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

The article argues AMD and Nvidia both have strong AI growth opportunities, with AMD highlighted for inference and agentic AI and Nvidia for its leadership in AI infrastructure. AMD is cited as trading at 35.5x 2027 estimates with roughly $35 billion in 2025 revenue, while Nvidia is described as trading at 16x fiscal 2028 estimates after 85% revenue growth to $81.6 billion last quarter and Q2 guidance for $91 billion. Overall tone is constructive on both stocks, but the piece favors Nvidia as the more attractive semiconductor stock at current valuations.

Analysis

The real market tell is not that both names benefit from AI, but that the profit pool is rotating from a single-vendor training stack toward a more fragmented compute architecture. That tends to compress pure GPU scarcity rents over time and lift the strategic value of complementary components: CPUs, networking, memory, and system integration. In that setup, the winner is the company with the broadest attach rate per rack, not necessarily the one with the most headline units shipped.

AMD’s upside is more about share capture than category growth. If inference and agentic workloads evolve as expected, its CPU position becomes more valuable because the workload mix shifts toward orchestration and memory-heavy serving, where platform breadth matters more than raw CUDA lock-in. The risk is that the market is already pricing a lot of that optionality; if customer ramp timing slips by even a couple quarters, the stock can de-rate sharply because the multiple is now anchored to distant earnings.

NVDA remains the cleaner expression of the AI capex cycle, but the better contrarian setup may be that its valuation is now low enough to absorb a deceleration in growth without a major multiple reset. The hidden risk for AMD is not AI demand itself, but that large buyers may dual-source to preserve bargaining power, turning today’s design wins into lower-margin volume. Intel’s main implication is negative: if agentic AI increases CPU content, the battlefield shifts further into a segment where it has already been losing share, reducing the odds of a near-term competitive reset.

The consensus is underestimating how quickly AI infrastructure buyers will optimize for total cost per token, not just peak FLOPS. That is structurally bullish for system-level vendors and memory-heavy architectures, and it argues for owning exposure to the full stack rather than treating GPU leadership as a winner-take-all outcome. Near term, the trade is less about catching a breakout and more about buying the pullbacks in names with the best earnings visibility while avoiding late-cycle valuation compression elsewhere.

Key Decisions for Investors

• Long NVDA on any 3-5% pullback over the next 2-4 weeks; use it as the cleaner AI capex expression with better risk/reward than chasing AMD after a strong run.
• Maintain a smaller strategic long AMD position for 6-12 months, but hedge with covered calls or a partial profit trim if the stock re-rates above 40x forward earnings, since dual-sourcing and margin pressure can cap upside.
• Pair trade: long NVDA / short INTC over 3-6 months. If agentic AI increases CPU share, Intel is the weakest incumbent levered to the wrong side of the mix shift.
• For higher-conviction exposure to AI infrastructure breadth, consider a basket long in NVDA + AMD with a slight overweight to NVDA, since the market is likely to reward systems integration and network attach more than standalone chip wins.
• Avoid chasing ARM here; the article’s setup implies optionality but not a near-term catalyst strong enough to justify paying for a cleaner beneficiary already available in NVDA.