3 ways Fidelity managers invest in AI stocks

Since ChatGPT's arrival helped spark the AI boom, investors have cycled through a series of potential stock market winners, from chipmakers and networking companies to suppliers of materials and power for data centers. Yet as AI models become more sophisticated and adoption expands, one early supply bottleneck remains among the most critical today: memory.

How to invest in AI

The AI buildout is still in its early stages. As hyperscalers continue to invest heavily in developing more sophisticated models, and businesses increase adoption of AI capabilities, demand for the building blocks for these platforms remains robust.

“These models are getting more capable, demand for the intelligence on offer is staggering, and the infrastructure required to deliver that intelligence hasn’t been able to keep up with the demand,” says Zach Turner, manager of the Fidelity^® Capital Appreciation Fund (). “I believe there's still opportunity to be found in owning the picks-and-shovels enablers of this AI infrastructure buildout.”

How investing in AI has changed

The enormous promise of AI has fueled historic levels of capital spending, helping drive strong corporate earnings growth and investor enthusiasm for the sector. Hyperscalers are projected to spend more than $1 trillion on AI development over the next 2 years.¹

As spending has accelerated, a succession of supply bottlenecks has emerged across the technology supply chain. Companies positioned at those chokepoints have often benefited from exceptional demand, pricing power, and profitability, which have helped fuel stock-price gains. For investors, it has been important to identify bottlenecks, assess how long they may last, and look ahead to potential future chokepoints.

Investors initially focused on processing chips and networking equipment before shifting their attention to semiconductor manufacturing equipment, materials, and the myriad companies that support data centers.

While leadership across the AI ecosystem has shifted in just the past few years, some Fidelity pros have been focused on identifying companies with more durable growth potential and competitive advantages they expect can persist.

“People have been looking for the next bottleneck and the next stocks to run. In reality, the biggest bottlenecks are almost the same now as they were in the beginning,” says Matthias Plancke, an equity analyst at Fidelity specializing in technology.

How AI infrastructure development has sustained demand for memory

The AI investing landscape has evolved rapidly since ChatGPT was introduced in late 2022. What began as a race to build, scale, and train AI models has shifted to inference, the process of putting those models to work using real-life data, and to agentic AI systems capable of handling increasingly complex tasks.

As users engage in longer conversations, upload more data, and ask AI to perform increasingly complex tasks, memory requirements have surged. Demand for memory hardware is doubling annually, while supply is increasing by roughly 20% to 30% annually, a mismatch that could persist for years.²

Some Fidelity managers believe this shift could make memory one of the most enduring constraints in today's AI infrastructure buildout.

“In the agentic AI era, the amount of data you have to cram into this finite memory resource goes up and up,” Turner says. “There are all kinds of networking bottlenecks, but the biggest one right now is memory.”

Turner estimates that roughly 200 zettabytes—hundreds of trillions of gigabytes—of data are generated each year. The vast amounts of data required to train, refine, and run AI models are expected to sustain high demand for memory and storage technologies, such as hard-disk drives, NAND flash memory, and dynamic random-access memory (DRAM).

“What makes investing in this space challenging is that AI isn’t static. Power needs are increasing. The throughput needs are increasing. This is changing how data centers are being designed and built,” says Vincent Montemaggiore, manager of the Fidelity^® Overseas Fund (). “Technology is adapting as AI developers transition from training models to more real-world usage.”

How some managers assess potential AI stock opportunities

Turner and Montemaggiore say they generally group AI-related businesses into 3 buckets based on the durability of their competitive advantages, which in many cases rests on the likely duration of supply constraints.

1. Long-term potential AI winners with durable advantages

The first category includes companies occupying critical positions in the AI ecosystem and expected to benefit from the AI buildout for years, not months. Turner and Montemaggiore point to areas such as semiconductor manufacturing, chip-making equipment, and chip-testing technologies as examples of businesses with potentially durable advantages.

“These businesses are critical within the ecosystem, are high-quality, have limited competition, and the sustainability of the cash flows is much clearer,” says Montemaggiore.

2. Medium-term potential winners in consolidated AI businesses with advantages that may fade

The second category includes businesses that can benefit from strong demand and limited competition today, but where supply constraints may eventually ease. Investors in these companies are often making judgments about how long those favorable conditions can last.

Examples here may be found in consolidated industries with more commoditized products. While supply-demand imbalances might be driving higher earnings, it’s more difficult to project how long these conditions can last.

“These stocks are relatively cheap because their projected earnings growth is so high, but nobody really knows the duration of that advantage. You're making a bet here,” Montemaggiore says.

3. Short-term potential AI winners with temporary advantages

The third category consists of more commodity-like businesses, where products are less differentiated and new supply can enter the market more easily. Those investments may carry greater risk because their competitive advantages may not be sustainable.

This category includes niche equipment and materials suppliers in the chip-making process, as well as several companies in the networking and optics value chains.

“From a portfolio manager perspective, as much as these stocks go up, they could also go down because the pricing power they can benefit from today could ultimately prove fleeting. The duration is not clear, and the market will be very forward-looking, which means the stocks could start to underperform well before their earnings peak,” says Montemaggiore.

What are some of the risks in AI investing?

Despite the enthusiasm surrounding AI, investors continue to debate whether today’s massive spending will generate sufficient returns, asking themselves, “Should I invest in AI?”

Some AI-related stocks have pulled back recently amid concerns about elevated capital spending levels, which has led to increased debt levels and raised questions about when AI companies might realize a return on their investments. If results fall short of expectations, development spending could slow, weighing on the torrid growth of the AI ecosystem.

“We've been on an incredible run in these stocks. The market naturally goes through phases where it becomes overly euphoric and may extrapolate the earnings trajectories of these companies,” Turner says. “It’s also a natural thing for the market to take a breather and retrace, from a price appreciation standpoint.”

Investors can learn more about the mutual funds mentioned in this article, including fund objectives and most recent complete holdings, by visiting the fund summary pages on Fidelity.com:

• Fidelity^® Overseas Fund ()
• Fidelity^® Capital Appreciation Fund ()