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

Ark Investment's Cathie Wood has made a name for herself as a champion of stocks with potentially disruptive technologies. Her flagship Ark Innovation ETF (NYSEMKT: ARKK) has had mixed results, up 49% over the past year but down 8.7% on average over the last five, as of the end of April.

Nonetheless, when she makes moves, it tends to be notable. Recently, she has been trimming her stake in Advanced Micro Devices (NASDAQ: AMD) while buying shares of rival Nvidia (NASDAQ: NVDA). Let's see if investors should be following suit.

Image source: The Motley Fool.

AMD: An inference and agentic AI opportunity

While Wood has trimmed her stake in AMD, it remains her fifth-largest position. The stock has quadrupled over the past year, driving its valuation to 35.5 times 2027 analyst estimates. That likely played a role in her recent decision to take some profits.

However, AMD is at the intersection of two powerful artificial intelligence trends that should help power its results going forward. The first is inference, with its graphics processing units (GPUs) well suited to take share. Inference is less technically demanding than training, which Nvidia's CUDA software does best, while its chiplet design packs in more memory, which is important for inference.

The company has locked in two large GPU partnerships and commitments, each worth an estimated $100 billion. In addition, it's believed that Anthropic will use its next-generation chips. After being an afterthought to Nvidia in AI model training, AMD now looks like a strong alternative for inference.

On top of that, the company also has a big opportunity stemming from agentic AI. While GPUs are great at providing the muscle to train AI models and run inference, they aren't good at sequential reasoning or working with outside tools, both of which are needed for AI agents. That job falls to central processing units (CPUs), which work as the brain of a computer. As a result, the ratio of GPUs to CPUs is expected to fall from 8:1 with training to 1:1 with agentic AI.

This is a huge opportunity for AMD, as it is a leader in the data center CPU market, having consistently taken share away from rival Intel. AMD has said this will be a $120 billion market over the next few years, while Nvidia has said it could become a $200 billion market.

With AMD generating just under $35 billion in total revenue in 2025, it is set for explosive growth in the coming years. As such, I think the stock still looks like a good one to own here.

Nvidia: The AI infrastructure leader

With Nvidia trading at a forward P/E of only 16 times fiscal 2028 (ending January 2028) analyst estimates, it's easy to see why Wood would be attracted to the stock. Not only is the stock cheap, but the company is also growing rapidly. Last quarter, its revenue surged 85% to $81.6 billion, while it guided for its Q2 revenue to soar 95% to $91 billion.

Nvidia remains the undisputed leader in AI model training, but it's also positioning itself for inference and agentic AI. The company has developed its own ARM-based CPUs, while its "acquisition" of Groq brought with it language processing units (LPUs) designed specifically for inference. Smartly, it can combine its GPU with LPUs to separately handle the pre-fill and decode phases of inference, offering a unique solution in this space. Together with its networking portfolio, Nvidia can now offer end-to-end servers dedicated to specific AI tasks, including training, inference, and agentic AI.

Which is the better stock to buy?

Given its valuation and continued growth, I think Nvidia looks like the more attractive semiconductor stock at the moment. However, I think investors can own both, as AMD has tremendous growth ahead.

Geoffrey Seiler has positions in Advanced Micro Devices. The Motley Fool has positions in and recommends Advanced Micro Devices, Intel, and Nvidia. The Motley Fool has a disclosure policy.