Chief people officers—and Jamie Dimon—say AI can’t learn ‘human skills.’ The world’s youngest self-made billionaires want to prove them wrong

San Francisco startup Mercor, valued at $10 billion last November and led by young founders now described as self-made billionaires, is hiring tens of thousands of contract experts to train AI systems on judgment and nuance, claiming it paid out about $2 million daily with an average hourly rate of $86 and 30,000 contractors hired last year. The company launched an AI Productivity Index (Apex) benchmarking models in medicine, consulting, banking and law—reporting GPT-5 top scores around 64.2% (as low as 59.7% in investment banking)—and argues that models exceeding ~60% production ability can reshape work even as WEF forecasts 39% of core skills disrupted by 2030 and 40% of firms planning workforce reductions due to AI.

Analysis

Market structure: Winners are AI infrastructure and cloud providers (NVDA, MSFT, AMZN) and data-center REITs (DLR, EQIX) because Mercor-style human-in-the-loop scaling multiplies GPU, storage and cloud consumption; losers include staffing/temporary-placement firms (RHI, MAN) and entry-level service providers as automation replaces routine white-collar tasks. Competitive dynamics favor firms that control model training pipelines and compute pricing power—expect higher gross margins for GPU/cloud vendors and price pressure for commoditized labor suppliers within 6–24 months. Cross-asset: stronger profit margins and productivity gains are disinflationary long-term (downward pressure on nominal yields), tightening credit spreads for tech leaders while raising idiosyncratic equity volatility around AI-readiness catalysts.

Risk assessment: Tail risks include regulatory intervention (EU AI Act, US gig-classification) or a rapid labor-demand contraction causing a 2–4 quarter consumer demand shock that could cut corporate revenues 3–7% in affected sectors. Short-term (days–months) risks are execution and model safety incidents that trigger selloffs; medium-term (6–18 months) risks are litigation and labor-classification rulings increasing operating costs by an estimated 10–25% for gig-centric firms. Hidden dependencies: proprietary data access, licensing terms with cloud/GPU suppliers and contractor retention; catalysts include major model releases, earnings beats from NVDA/MSFT, or adverse NLRB rulings.

Trade implications: Direct plays—overweight NVDA and MSFT/AMZN (cloud) for 3–6 month to 2-year horizons; underweight/short staffing names (RHI, MAN) and office REITs (SLG, VNO) as entry-level displacement reduces demand. Pair trade—long NVDA (2–3% portfolio) vs short RHI (1–1.5%) to capture divergence in revenue drivers over 3–12 months. Options—buy 3–6 month call spreads on NVDA to capture upside while funding with short calls or buy 12–24 month LEAPS on MSFT (10–15% OTM) to play durable cloud AI. Rotate capital from commercial office REITs into data-center REITs (DLR, EQIX) over next 6–18 months.

Contrarian angles: Consensus underrates that human-in-the-loop demand will create a high-skill premium—specialist contractors may command 2–4x wages, supporting software tools that coordinate them (a pick-and-shovel opportunity underowned by public markets). The market may be overpricing perpetual NVDA multiple expansion but underpricing recurring revenue capture by cloud providers and data-labeling platforms that can lock customers long-term. Historical parallel: manufacturing automation compressed low-skill jobs but ultimately expanded higher-skill roles and service ecosystems over 5–10 years; unintended consequences include regulatory backlashes and unionization that could flip winners to losers quickly if labor rules change.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in NVDA over the next 2 weeks, funded via a 3–6 month call spread (buy near-term calls and sell higher strike) to capture continued GPU demand while limiting premium spend; add on pullbacks of 10%+.
• Allocate 1.5–2% to MSFT via 12–24 month LEAPS (10–15% OTM) as a core cloud/AI tooling exposure; if MSFT outperforms by >8% in a month, trim 25% to lock gains.
• Initiate a 1–1.5% short position in Robert Half (RHI) or Manpower (MAN) to reflect structural headwinds to staffing demand; hedge with a 3–6 month put spread and target to cover if RHI/MAN trade down 15% or after 6 months.
• Reduce office-REIT exposure (SLG, VNO) by 40–60% within 30 days and reallocate proceeds to data-center REITs (DLR, EQIX) overweight by 3–4% of portfolio, expecting 6–18 month re-rating from higher AI compute demand.
• Monitor regulatory catalysts: if EU AI Act final vote or NLRB gig-worker rulings in the next 30–90 days impose meaningful restrictions, cut long positions in gig-dependent AI firms by 50% within 5 trading days.