Firms and Artificial Intelligence: A Regional Update

The Richmond Fed's Dec. 2025 Fifth District business survey shows 70% of respondent firms provide employees with AI tools (40% public-only; 13% company-provided only) and 56% report using AI in operations. In June 2024, 16% had automated tasks with AI in the prior two years while 45% expected to do so by 2026; aggregate outcomes through Dec. 2025 generally matched or exceeded those expectations. Firms report AI is being used mainly to improve efficiency on tasks (e.g., drafting text, data analysis, generating graphs) rather than to cut labor, suggesting gradual productivity gains as adoption expands but limited near-term operational disruption.

Analysis

Market structure: Near-term winners are hyperscalers and chipmakers that supply AI compute and platform services (NVDA, MSFT, GOOGL, AMZN, AMD, INTC to a lesser degree). SMB IT services, staffing firms, and pure-play task automation vendors without scale are at risk as adoption is task-focused now; cloud/compute pricing power implies incremental AI spend could lift hyperscaler gross margins by +1–3ppt over 12–24 months if adoption follows 2024–25 survey momentum.

Risk assessment: Tail risks include regulatory restraints (EU AI Act enforcement, major privacy suits) and a security/ethics shock that could prompt spending freezes — both plausible over 6–24 months and could subtract 10–25% from forward multiple for exposed names. Hidden dependencies: data pipelines, integration costs, and workforce retraining mean productivity gains likely lag revenue adoption by 6–18 months; catalysts are large enterprise capex announcements or multi-quarter acceleration in cloud AI spend.

Trade implications: Favor infrastructure and security over point solutions — overweight NVDA (compute), MSFT/GOOGL/AMZN (cloud/platform) and PANW/ZS (security), underweight ManpowerGroup (MAN) and select small-cap IT services with <20% recurring revenue. Use options to express convexity: buy 6–9 month NVDA calls (delta 0.35–0.45) on <15% pullbacks; sell covered calls on MSFT to harvest premium while holding exposure.

Contrarian angle: Consensus overweights immediate labor-replacement narratives; surveys show task augmentation dominates so inefficiencies in application-layer startups could persist. Historical parallel: early internet app froth consolidated into infra winners — bias capital to durable moats (chips/cloud/security) and avoid paying multiples for unproven operational AI integrations that may take 2+ years to realize value.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA within 30 days, add more on pullbacks >10%; hedge with 6–9 month calls (buy NVDA 6-month calls, delta ~0.40) to capture continued AI compute demand.
• Allocate 3–4% combined long exposure to MSFT (1.5–2%), GOOGL (1–1.5%) and AMZN (1%) as core cloud/platform plays; implement buy-write on MSFT (sell 3–6 month calls ~5–8% OTM) to monetize theta while retaining upside.
• Short 1–2% position in ManpowerGroup (MAN) or similar staffing firms: thesis is secular task-augmentation reducing low-skill placement demand within 12–36 months; cut if MAN reports revenue growth >5% YoY or staffing rates rise >200 bps sequentially.
• Implement a pair trade: long NVDA (2%) / short INTC (1%) to express secular GPU-led share shift; unwind if NVIDIA revenue growth decelerates below +20% YoY for two consecutive quarters or if Intel posts >$2bn incremental AI revenue in a quarter.
• Increase duration exposure (buy 2–4% TLT or long-dated Treasuries) on a 12–24 month view as successful AI diffusion is deflationary; scale back if 10y yield breaks above 4.0% or CPI prints surprise upside >0.4% month-on-month.