PwC’s global chairman says most leaders have forgotten ‘the basics’ as 56% are still getting ‘nothing’ out of AI adoption

PwC’s 29th global CEO survey of 4,454 CEOs across 95 countries reveals a pronounced execution gap in AI: only 10–12% of companies report revenue or cost benefits while 56% say they’re getting nothing, echoing an MIT finding that 95% of generative-AI pilots fail. CEO confidence in 12‑month revenue growth has plunged to 30% (from 38% in 2025 and 56% in 2022), prompting PwC chairman Mohamed Kande to argue leaders must simultaneously run, transform and build new business models and shore up foundations such as clean data, processes and governance. For investors, the survey signals meaningful near-term growth and execution risk across corporates even as PwC frames the period as the start of a longer-term wave of industry reconfiguration driven by AI and innovation.

Analysis

Market structure: The immediate beneficiaries are cloud/platform owners (MSFT, GOOGL, AMZN), GPU/semiconductor leaders (NVDA, AMD, ASML) and data-infrastructure/security vendors (SNOW, MDB, PANW) because PwC’s data shows only ~10–12% are extracting revenue/cost benefits — that tight subset will consolidate share and pricing power. Losers are mid/ small-cap 'pilot' AI vendors and incumbents that haven’t invested in clean data/governance; expect margin compression there and increased M&A as acquirers buy capabilities, not customers.

Risk assessment: Tail risks include rapid regulatory action (EU/US AI rules or liability regimes) that could cut TAM for certain AI services by 10–30%, export controls on chips, or a sudden GPU supply shock; talent and data-quality constraints are second‑order risks that can delay monetization by 6–24 months. Time horizons: immediate (days) for sentiment/volatility around earnings, short-term (3–12 months) for pilot-to-production conversion rates, long-term (1–5 years) for industry reconfiguration and capex cycles.

Trade implications: Favor positions that capture infrastructure and integration (long NVDA via spread, MSFT/GOOGL equities, SNOW) and hedge with cyber names and selective puts; avoid/short high-multiple pure-play AI apps lacking client references. Cross-asset: expect upward pressure on corporate issuance (capex funding), a stronger USD into safe-haven tech flows, and higher electricity/renewables exposure from data-center growth.

Contrarian angles: Consensus overemphasizes model capability; it underestimates data+process integration services — Accenture (ACN) and big consults are likely underpriced optionality. The market may be overstating near-term revenue from incremental AI features while understating multi-year margin capture by platform owners; that dispersion creates pair-trade opportunities over 6–24 months.

Key Decisions for Investors

• Establish a 2–3% notional long position in NVDA via a 3-month call spread (buy ATM call / sell 30% OTM) ahead of next quarterly results to play continued GPU scarcity and data‑center demand; add on a pullback >15%, trim if IV > +40% vs 30‑day average.
• Allocate 2% each to MSFT and GOOGL equities (total 4%) with a 6–12 month horizon to capture cloud AI monetization; reduce by half if either reports forward cloud revenue guidance missing consensus by >3 percentage points.
• Initiate a 2–3% position in SNOW (Snowflake) for 12–24 months to play data-monetization; exit if ARR growth decelerates below 30% YoY or gross margins drop >500bps versus last quarter.
• Establish a pair trade: long ACN 1.5% (systems-integration optionality) and short an equal-weighted basket (total 1.5%) of 5 revenue-negative small-cap AI application stocks (screen: >50x EV/NTM revenue, negative EBITDA) to exploit execution risk; rebalance quarterly.
• Reduce cyclical/legacy-tech exposure by 2–4% and redeploy into renewable power utilities or copper miners (electrification/data-center energy plays) over next 6–18 months to hedge higher data-center energy demand; monitor EU AI Act and US regulatory moves over next 30–60 days—if material liability language appears, de-risk ad/consumer-AI names by 50–100bps of portfolio weight.