US macroeconomic overheating and capacity constraints Report Interpretation

Overview

Goldman Sachs examines whether inflationary pressure from the AI buildout is creating hidden pockets of overheating in the US economy. Its conclusion is that aggregate and sector-level evidence points to limited, localized constraints rather than broad-based overheating.

Core views

Goldman Sachs begins from the view that the macroeconomy is not broadly overheating. Its labor-market slack tracker, a composite of ten utilization measures scaled to the unemployment rate, is 1 percentage point higher than at the end of the prior cycle, when the Federal Reserve struggled to sustain 2% inflation. Wage growth is running below the pace the report estimates is compatible with 2% inflation, while aggregate industrial capacity utilization is not particularly high. The central question is whether AI-related investment is nevertheless creating narrower resource constraints that could add inflation pressure. The labor-market evidence indicates few such pockets. Goldman Sachs calculates industry-level jobs-workers gaps as job openings minus unemployed workers divided by each sector’s labor force. In most industries, this gap has fallen below pre-pandemic levels. Wholesale trade, healthcare and professional services still show somewhat elevated gaps, but these have eased substantially from pandemic peaks and have not renewed their earlier pressure. A separate measure derived from Beige Book language on shortages and wage pressure is roughly at its historical average across all 12 Federal Reserve Districts, arguing against regional labor shortages. Wage pressure is also not broad-based: approximately 35% of employment is in industries with nominal wage growth above 4%, compared with a 25% average in 1990-2019 and a roughly 90% peak in 2022; only 5% of industries are above 6%. Manufacturing capacity constraints also appear concentrated. Most industries are operating well below maximum potential output, even using a stricter ceiling based on each sector’s highest observed utilization rate during recent business cycles. Electrical equipment and machinery manufacturers are notable exceptions and are likely benefiting from AI-infrastructure demand. More granular production data show rapid growth in batteries, communication wire and cable, and power-transmission equipment, consistent with these sectors approaching capacity limits. However, overtime hours—an indicator that historically leads capacity-utilization increases by three to six months on average in a panel of about 20 major manufacturing industries—remain below recent-cycle peaks in most sectors. The report therefore sees limited pipeline risk of a broader manufacturing capacity squeeze. For services, Goldman Sachs builds alternative capacity-utilization measures across seven industries representing about 30% of US GDP, using operational indicators such as transportation load factors, warehouse storage use, retail sales per square foot, billable-hours workweeks, hotel occupancy, real-estate vacancy rates and construction-equipment operating hours. These measures are validated against sectoral price indexes: higher hotel occupancy and rail freight load factors have generally coincided with rising relevant CPI and PPI inflation. Most services industries remain below their utilization peaks of the past two decades. Professional and business services are the main exception, with consulting employment picking up since late 2025 as firms help clients manage the AI transition. The GDP-weighted services utilization index has returned to its pre-pandemic level. Although the ISM semiannual survey shows higher utilization, the latest ISM commentary expects robust capacity expansion through the rest of 2026, which Goldman Sachs says should lower that measure and narrow the difference. Adding the services-capacity index to a Phillips curve model that already includes the services unemployment rate improves the model’s ability to forecast services inflation. With capacity constraints only slightly above their long-run average, the report estimates they are adding about 10bp to year-over-year core services ex-housing PCE inflation, far below the 30-40bp contribution estimated for 2021-22. Finally, Goldman Sachs combines labor, manufacturing and services indicators into a composite bottlenecks tracker measuring the share of industries with particularly elevated jobs-workers gaps, wage growth or capacity constraints. The tracker has risen slightly in recent months, largely because the jobs-workers gap widened in a few subsectors after labor-market weakness in the second half of 2025. It nevertheless remains in line with its pre-pandemic level. The report concludes that pressures have not become unusually broad-based and that overheating pockets remain limited so far.

Analysis framework

The report tests the overheating question sequentially across industry labor tightness, regional Beige Book commentary, wage-growth breadth, manufacturing utilization and overtime, and alternative operational measures of services capacity. It then aggregates those measures into a composite bottlenecks indicator and relates services utilization to inflation through a Phillips curve model.

Methodology notes

• Industry AnalysisSupply-demand framework

  Sector-level capacity and labor-tightness analysis

  The report compares sectoral labor gaps, wage growth and utilization with historical levels to judge whether supply capacity is binding relative to demand.

• MacroeconomicsPhillips curve

  Services inflation model augmented with capacity utilization

  Goldman Sachs adds its services-capacity index to a model already using the services unemployment rate to estimate the inflation effect of capacity constraints.

• Other

  Composite bottlenecks indicator

  The report combines the share of industries with elevated labor gaps, wage growth or capacity constraints to assess whether pressures are broad-based.

Key data

• Labor-market slack tracker1pp higher than late last cycleCompared with the period when the Fed struggled to sustain 2% inflation.
• Industries with nominal wage growth above 4%About 35%Above the 25% average over 1990-2019 but below the roughly 90% peak in 2022.
• Industries with nominal wage growth above 6%5%Shows limited breadth of very rapid wage growth.
• Services industries covered by alternative utilization measuresSeven industries representing about 30% of GDPThe measures include transportation, warehousing, retail, professional services, accommodation, real estate and construction.
• Estimated services-capacity contribution to core services ex-housing PCE inflationAbout 10bpYear-over-year contribution, down from 30-40bp in 2021-22.
• Overtime lead over manufacturing utilization3-6 months on averageEstimated using about 20 major manufacturing industries over the last two decades.

Impact & implications

The report’s indicators suggest that current inflation risk from capacity constraints is modest at the economy-wide level. AI-related demand may be tightening electrical equipment, machinery and some professional-services activity, but these pressures are not yet broad enough to signal generalized overheating.

What to watch

• Whether AI-related demand further pushes electrical equipment and machinery utilization toward capacity limits.
• Whether overtime hours rise across manufacturing sectors, which historically precedes higher utilization by 3-6 months.
• Whether professional and business services remain near capacity as consulting activity supports AI adoption.
• Whether the composite bottlenecks tracker continues to rise beyond its pre-pandemic range.
• Whether expected services-sector capacity expansion through the rest of 2026 materializes.