College students are voting with their feet on AI. Goldman has the receipts

Goldman Sachs found 2025–26 enrollment in computer science and computer programming majors fell by more than 10%, while healthcare and engineering majors rose about 3% on average, the first statistically significant sign students are shifting away from AI-exposed fields. Goldman estimates AI is cutting roughly 11,000 U.S. jobs per month and says recent college graduate unemployment has diverged sharply higher since 2024. The article suggests a behavioral response to AI risk, with students moving toward lower-automation fields such as nursing and engineering.

Analysis

This is a slow-burn labor-supply shock, not an immediate macro shock. The first-order read is bearish for broad AI-enabler narratives because the marginal student is now rationally avoiding the most crowded, automation-exposed funnels into white-collar work; the second-order effect is even more interesting: the talent pool for future software, fintech, and consulting hiring may tighten just as those firms are using AI to reduce headcount, amplifying wage polarization at the top and hollowing out entry-level training layers.

Healthcare and engineering are the natural refuge trades, but the market should not extrapolate equal benefit across the group. Healthcare is the cleaner medium-term beneficiary because the supply response is constrained by seat capacity and licensing bottlenecks, which makes labor scarcity more durable and supports wage pressure; engineering is more cyclical and capital-intensive, so the increment in enrollment may lag into actual supply for years. That argues for a longer-duration positive view on firms with staffing leverage rather than a generic bet on “education winners.”

For Goldman specifically, the message is mixed-to-negative. The bank’s research credibility rises, but the underlying thesis points to a structurally tougher pipeline for junior hiring in advisory, markets, and wealth—businesses that depend on high-throughput analyst classes. Over 6-18 months, this can become a margin story for incumbents that use AI to compress junior costs, but over 2-4 years it becomes a franchise-risk story if the training pipeline degrades and clients increasingly source more specialized or automated services.

The consensus may be overestimating how quickly students can fully arbitrage this shift. Reallocation into nursing and engineering is rational, but capacity constraints mean the labor-market payoff is delayed, which can create a temporary overshoot in relative valuations of education, training, and healthcare labor names. The more interesting contrarian setup is that the labor market may end up bifurcating: fewer generalists, more credentialed specialists, and a persistent shortage in roles that are hard to automate but also hard to scale through schooling.

Key Decisions for Investors

• Long XLV / short XLY or XLC over 6-12 months: healthcare labor scarcity should be more durable than the AI-disrupted entry-level white-collar funnel; target 8-12% relative outperformance, stop if wage inflation accelerates into margin compression for managed-care and providers.
• Buy LEAP call spreads on HCA or MDT with 9-15 month tenor: these names can benefit from tighter labor supply and stronger pricing power; use spreads to cap premium while targeting a 2:1+ payoff if staffing pressure persists.
• Short basket of AI-exposed white-collar intermediaries (KX, UIPath-adjacent software services, or a basket proxy via XLF/XLI overweights) for 3-6 months: the near-term risk is not revenue collapse but slower entry-level hiring and weaker downstream spending by recent grads; size modestly because the move is gradual.
• Long GS vs short a diversified brokerage/consulting proxy on a 6-12 month horizon: GS benefits from the research relevance and higher AI adoption internally, but the thesis is that firms with more junior-heavy cost structures face more margin pressure than GS; use as a relative-value trade, not an outright long.