Ark Investment's Cathie Wood Just Sold AMD Shares and Bought Nvidia. Is That the Right Move for Investors?

The article argues that AMD has major upside in AI inference and agentic AI, citing two large GPU partnerships worth an estimated $100 billion each and a potential $120 billion data center CPU opportunity. Nvidia is still presented as the AI infrastructure leader, with Q2 revenue guidance of $91 billion implying 95% growth and a forward P/E of 16x fiscal 2028 estimates. Overall, the piece is constructive on both stocks but says Nvidia is the more attractive buy today.

Analysis

The market is still pricing this as a simple share-shift between two GPU vendors, but the more important second-order effect is margin structure: inference and agentic workloads reward total platform breadth, not just raw FLOPs. That favors the company that can bundle silicon, networking, CPUs, and software into a single procurement decision, because hyperscalers will prefer fewer integration points as deployments move from pilots to production. In practice, this means the competitive battle is likely to show up first in gross-margin mix and attach rates, not just unit share.

AMD’s upside is less about displacing the leader in training and more about becoming the default hedge against concentration risk for cloud buyers. The real catalyst is evidence that its large design wins convert into recurring platform revenue over the next 2-4 quarters; without that, the stock’s multiple can remain elevated while fundamentals lag expectations. The key risk is that inference demand may be more price-sensitive than the market assumes, which would compress the TAM narrative and keep AMD in a “good story, slower monetization” box.

Nvidia’s setup is stronger than the headline suggests because it is using adjacent CPU and networking products to defend share while broadening wallet share per rack. The contrarian issue is that the market may be underestimating how quickly buyers will dual-source inference silicon to pressure pricing, especially if agentic AI shifts more compute toward general-purpose CPUs and specialized inference engines. That creates a narrower window for excess returns in NVDA: still the quality leader, but the edge may increasingly come from ecosystem monetization rather than pure growth acceleration.

Key Decisions for Investors

• Stay long NVDA vs. AMD as a pair trade for the next 3-6 months; the cleaner earnings durability and platform breadth should outperform if hyperscaler capex stays strong. Risk: if AMD starts converting wins into visible revenue faster than expected, the spread can compress quickly.
• Use AMD only on pullbacks, not strength: buy 3-6 month dips tied to post-rally digestion, with a focus on confirmation of revenue conversion from announced wins. Risk/reward is attractive only if forward multiple stops expanding before fundamentals catch up.
• Add a relative-value long NVDA / short INTC basket if you want direct exposure to agentic CPU attach; the market is likely to keep rewarding the integrated AI stack while Intel remains a lagging beneficiary. Time horizon: 6-12 months.
• For more tactical exposure, sell out-of-the-money AMD calls against a core long or buy call spreads into any earnings-driven volatility. That preserves upside if execution is real while monetizing the elevated valuation premium.
• Watch for evidence of pricing pressure in inference silicon over the next 1-2 quarters; if ASPs soften, rotate away from AMD-beta names into NVDA and networking suppliers that capture infrastructure spend even when compute economics tighten.