What CEOs say about AI and what they mean about layoffs and job cuts: Goldman Sachs peels the onion

Goldman Sachs senior economist Ronnie Walker finds that while AI has a modest, unproven role in economy-wide layoffs, firms that discuss AI in workforce contexts have sharply cut job openings and AI-related layoffs rose to just above ~15% in the quarter. Corporate commentary across nearly all S&P 500 firms shows softer labor market conditions, reduced wage concerns, and targeted hiring pullbacks—especially at companies with high tariff exposure—while S&P 500 real revenues (ex-energy) rose 4.1% year-over-year versus 2.2% real GDP growth. Retail exposure is bifurcated: same-store sales grew ~0.2% for lower-income zip-code retailers versus ~2.5% for middle/high-income, implying persistent downside risk for low-end consumers into 2026.

Analysis

Market structure: Winners are AI-capex suppliers and cloud/software firms (think NVDA, MSFT, AMZN) and premium consumer plays (AXP benefits from high‑income spending); losers are low‑end retailers and labor‑intensive manufacturers exposed to tariffs. Preemptive hiring cuts transfer near‑term demand from payrolls to capex, improving unit economics for automation vendors while compressing volumes for front‑line retail. Labour slack reduces wage inflation upside, which should lower longer‑run terminal rate expectations if sustained.

Risk assessment: Tail risks include rapid regulatory limits on model deployment, a demand shock from broad layoffs, or tariff escalation; each could knock 3–7% off affected sector revenues in 6–12 months. Immediate signals (days–weeks) are earnings commentary and weekly job openings; medium (3–6 months) is JOLTS/payrolls and capex cadence; long term (2026+) is structural “jobless growth.” Hidden dependency: AI adoption raises concentrated GPU/semiconductor demand and supply chokepoints.

Trade implications: Tilt portfolios into semiconductor hardware and cloud software with size limits (2–3% notional per position) and pair that with shorts in low‑end retail (XRT or specific names) and tariff‑exposed industrials. Buy duration (5–7y tenor) if monthly inflation prints remain <0.2% m/m for two prints. Use 3–6 month call spreads on NVDA and 3–6 month put spreads on XRT to express asymmetric risk/reward around earnings and JOLTS.

Contrarian angles: Consensus underestimates second‑order demand loss from preemptive hiring freezes — shorting low‑end discretionary now priced as resilient may be premature. Conversely, market may be underpricing near‑term margin pressure for small AI vendors that must spend heavily on talent/capex before revenue recognition. Watch for a paradox where stronger index EPS from large AI winners masks deteriorating breadth and consumer weakness in 2025–26.

Key Decisions for Investors

• Initiate a 2% portfolio long in NVDA via a 3–6 month call spread (buy 20% OTM, sell 40% OTM) sized to cap max loss at ~2% of portfolio; rationale: sustained GPU demand from enterprise AI; exit or roll if NVDA misses revenue guide by >5% or implied vol rises >40% from today.
• Establish 1–2% long position in AXP (American Express) with a 3–9 month horizon to capture high‑income consumer resilience; trim if travel/merchant spend ex‑fuel falls >3% sequentially or same‑store sales for upper‑income retailers decline by >200 bps.
• Short low‑end retail exposure: 1.5% notional short XRT or buy 3–6 month put spread on XRT (10–25% OTM) to express downside from jobless growth and tariff pain; cover if JOLTS openings recover >8% month‑over‑month or same‑store sales for low‑income retailers exceed +1% for two consecutive quarters.
• Increase duration exposure: add 2–3% notional to 5–7y Treasury futures or IEF; objective to gain from lower wage inflation and softer terminal rate expectations; unwind if CPI m/m prints >0.3% for two consecutive months or payrolls exceed 300k in a single month.
• Implement a monitoring trigger: track Goldman’s weekly ‘AI mentions in workforce’ metric and job openings — if AI workforce mentions >20% of S&P 500 transcripts and job openings fall >10% year‑over‑year among AI‑mentioners, increase short exposure to labor‑intensive discretionary names by another 1–2%.