The AI "Capex Trap" Is Wall Street's Biggest Fear Right Now. Here's Why I Think It's Overblown.

Amazon plans $200B in capex this year after $132B in 2025; 2025 revenue rose 12% (AWS +20%), free cash flow was $11B (down from $38B in 2024), and book value increased to $411B from $286B. Alphabet plans $175–185B in capex (vs $91B in 2025); 2025 revenue grew 18% (Google Cloud +48%), free cash flow was $73.3B (vs $72.8B), and book value rose to $180B from $125B. The piece argues these large AI-related capex programs have coincided with improved balance sheets and book values, implying the spending is accretive and could present buying opportunities rather than reasons to sell.

Analysis

Hyperscaler-led capex is creating concentrated, multi-year demand pockets that amplify upstream margins for a narrow set of suppliers while compressing the economics of second-tier vendors. The concentration raises bargaining leverage: large cloud builders can steer design wins toward preferred partners and lock in preferential supply, which accelerates vertical integration and raises the effective entry cost for new competitors. Expect incremental demand to be lumpy by vendor and geography — order books will skew to firms that can guarantee scale, speed, and service-level agreements.

A meaningful tail risk sits in rapid efficiency improvements at the software and model level that materially lower per-inference hardware needs; that mismatch (supply ramp versus demand-efficiency gains) can create a 6–18 month window of price-led destocking. Macroeconomic shocks or tighter financing cost could force deferred capex into cancellations or slower rollouts, turning a multi-year growth story into a valuation reset for overly levered suppliers. Regulatory interventions (export controls, antitrust) are lower probability but would shift order flows and rerate regional winners and losers over 12–36 months.

For positioning, prioritize balance-sheet optionality and operating leverage protection rather than pure hardware exposure; favor owners of demand (cloud providers) and vertically integrated suppliers who capture both recurring revenue and implementation margins. Use duration layering: trade near-term volatility with options and capture directional exposure with pair trades that express relative-share gains. Monitor three short-term readouts — order cadence, pricing on spot GPU/accelerator markets, and colo availability — as the quickest signals of demand durability and margin pressure.

Key Decisions for Investors

• GOOGL — Buy a 12-month call spread (buy 1x 12m ATM call, sell 1x 12m +25% call). Timeframe: 9–12 months. R/R: limited downside = premium paid (target <3% notional), upside = 30–60% if re-rating and cloud monetization accelerate; exit/trim on ad-revenue miss or if spread premium compresses >50%.
• AMZN / INTC pair — Long AMZN equity (2% portfolio) vs short INTC equal notional for 6–12 months. Timeframe: 6–12 months. R/R: asymmetric — expect 1.5–3x relative upside if hyperscalers capture more value; cap losses by sizing to max 15% absolute drawdown per leg and use a 12% relative stop between legs.
• NVDA — Sell 30–60 day covered calls or call spreads to monetize elevated implied volatility while owning the equity (or sell calls if long exposure is not desired). Timeframe: roll monthly for 3 months. R/R: collect 2–5% premium per month; protect tail risk with OTM puts (10% downside) to cap gap risk; avoid naked short duration risk beyond 60 days.
• Tactical short (select small-cap AI infra names) — Buy 6–9 month put spreads on over-funded/low-cash runway suppliers (name-specific selection required). Timeframe: 3–9 months. R/R: limited debit risk (100% of premium) vs outsized payoff if order reversals or pricing collapses; close on signs of renewed multi-quarter contract visibility.