AI will create more jobs than it kills, McKinsey says. The catch: 11 million Americans may need new careers

McKinsey projects AI and automation will reduce demand for roughly 36 million U.S. jobs by 2035, while growth in other occupations creates about 41 million jobs. In its base case, 11 million workers—7% of the workforce—would need to change occupations entirely, requiring about 770,000 cross-field transitions annually, 3.6 times the historical average. Displacement will be concentrated in lower-paid office support, retail, and transportation roles, while healthcare, construction, and management expand; only one in seven affected workers has a direct transition path, and 85% of growing jobs require credentials. Geographic mobility is also constrained because 76% of growing jobs cannot be performed remotely.

Analysis

The investable implication is not broad unemployment; it is a tighter market for credentialed, location-bound labor. That should preserve wage inflation and constrain capacity in healthcare, construction, power-grid work and data-center buildouts, favoring equipment and productivity vendors such as United Rentals (URI), Caterpillar (CAT), Quanta Services (PWR) and Vertiv (VRT) over labor-intensive operators whose margins depend on abundant entry-level staff.

Healthcare is the clearest near-term bottleneck: providers can pass some labor costs through reimbursement, but staffing agencies and lower-acuity operators remain exposed if wage escalation outruns rate updates. A labor-transition cycle also increases the value of accredited training capacity; Adtalem (ATGE), Universal Technical Institute (UTI) and Strategic Education (STRA) are plausible beneficiaries, although the thesis depends on enrollment conversion and employer-funded reskilling rather than headline demand alone.

Consensus may overprice the deflationary labor narrative around AI. In the next 1-3 months, this report is unlikely to move markets, but over 6-18 months the binding constraint becomes deployment: automation vendors gain only where customers can redesign workflows and finance implementation. The better expression is therefore selective long productivity infrastructure versus labor-dependent service models, not a blanket long AI/short labor trade. Falsification would be a meaningful decline in healthcare/construction job openings, easing hourly wage growth, or a capex slowdown in data-center and grid projects.

Key Decisions for Investors

• No immediate index-level trade: treat this as a 6-18 month labor-scarcity theme, not a near-term recession signal. Reassess after the next two JOLTS and Employment Cost Index releases.
• Build a 6-12 month long basket in URI, PWR and VRT on market pullbacks; these companies monetize customer capacity constraints through equipment, electrification and data-center infrastructure rather than direct payroll expansion. Target 15-25% upside versus 10-12% downside; exit if U.S. nonresidential construction starts and data-center capex guidance weaken materially.
• Pair trade over 9-12 months: long ATGE and UTI versus short XRT. Credential scarcity can support enrollment and pricing while retail faces continued labor substitution and store-level wage pressure; size modestly because consumer demand can dominate the spread. Stop if starts/enrollment fail to accelerate over two reporting periods.
• Avoid broad long exposure to healthcare staffing names such as AMN until evidence emerges that wage inflation is easing faster than bill rates. A reversal in temporary nurse demand or provider commentary on internal staffing normalization would be the catalyst to revisit.
• Monitor HCA, UHS and skilled-nursing operators for margin risk rather than shorting immediately: reimbursement resets and patient volumes can offset labor costs. Consider downside hedges only if quarterly labor-cost guidance rises while same-facility revenue guidance is unchanged.