Since March, one theme has largely dictated the direction of global markets: artificial intelligence infrastructure. Despite geopolitical tensions, interest rate volatility, and weakness across several segments of the real economy, AI-related companies have accounted for a disproportionate share of equity market performance. The gains have been remarkable, with several key companies rising more than 100% in just a few months, while many traditional defensive sectors have remained largely flat.

This helps explain the strong performance of our Global Long-Term Thematic Equity Model in recent months, as the portfolio holds several companies with direct exposure to the artificial intelligence ecosystem. Given the questions this has generated from clients, we felt it would be useful to explain our thinking in greater detail.

A Major Theme, But Still in Its Early Stages: Selling the Tools

We believe artificial intelligence represents one of the most important investment themes of the coming decades. Its potential impact on productivity and everyday life is likely to be significant. While the numbers involved are already enormous, the real-world adoption of AI across the broader economy remains limited relative to what it could ultimately become.

Today, major industry participants such as Microsoft, Amazon, Alphabet, Meta, OpenAI, Anthropic and many others are engaged in what has become an existential race. Capital spending is no longer discretionary; it has become necessary to protect competitive positions. In many cases, these investments are now growing faster than already substantial cash flows, forcing companies to issue additional debt or equity to finance the next generation of data centers. As a result, we remain cautious toward businesses that bear most of the financial risk associated with this race.

For much of the past two years, the market rewarded companies that announced the largest AI investments. The more a company planned to spend, the more its stock tended to rise. That dynamic has begun to change in recent months.

History shows that during major industrial revolutions, it is often the suppliers of the essential tools who create the most durable value. We do not know which language model or software application will dominate five years from now. We do, however, have a better understanding of what all of these participants will be required to purchase in order to operate their infrastructure. For that reason, we prefer investing in providers of critical physical components and resources.

From Powerful GPUs to Custom Chips and Connectivity

For the past two years, the market’s attention has been focused almost entirely on GPUs, highly powerful processors capable of performing many calculations simultaneously. They were essential to the first phase of artificial intelligence: training large language models (LLMs). Training these models requires enormous computing power, and Nvidia was uniquely positioned to meet that demand.

We are now entering the second phase: the real-time use of AI by millions of users simultaneously. In this environment, the raw power of an individual chip becomes less important than cost per query, energy efficiency, and the ability of the overall system to operate at scale. To reduce costs and decrease reliance on a single supplier, major data center operators are increasingly turning toward custom-designed chips and much more interconnected architectures.

The key challenge facing modern data centers is the physical bottleneck created when thousands of chips must communicate with one another. When communication is slow, valuable computing power is wasted. It is precisely this segment, the connectivity and “pipes” of artificial intelligence, where our highest conviction lies.

 Marvell: Our Largest Position

Marvell Technology (MRVL) is currently our largest position. We first began purchasing the stock in 2020 at approximately $33 per share, initially for its exposure to 5G during the pandemic. After taking profits as that cycle matured, we gradually rebuilt the position below $80 as it became increasingly clear that the company was emerging as an important player in artificial intelligence.

We believed that the next phase of growth would be driven less by general-purpose processors and more by custom chips and the infrastructure required to operate data centers efficiently. For a long time, the market remained skeptical of this thesis and of Marvell’s positioning. Over the past several months, that view has changed dramatically. The stock has nearly quadrupled since February.

Marvell occupies a unique position within the AI ecosystem. On one hand, it helps develop custom chips for major cloud providers including Amazon, Microsoft, Google and Meta. On the other hand, it plays a critical role in the networking infrastructure that connects those processors together. In many respects, Marvell serves as the nervous system of modern data centers.

Traditionally, servers communicate through copper cables carrying electrical signals. However, this approach becomes increasingly inefficient as data volumes grow. Copper generates more heat, consumes more energy and slows data transmission over longer distances. Marvell is now one of the global leaders in optical networking solutions that replace portions of these electrical connections with significantly faster and more efficient light-based communication.

This expertise has become strategically important across the industry. Nvidia recently announced a strategic partnership with Marvell aimed at improving connectivity between its AI processors. The CEO of Nvidia also stated that he believes Marvell could become the next company to reach a US$1 trillion market capitalization, at a time when Marvell’s market capitalization was still approximately US$150 billion. This recognition was further reinforced by Marvell’s inclusion in the S&P 500 on June 22.

