Better Stock to Buy Right Now: Amazon vs. Costco

Amazon reported Q4 2025 net sales of $213.4B (+14% YoY) and earnings of $21.2B (+6%), but shares have lagged amid management’s planned ~$200B capex for AI infrastructure; forward P/E ~26. Costco posted fiscal Q2 2026 net sales of $68.2B (+9.1% YoY) and net income of $2.04B (+13.8%), with membership renewal at 92.1% after fee hikes and a forward P/E ~48. The author favors Amazon for long-term investors due to cheaper valuation and a material AI tailwind despite Costco’s defensive qualities and premium pricing.

Analysis

Winners extend beyond the obvious AMZN/NVDA axis: the $-intensive buildout of hyperscale AI capacity creates a multi-year demand waterfall for GPUs, custom silicon, rack-level power distribution, and colocation capacity. That favors NVDA and server integrators, stresses legacy foundry/OSAT capacity, and creates a secondary rally path for utilities and copper-exposed names if build schedules accelerate over the next 12–24 months.

Second-order losers include low-margin consumer suppliers tied to bulk retail channels and small marketplace sellers who face rising fulfillment fees as Amazon directs more spend into owned infrastructure; Costco’s supplier base could see margin pressure if it leans harder on private-label sourcing to defend price points. Also watch commercial real-estate and regional power grids near major cloud campus expansions — localized input inflation and permitting delays are a real operational risk that can push AWS monetization timelines out by quarters.

Key risks and catalysts: execution risk on the capex program (cost overruns, power/permits, or slower appliance of agentic AI) is the clearest path to further multiple compression in AMZN over 6–24 months, while a macro-driven drop in discretionary bulk spending or a membership churn inflection is the immediate threat to COST over 3–9 months. Regulatory or geopolitical constraints on GPU supply (export controls, China restrictions) are binary catalysts that would re-rate NVDA and reprice the AI optionality embedded in AMZN.

Contrarian read: consensus frames AMZN as a cheap AI call and COST as a defensive safe haven; I view that as an asymmetric trade corridor — AMZN’s optionality is real but back-loaded and execution-sensitive, while COST’s premium already discounts near-perfect consumer resilience. The most efficient way to express that view is a calibrated pair trade that shortens lottery-ticket exposure and buys concentrated optionality on AI upside.

Key Decisions for Investors

• Pair trade: Long AMZN / Short COST (size 1.0 / 0.6) with a 6–12 month horizon to express AI optionality vs defensive premium. Trim/stop if AMZN falls >20% from entry or COST outperforms by >15%; target asymmetric payoff of ~2–3x if AWS monetization accelerates within 12 months.
• Buy NVDA call spread (3–9 month expiries) to capture near-term GPU tightness while capping premium paid — starter position 1–2% NAV. Close or widen if export-control headlines pop or NVDA IV spikes >40% above history.
• Long-dated AMZN LEAPs (12–24 months) funded by selling 3–6 month calls to reduce cost; treat as optionality play sized for 2–4% NAV with a multi-year payoff. Accept near-term dilution of return if AWS capex cadence slips; cut if realized free-cash-flow turns negative for two consecutive quarters.
• Hedge Costco exposure: modest long-dated puts on COST (9–12 months) as insurance against membership churn or a macro shock — allocate 0.5–1% NAV. This converts the pair into a more convex position if consumer weakness materializes.