Why AMD Is the Most Underrated AI Stock Right Now

Published March 17, 2026, the author states a preference for buying Advanced Micro Devices (AMD) over Nvidia, arguing AMD could outperform its larger rival for years; morning stock prices cited were from March 11, 2026. The piece is opinion-based with no new financial results or guidance and therefore limited actionable news flow, though it aims to influence investor positioning. Promotional context cites Stock Advisor marketing numbers (total average return 930% vs. S&P 500 187%), and disclosures note Matt Frankel holds AMD and The Motley Fool holds/recommends AMD and Nvidia.

Analysis

AMD's risk-reward now looks driven less by headline AI optics and more by second-order execution: TSMC capacity allocation, multi-chiplet economics (CPU+GPU bundling), and successful penetration of cloud hyperscalers with validated software stacks. If AMD can convert a single hyperscaler GPU design win every 12–18 months, that would force a reallocation of HBM/TSMC capacity and create 10–20% incremental revenue upside per major win, with operating leverage in place to convert a large share to EBITDA.

The main competitive lever that can reverse the bullish case is software lock‑in and ecosystem depth (CUDA + software toolchains). That is a 12–36 month structural risk: even a technically superior part can be priced out if development and migration costs remain non-trivial for ML frameworks. Another asymmetric risk is foundry friction — loss of premium wafer allocation at TSMC could compress ASPs and push AMD into margin-leadership tradeoffs that materially slow outperformance.

Monitorable catalysts and timing: (1) quarter-to-quarter server GPU shipment cadence and invoice recognition (near-term 0–3 months), (2) announced design wins with hyperscalers and SDK maturity signals (3–12 months), and (3) TSMC capacity commitments and HBM supply contracts (12–24 months). Watch gross margin expansion or contraction as the clearest 3–6 month read on whether share gains are translating into durable economics; a sustained 200–400bps margin improvement would validate multi-year upside, while a 300–500bps decline would indicate structural pressure.

Key Decisions for Investors

• Pair trade (12–18 months): Long AMD (+6–8% portfolio) / Short NVDA (~40% notional of long). Rationale: express idiosyncratic share-gain view while hedging macro/AI beta. Target: 30–60% gross return on the pair if AMD captures incremental DC GPU share; stop-loss: -15% on AMD leg or -10% on pair relative move.
• Bullish options (6–12 months): Buy AMD 9-month call spread (delta ~0.40 long, hedge with 1:1 higher strike short) sized to 2–3% portfolio risk. Expectation: 50–100% return if design-win cadence or margin beats materialize; max loss = premium paid.
• Protective hedge for existing longs (0–6 months): Buy 3–6 month 10% OTM puts on AMD sized to cover 25–50% of position or implement a collar by selling 12-month 25% OTM calls to fund puts. Use if headline-driven volatility spikes and you need to lock upside while retaining some participation.
• Event-driven tactical short (3–9 months): Small, conviction-weighted short of companies dependent on HBM supply tightness (watch suppliers or indexed ETFs) if TSMC/packaging confirmations show capacity squeeze — target asymmetric payoff if AMD proves it can outcompete incumbents for wafer/stack allocation.
• Monitoring triggers (risk management): Add to long AMD on (a) two consecutive quarters of data-center GPU shipment growth >20% QoQ or (b) announced hyperscaler validated software stack adoption; materially reduce exposure if gross margins compress >300bps YoY or if TSMC allocation guidance turns negative.