more direct indicators of AI investment intensity.

  Risks

  If AI capex slows, memory prices decline, or demand expectations are revised down, valuation and earnings expectations may come under pressure.

• AI hardware and data centers

  Reflects the physical investment foundation for enterprise and cloud AI deployment.

  Strengths

  U.S. AI hardware investment is up $425bn versus 2022, South Korean AI-related shipments are at record highs, and U.S. AI hardware net imports remain at $52bn.

  Weaknesses

  The hardware cycle is easily affected by inventories, trade, supply chains, and the pace of capital spending.

  Comparison

  Hardware investment is stronger than the broad labor-market impact and is currently a more certain dimension for observing the AI cycle.

  Risks

  Supply bottlenecks, trade restrictions, data-center power constraints, or tighter enterprise budgets could weaken growth.

• Data-center-related construction employment

  A positive spillover channel from AI investment to the real economy and employment.

  Strengths

  Construction employment in data-center-exposed categories has risen by about 251k relative to trend since 2022, with the latest three months adding about 9k per month.

  Weaknesses

  The employment contribution is concentrated in the construction phase and in specific regions, making it difficult to fully offset all AI-related job displacement pressure.

  Comparison

  It serves as a counterweight to employment drag in marketing, design, customer service, and some technology roles.

  Risks

  If data-center construction slows or financing costs rise, related employment support may weaken.

• Information, professional services, financials, and education sectors

  Leading sectors in enterprise AI adoption.

  Strengths

  More than one-third of firms in these sectors have already adopted AI, with adoption in publishing and some financial subsectors reaching or exceeding 50%.

  Weaknesses

  Higher adoption rates also imply that some roles may be more exposed to automation-driven substitution and organizational restructuring.

  Comparison

  Compared with low-adoption sectors, these sectors are more likely to show productivity gains and changes in job structure first.

  Risks

  If efficiency gains cannot be converted into revenue growth, the impact may show up more as cost cutting and labor pressure.

Key data

• Current U.S. enterprise AI adoption rate20.6%Up 1.1 percentage points from May.
• Expected AI adoption rate over the next 6 months23.9%From the U.S. Census Bureau BTOS survey.
• Large-enterprise AI adoption rate40.7%Firms with more than 250 employees continue to lead.
• Semiconductor company revenue expectation$826bnEquity analysts expect global revenue to reach an annualized level by the end of 2026.
• Increase in U.S. AI-related hardware investment$425bnOn a three-month annualized basis, about 1.3% of GDP above 2022 levels.
• South Korea AI-related hardware shipments$36bn/monthRose to a record high in May 2026.
• U.S. AI hardware net imports$52bnRemained elevated in April 2026.
• Increase in data-center-related construction employment+251kIncrement relative to broader construction-employment trends since 2022; the latest three-month trend is about +9k per month.
• Headcount affected by AI-related layoff announcements38.6kThe number of corporate layoffs attributed to AI continued to accelerate in May 2026.
• Russell 3000 earnings-call mentions24%Companies mentioning AI- and labor-related keywords on Q1 2026 earnings calls.
• Average productivity gain in academic research23%Based on research evidence from already deployed generative AI use cases.
• Average efficiency gain in corporate case studies34%Anecdotal corporate evidence shows efficiency gains slightly above the academic-research average.

Impact & implications

From an investment perspective, the AI theme continues to be supported by hardware, semiconductors, memory, data centers, and enterprise software deployment, with demand evidence not yet showing clear cooling. From a macro perspective, AI's positive impact on productivity is accumulating, but its shock to the labor market remains concentrated in a small number of occupations and industries and has not yet become a broad employment headwind. For asset allocation, investors should simultaneously track the durability of hardware capex, the speed of enterprise adoption diffusion, whether productivity improvements continue to show up in official data, and the social and regulatory risks of AI substituting for labor.

Risks

• The BTOS survey definition changed in December 2025, creating a level shift in the adoption-rate series, so cross-period comparisons should be made cautiously.
• The relationship between AI adoption rates and employment, hours worked, unemployment, or productivity is not equivalent to causality.
• If AI-related layoffs and job substitution continue to accelerate, they may create social, regulatory, and corporate reputation risks.
• Expectations for semiconductors and AI hardware are already strong; if capex or demand falls short of expectations, related assets may pull back.
• The productivity-gain evidence in the report mainly comes from limited deployment scenarios, and whether it can diffuse across entire industries still needs to be verified.

What to watch

• Whether U.S. enterprise AI adoption continues rising from 20.6% toward the 23.9% expected over the next 6 months.
• The speed of adoption diffusion across industries such as information, professional services, finance, education, publishing, and transportation.
• Semiconductor revenue forecasts, forecast revisions for memory producers, South Korean AI hardware shipments, and U.S. AI hardware net imports.
• AI-related layoff announcements and the share of Russell 3000 earnings calls mentioning AI- and labor-related keywords.
• Whether employment growth, hours-worked growth, and official productivity data continue to diverge in high-AI-adoption industries.
• Whether data-center-related construction employment can continue offsetting part of the negative impact from AI-substituted jobs.