Cathie Wood Thinks AMD Will Challenge Nvidia This Year

Cathie Wood warns AMD will pose tougher competition to Nvidia in the AI market this year, but Nvidia's entrenched advantages — a 92% discrete GPU market share in 2025, a sticky CUDA ecosystem, and successive architectures through Blackwell — make it the default for high-end AI workloads. AMD's data-center mix (nearly half of revenue) and lower-cost Instinct MI300X (~$15,000 vs Nvidia H100 ~ $25,000) give it share gains among major cloud customers, yet analysts still project Nvidia revenue and EPS CAGRs of ~47% and ~45% from fiscal 2025 to 2028. With the AI market expected to grow at a ~30.6% CAGR from 2026–2033 and Nvidia trading near 26x next-year earnings, the piece frames both firms as viable long-term AI plays while favoring Nvidia's durable moat.

Analysis

Market structure: Nvidia (NVDA) remains the incumbent with a de-facto standard (CUDA) and ~92% discrete GPU share in 2025, translating to outsized pricing power (H100 ≈ $25k vs AMD MI300X ≈ $15k). Winners include NVDA, hyperscalers (MSFT, GOOGL, META) that capture AI value-add and HBM/memory suppliers; losers are legacy x86 incumbents weaker on accelerator economics and any low-cost GPU vendors unable to match software ecosystem lock‑in. Global AI TAM growing ~30% CAGR (2026–33) supports multi-player growth, not zero-sum share collapse.

Risk assessment: Key tail risks are renewed export controls or China market curbs (material revenue impact within 0–12 months), an AMD or hyperscaler architectural leap that breaks CUDA stickiness (12–36 months), and HBM/PCB supply shocks pushing component costs +20–40% near-term. Immediate risk drivers (days–weeks) are quarterly guidance and inventory comments; medium-term (months) are AMD product cadence and HBM supply; long-term (years) are regulatory/antitrust actions and software portability trends. Hidden dependency: hyperscalers’ willingness to co-design silicon can accelerate switching despite CUDA inertia.

Trade implications: Core conviction is long NVDA with tactical AMD exposure. Establish a 12–24 month core NVDA position (2–4% portfolio) and a smaller 1–2% AMD satellite to capture share gains; use 9–18 month options to control downside. Pair trade: long AMD / short INTC to express server CPU/GPU secular reallocation. Manage entry/exit with hard stop-losses (NVDA -20%) and profit trims (+30%); scale on quarterly datapoints (market share, HBM supply, export policy).

Contrarian angles: Consensus underestimates hyperscalers’ agency to diversify away from NVDA if open software stacks (e.g., ROCm, MLIR tooling) materially close the portability gap over 24–36 months—this would compress NVDA’s pricing premium. Conversely, market may be underpricing NVDA’s software moat and pricing power in scenarios where HBM remains constrained; asymmetric outcome favors owning NVDA with limited-cost option exposure to AMD upside. Watch for regulatory moves that could bifurcate global stacks and create region-specific winners.

Key Decisions for Investors

• Establish a 2–4% portfolio core long in NVDA (ticker NVDA) with a 12–24 month horizon; if using options, buy 9–18 month LEAPS (example: Jan 2027 calls near-the-money) sized to limit downside to ~2% portfolio; apply stop-loss at 20% below entry and trim 30% at +30% gains.
• Add a 1–2% tactical satellite long in AMD (ticker AMD) to capture data-center share gains; express via a 6–12 month call spread (buy-to-open AMD call, sell higher-strike AMD call) to cap cost—e.g., 1:1 spread sized ≤1.5% portfolio and take profits if AMD gains >1 percentage point discrete GPU/server share in next two quarters.
• Implement a pair trade: long AMD (1% notional) / short INTC (1% notional) to express server secular substitution; rebalance quarterly or close if spread moves >15% or if Intel reports >2% sequential server share recovery.
• If holding NVDA stock, sell short-dated (30–60 day) covered calls when implied volatility >40% or IV exceeds 10% above realized volatility to harvest premium; otherwise hedge with 3–6 month protective puts if nearing earnings with elevated guidance risk.