Over 80% of companies report no productivity gains from AI so far despite billions in investment, survey suggests — 6,000 executives also reveal 1/3 of leaders use AI, but only for 90 minutes a week

A National Bureau of Economic Research survey of more than 6,000 executives across Europe and the U.S. finds 70% of firms are using AI but over 80% report no productivity or employment impact to date; executives expect modest future gains (productivity +1.4%, output +0.8%) and a small headcount reduction (-0.7%) over three years. Among executives, one-third use AI personally but average only 1.5 hours per week and 25% do not use it at all, underscoring limited near-term ROI despite heavy investment: AI firms captured 61% of global venture capital in 2025 (~$258.7 billion). The data suggest continued enthusiasm and capital allocation into AI but limited current operational benefit, warranting caution for investors pricing rapid productivity or large-scale labor displacement.

Analysis

Market structure: Current data (80%+ of firms reporting no productivity gains, executives using AI ~1.5 hrs/week, expected +1.4% productivity over 3 years) implies demand is front-loaded for tooling and hardware while monetization is lagging. Winners in near term are cloud incumbents and AI infrastructure sellers (MSFT, AMZN, NVDA) capturing CAPEX and services revenue; losers are marginal, over-levered hardware suppliers and early-stage AI service vendors facing high burn if adoption stalls. Cross-asset: weaker productivity outcomes increase equity dispersion, boost tech credit spreads for levered hardware names, pressure commodity memory prices and raise idiosyncratic volatility in options on semiconductors and cloud names.

Risk assessment: Tail risks include rapid regulatory constraints on enterprise AI, a sudden inventory glut in GPUs/memory causing >20% price deflation within 6-12 months, or a high-profile AI failure causing reputational/legal costs. Immediate risks (days/weeks) are knee-jerk repricing around earnings; short-term (3-6 months) is capex pullback; long-term (12-36 months) is structural productivity realization or failure. Hidden dependencies: enterprise deployment costs, data quality, and labor pushback; catalysts include quarterly capex guidance, GPU inventory reports, and major vendor layoffs or wins.

Trade implications: Prefer selective, size-limited long exposure to durable cloud/software moats (MSFT) and short/trim cyclical hardware exposures if inventory or capex guidance deteriorates. Use pair trades (long enterprise software vs short semiconductor-capex suppliers) and volatility-targeted option structures: 9–12 month call spreads on MSFT (10–25% OTM) sized 0.5–3% of portfolio and puts on a semiconductor ETF as portfolio hedges. Time entries around earnings and capex updates; add on >5–10% pullbacks or remove on >15% run-ups.

Contrarian angles: Consensus underestimates backloaded productivity — benefits may concentrate in a few firms that execute integrations, creating winner-take-most economics over 12–36 months. Conversely, market may be underpricing a near-term hardware oversupply; this creates a tactical short window in memory/GPU-linked equities. Historical parallel: early PC-era Solow paradox: multiyear lag before productivity inflection, suggesting patience and selection matter more than broad thematic exposure.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in MSFT over the next 1–3 quarters, buying 50% now and 50% on a >5% pullback; hedge downside with a 9–12 month 8–12% OTM protective put sized to 1% of portfolio if downside protection is required.
• Trim 20–30% exposure to high-multiple, capex-sensitive hardware names (e.g., NVDA, MU, LRCX) if forward EV/Sales >30x or if reported GPU/memory inventory days rise >30% QoQ; redeploy proceeds into cloud/software leaders (MSFT, CRM) with stronger SaaS monetization.
• Construct a pair trade: long 1–2% in enterprise software/cloud (MSFT, CRM) vs short 1–2% in semiconductor-capex suppliers or SMH ETF; add to the short leg if next two quarterly capex guidances decline >15% YoY.
• Implement options trades: buy 9–12 month MSFT call spreads 10–25% OTM sized 0.5–1% of portfolio to capture asymmetric upside if AI monetization accelerates; fund by selling 30–60 day call spreads or short volatility on a semiconductor ETF given anticipated inventory-driven compression.
• Set tactical monitoring triggers for active trades: if GPU/memory wholesale prices drop >15% in 60 days or cloud capex guidance falls >15% at two major providers, increase shorts in hardware by a further 10–15% and rotate into software names within 30 days.