How Will AI Affect the US Labor Market?

6-7% of workers could be displaced under Goldman Sachs Research's baseline 10-year AI adoption, with a 0.6 percentage-point rise in unemployment if the transition is spread over a decade and materially larger impacts if frontloaded. GS finds ~300 million jobs globally exposed to AI and that AI could automate tasks equal to ~25% of U.S. work hours; construction jobs tied to data-center build-out have risen by ~216,000 since 2022 and ~500,000 net new U.S. power-sector jobs are needed by 2030. Goldman expects U.S. unemployment to inch to 4.5% (from 4.3% in January) and warns that earlier job losses could cause underperformance versus forecast and influence Fed policy toward rate cuts.

Analysis

AI-driven demand for compute and electrification is a multi-year capex wave that will reallocate profits toward asset-heavy providers (data-center landlords, power-equipment suppliers, utilities) rather than labor-heavy service providers. If adoption is backloaded over a decade, the market prizes steady durable cashflows; if adoption is frontloaded, expect a concentrated, 12–24 month surge in hardware orders that tightens component supply and forces near-term margin divergence across suppliers. A frontloaded scenario also creates a distinct monetary channel: a transient unemployment uptick could precipitate earlier-than-expected Fed easing, which would re-rate long-duration growth and REIT cashflows even as wage pressure in skilled trades keeps cyclicals supported. Tariff uncertainty and semiconductor supply risk are the primary levers that can compress or extend this cycle — each can swing capex timing by quarters-to-years and therefore alter which securities capture the upside.

Second-order supply effects matter for trade selection: rising data-center build lifts not just hyperscalers but copper, switchgear, and medium-voltage transformer makers and their logistics partners; simultaneous labor tightness increases subcontractor pricing power, widening margins for distributors and tool retailers. Conversely, incumbents whose revenue maps closely to routine human interaction (outsourced call centers, gig creative platforms) face both demand substitution and margin compression from AI tools, creating asymmetric downside. Watch policy/corporate hiring signals and semiconductor OEM order books as high-frequency indicators — they will lead labor prints and capex guidance by 3–6 months.

Key Decisions for Investors

• Long NVDA (or buy 12-month $650 calls) — play frontloaded compute demand. Position size 3–6% notional; target 30–100% upside if order wave sustains; hard stop if NVDA falls 35% from entry or if semiconductor inventory days climb >25% sequentially.
• Long data-center REITs (EQIX, DLR) for 12–36 months — buy into pullbacks; expect lease-up and pricing power to outpace general REIT indices if AI capex continues. Use a 6–8% allocation, take profits on a 25–40% rally or if cloud capex guidance turns negative for two consecutive quarters.
• Pair trade (bullish infrastructure / bearish labor-substitutable services): Long FAST or HD (exposure to construction supply chain) vs short TTEC or FVRR for 6–12 months. Rationale: rising specialized construction demand boosts distributors while AI disintermediates outsourced service providers. Keep pair delta-neutral; unwind if unemployment rises >0.5ppt and retail sales collapse.
• Long utility/grid modernization names (NEE, EXC) — 12–36 month horizon to capture regulated returns on electrification. Size 4–7%; tail risk is regulatory pushback or deflationary technology reducing power intensity per compute dollar. Close or hedge if renewable permitting or transmission build-out slows materially for two consecutive quarters.