 Our Other AI-Related Holdings

While Marvell remains our primary AI investment, it is far from our only exposure to the theme.

ASML, based in the Netherlands, holds a near-monopoly on the advanced lithography machines required to manufacture the world’s most sophisticated semiconductors. Regardless of which company ultimately wins the AI race, the most advanced chips will need to be produced using ASML’s technology.

Qualcomm provides exposure to the arrival of artificial intelligence directly within devices. As AI increasingly moves into smartphones, computers, automobiles, smart glasses and connected devices, demand for local processing power is likely to increase substantially.

Palo Alto Networks is one of the world’s leading cybersecurity companies. As systems become more intelligent, automated and connected, protecting data, networks and infrastructure become increasingly critical.

We maintain high long-term conviction in all four companies. Despite their strong recent performance, we have chosen to maintain our positions. This increases the volatility of our Global Long-Term Thematic Equity Model, but we believe their long-term value creation potential remains substantial.

 AI Benefits More Than Just Technology Companies

Artificial intelligence does not benefit only technology companies. Data centers must be built, powered, cooled and connected. This requires substantial amounts of materials, energy and infrastructure. Interestingly, many of these indirect beneficiaries have yet to fully participate in the recent market rally, particularly in energy, natural resources and engineering. We are currently overweight these sectors.

We also see growing signs of an industrial boom emerging across the American Midwest, where access to electricity, infrastructure and skilled industrial labour creates an important competitive advantage for future AI-related investment.

The Canadian stock market also appears particularly well positioned due to its significant exposure to natural resources, energy and infrastructure. It is also important to remember that other long-term themes, including autonomous vehicles, automation and robotics, are closely tied to artificial intelligence and often rely on many of the same components, infrastructure and resources.

Volatility and Risk Management

We fully expect AI-related investments to remain volatile. Every major technology company has experienced significant corrections throughout its history. Volatility is simply part of the normal path of high-growth businesses.

That said, our role is not merely to buy a promising theme. Our role is to manage risk. We continuously monitor valuations, cash flows, debt levels, revenue quality and demand trends. We take profits when we believe risk has become excessive and ensure that no single position becomes overly dominant within portfolios. As an example, no individual holding can exceed 10% of our total equity exposure. In addition, we remain underweight technology relative to our benchmark index.

We also maintain a strong commitment to diversification. Even within our most growth-oriented models, we do not seek to own only the most popular stocks. Our low-volatility models continue to represent an important portion of our portfolios. They provide exposure to more stable, profitable businesses that are less dependent on enthusiasm surrounding artificial intelligence.

This approach allows us to participate in the growth of AI while maintaining a prudent approach to risk and capital preservation.

Conclusion

We know that artificial intelligence will eventually experience periods of excess. There will be corrections. Some companies will become too expensive, and some projects will fail to generate the returns investors expect. Retail investors are already beginning to enter certain names at valuations that are difficult to justify. This is precisely why discipline remains essential.

For now, however, demand remains exceptionally strong. The current cycle is characterized more by a shortage of supply than by any lack of demand. In many areas, the primary constraint is not customers, but rather electricity, components, manufacturing capacity and infrastructure. If these constraints did not exist, investment levels would likely be even higher.

We therefore believe that artificial intelligence is still far from reaching its full potential, but we intend to remain highly disciplined in how we invest in this theme.

We remain invested with conviction while maintaining the discipline required to reduce or exit a position when risk no longer justifies the potential return.

Our objective remains unchanged: to participate in the most important long-term growth drivers while maintaining rigorous risk management. We believe this approach will allow us to create durable value for our investors.

RBC Dominion Securities Inc.* and Royal Bank of Canada are separate corporate entities, which are affiliated. *Member–Canadian Investor Protection Fund. RBC Dominion Securities Inc. is a member company of RBC Wealth Management, a business segment of Royal Bank of Canada. ® / ™ This information is not investment advice. The information contained herein has been obtained from sources believed to be reliable at the time obtained but neither RBC Dominion Securities Inc. nor its employees, agents, or information suppliers can guarantee its accuracy or completeness. This report is not an offer to sell or a solicitation of an offer to buy any securities. This report is furnished on the basis and understanding that neither RBC Dominion Securities Inc. nor its employees, agents, or information suppliers are to be under any responsibility or liability whatsoever in respect thereof.