Amazon Stock: Buy, Sell, or Hold?

AWS revenue accelerated 24% YoY to $35.6B in Q4 2025, while Amazon reported net sales up 12% YoY to $213.4B and operating income of $25.0B (+18% YoY); Q1 2026 sales guidance midpoint $176B implies ~13% YoY growth. Trailing-12-month operating cash flow rose 20% to $139.5B, but free cash flow fell to $11.2B from $38.2B due to a $50.7B YoY increase in capex; management expects to spend about $200B in capex this year, largely for AI/AWS capacity. Bottom line: underlying cash generation and accelerating, higher-margin AWS growth support the aggressive AI build-out, but the massive capex and ~30x P/E leave limited room for execution or demand risk.

Analysis

The most important dynamic is not the spend itself but the supply-cycle it creates: hyperscaler-driven front-loading of AI racks will tighten upstream markets for HBM, GPUs and high-end power/cooling gear for a defined window, then flip to structural oversupply as depreciating server fleets and second-hand rack markets materialize. That flip will compress gross margins for chip vendors that price aggressively into the build period and create renewed price competition among cloud providers for monetization of underutilized capacity.

Second-order winners include firms that sell turnkey data-center infrastructure and long-term power contracts; losers are third-party colocation players and legacy ERP vendors that assume steady high-margin cloud pricing. Financially, the fund allocation choice Amazon makes—owning capacity versus leasing—shifts capex from service providers to balance sheets of hyperscalers and increases optionality to monetize scarce high-performance compute through new higher-margin SaaS/AI services.

Key risks are timing and demand persistence: if enterprise AI adoption slows or model architectures pivot away from current GPU-heavy stacks, the payback window on this infrastructure extends materially and equity multiples re-rate. Conversely, persistent high utilization would compound recurring, high-margin revenue and de-risk the capex — outcomes that diverge sharply on a multi-quarter horizon and argue for asymmetric, time-limited exposures rather than undifferentiated long-only positions.

Key Decisions for Investors

• Long AMZN via calendar spread (buy 24-month LEAP calls, sell nearer-term calls for premium) — objective: capture multi-year optionality on cloud monetization while reducing carry; risk: large short-term drawdowns if utilization lags; target 2.5x payoff if AWS monetization accelerates within 12–24 months.
• Paired trade: long AMZN equity / short NVDA equity (size 60/40) for 6–12 months — rationale: isolate cloud revenue share re-rating vs semiconductor multiple risk; stop-loss if NVDA outperforms by >25% relative to AMZN over 3 months or if NVDA announces outsized new supply deals that lock in demand.
• Tactical long INTC 9–18 month call spread — thesis: potential re-entry of hyperscalers into x86/accelerator diversity and domestic-capex incentives could re-price Intel; execution risk high so cap gains targeted 2x with defined debit and limited downside.
• If seeking lower-volatility exposure, buy select utility/infrastructure names with heavy data-center footprints (regulated franchises near major cloud hubs) for 12–36 months — payoff comes from long-term contracted power demand; monitor regional transmission capacity announcements as a catalyst.