The AI Chip Split - And Why Amazon Is On The Right Side

AMZN is trading at $212, down 14% since prior coverage; the analyst maintains a Buy despite a Q4 EPS miss. Management guided to $200B capex by 2026, described as demand-driven as AWS accelerates. The stock is described as cheap at 26.8x forward P/E, with upside cited from AWS partnerships (Anthropic, OpenAI) and custom silicon (Trainium).

Analysis

The market is treating the recent guidance as a binary signal on capital intensity rather than a multi-year strategic pivot; that misread opens a window for active positioners. A demand-driven, multi-year data‑center build disproportionately benefits layer‑1 infra suppliers (switching, storage, memory, custom silicon fabs) while creating negative optionality for OEMs that rely on broadly distributed enterprise refresh cycles. Expect tighter component lead times and higher spot pricing for GPU/DRAM in the next 3–9 months even if AWS levers Trainium/Inferentia to displace some third‑party GPU spend.

Tail risks cluster around macro-driven capex pauses and model performance/regulatory shocks to the AI stack. In days, earnings and guidance updates will drive volatility; in months, supplier order cadence and foundry allocations will reveal true capex stickiness; in years, the P&L impact of custom silicon (higher gross margins, lower third‑party royalty leakage) will compound. A sharper-than-expected enterprise pullback, or Trainium failing to meaningfully cut total cost of ownership vs GPUs, would flip the narrative quickly.

There is a clear tactical taxonomy for positioning: alpha from supply‑chain beneficiaries, convex optionality in AMZN equity via LEAPS or defined‑risk spreads, and hedged pair trades to isolate infrastructure vs GPU demand. Risk management should be calendarized to milestone cadence (next two earnings, next AWS re:Invent, quarterly supplier bookings) rather than headline price moves. The consensus is pricing the story as either full doom or full glory; the highest-probability outcome is middling execution with asymmetric upside if AWS demonstrates 20–30% lower inference $/token versus GPU peers within 12–18 months.

Key Decisions for Investors

• Buy AMZN stock, time horizon 12–24 months; target asymmetric upside of ~30–50% if AWS margin capture accelerates. Hedge with 12–18 month puts (buy puts at ~20% OTM) to cap drawdown to ~15–20%.
• Buy AMZN Jan 2027 LEAPS calls (or equivalent 18–30 month call spread) to capture multi‑year margin optionality; size 2–4% NAV. Use a sell‑to‑close or roll strategy if AWS QoQ revenue growth decelerates below company guide on two consecutive quarters.
• Long supply‑chain beneficiaries (examples: ANET, SMCI, MU) for 6–12 months to play data‑center kit demand; pair 30–50% notional with a small NVDA short to isolate exposure to AWS custom silicon substitution. Take profits/ reassess at supplier booking inflection points (quarterly).
• Income/volatility play: sell AMZN 3–6 month puts at strikes ~10–15% below current levels to collect elevated IV; size so max assignment equals desired long entry. Cap downside with bought puts 20–25% OTM to define worst‑case loss and preserve optionality.