Goldman says AI disruption will have a 'scarring effect' and lead to a yearslong pay cut for displaced workers

Goldman finds tech-displaced workers take an average 3% real earnings cut and experience 10 percentage points less cumulative real earnings growth over the following decade. They also take about one month longer to find a new job, face elevated unemployment risk for the next 10 years, and Goldman estimates AI reduced new job growth by ~16,000 payrolls/month and could displace up to 7% of US workers over 10 years. Retraining helps: workers who retrained after tech-driven displacement saw ~2 percentage points higher cumulative real wage growth and ~10 percentage points lower unemployment probability over the next decade.

Analysis

AI-driven displacement will not be a symmetric shock: gains in corporate productivity concentrate at the top of firms and geographies while consumption and balance-sheet formation are most harmed in mid-skill cohorts and second-tier tech hubs. That bifurcation creates durable regional pockets of weaker mortgage demand and slower household formation, which can erode CRE and regional bank cashflows even as national GDP holds up.

From a competitive standpoint, vendors that convert headcount savings into recurring software and cloud spend (AI model providers, orchestration layers, tools that embed model outputs into workflows) capture most upside, while businesses monetizing mass low-to-mid-skill labor (traditional staffing, brick-and-mortar retail, legacy back-office outsourcers) face secular margin pressure. Reskilling platforms and IT staffing firms sit at the intersection — they win if retraining budgets scale, lose if firms prefer automation-only playbooks.

Key catalysts that will magnify or reverse these dynamics are macro (recession vs. tight labor market), policy (scaled retraining subsidies or unemployment insurance reforms), and corporate behavior (how much productivity savings are recycled into wages vs. buybacks). Effects play out on different clocks: hiring patterns and payroll prints move in months; wage-scarring, homeownership and CRE reallocation run for years.

Contrarian angle: the market largely prices AI as a tech-capex story while underweighting the services layer that will consume most of the increased developer/user productivity (embedding, monitoring, compliance, retraining). If retraining adoption accelerates, expect faster reallocation of labor into higher-value roles and a compression of the currently anticipated long-run human-capital drag — a catalyst that can re-rate software/service integrators faster than hardware names.

Key Decisions for Investors

• Long NVDA (or NVDA 6–9 month call spread): play durable AI infrastructure pricing power. Entry: on any 5–10% pullback; target 30–60% upside, stop 40% downside from entry (volatility-aware).
• Long ASGN (ASGN) or buy ASGN 12-month calls: hedge between short-term staffing disruption and longer-term IT reuptake from retraining budgets. Horizon 6–18 months; asymmetric payoff if corporate tech hiring resumes.
• Short KRE (regional bank ETF) or a basket of regional banks with heavy mortgage/CRE exposure (e.g., ZION, CFG) — horizon 3–12 months. Risk: deposit flight or rate moves can flip the trade; target 20–40% downside vs defined risk sized to 1–2% NAV.
• Buy a 12–18 month put spread on an office-heavy REIT (e.g., VNO or SLG) to express prolonged downsizing of office footprints. Structure to limit premium and target 2–3x payoff if office cashflows compress materially.
• Pair trade: long MSFT (or GOOGL) vs short KRE — thematic long AI/software monetization vs short regional bank/CRE consumer-weakness exposure. Use 6–12 month horizon; expected asymmetric payoff if productivity wins coincide with localized demand erosion.