The artificial intelligence revolution is fueling an unprecedented expansion of data center infrastructure. Major cloud service providers—Google, Meta, Microsoft, and Amazon—are locked in an aggressive race to meet surging demand for AI services, collectively committing hundreds of billions of dollars to stay competitive. That spending pipeline remains wide open despite stock market volatility and lingering concerns over whether AI-related equities are in a bubble.

"Hyperscalers are going to spend maybe between $750 and $800 billion [a year]. Some forecasts even have it up to a trillion, and that would put it at 2.5 to 3% of U.S. GDP, which is just extraordinary for a capital market," said John Mowrey, chief investment officer at NFJ Investment Group.

A significant share of this capital is flowing into data centers, which present a pick-and-shovels approach for investors seeking exposure to the AI wave. Fortune identified four distinct entry points into the data center economy: semiconductor chips, real estate, energy, and cooling. Industry experts highlighted standout stocks within each category, reflecting how the buildout is spreading beyond headline AI names into the industrial and physical systems needed to keep the technology running.

Chips: The Brains Behind the Buildout

Every AI data center depends on multiple interconnected components, but none function without semiconductors—the processing brains that make these facilities work.

Even as companies race to construct the physical buildings to house these operations, the specialized processors that run AI workloads cannot be manufactured quickly enough. Demand has outstripped the industry's production capacity, making semiconductors one of the most acute bottlenecks in the entire buildout, according to Craig Ellis, research director and senior semiconductor analyst at B. Riley Securities.

For investors, that shortage is not necessarily a drawback.

"We're at a point where undersupply is so severe that there needs to be a multi-year period of unusually strong capex growth in front of us, and that capex growth is something that is very investable because it has the potential to continue to lift expectations for revenues and earnings," Ellis said.

To capitalize on this opportunity, Ellis recommends that investors look beyond chip giants like Nvidia, AMD, and TSMC and instead focus on the companies that manufacture the equipment used to produce semiconductors.

Applied Materials (AMAT), the world's largest semiconductor equipment company, is a prime example. Its core Semiconductor Systems division accounts for approximately 73% of total revenue, according to its most recent annual filing, supplying the firms that manufacture chips for computing, logic, and memory applications. Its advantage lies in breadth: nearly every advanced chip and display passes through Applied's tools at some stage, distributing its exposure across the entire chipmaking ecosystem rather than concentrating on a single chip type.

Ellis also highlighted Lam Research (LRCX) as a compelling alternative. The company produces equipment for memory and storage chips, and while its product range is narrower than Applied Materials', Ellis sees Lam as especially well positioned to benefit from a surge in new capacity investment over the next two years. B. Riley Securities has raised its earnings estimates for Lam Research by 25% to reflect that outlook.

Marvell Technology (MRVL) is another strong contender, particularly in networking—the infrastructure that enables thousands of chips within a data center to communicate fast enough to function as a single integrated system. Though less recognized by the general public, Marvell commands approximately $195 billion in market value, according to analytics firm FactSet, placing it in the same league as Nvidia and AMD. It has even attracted a direct investment from Nvidia as part of a partnership on next-generation networking technology.

Currently, the majority of Marvell's revenue comes from Amazon Web Services, Amazon's cloud computing division. While that relationship has been highly lucrative, it also ties Marvell's performance heavily to a single client. Ellis views this concentration as an opportunity rather than a warning sign: if Marvell can expand its customer base and scale its networking products, he believes the company could generate significantly more profit than it does today.

These investments carry risks, however. Chip stocks are known for sharp price swings and can be disrupted by geopolitical tensions or changes in how hyperscalers choose to finance their spending.

Still, Ellis maintains that the industry is transitioning from a boom-and-bust cycle toward more durable, long-term growth. Even after the SOX—the primary index tracking AI-linked chip stocks—fell 26% on concerns that the spending frenzy may not pay off, Ellis argues the decline already accounted for most of that risk, making the current moment a reasonable entry point. The data supports his view: since the July 28 sell-off, Applied Materials, Lam Research, and Marvell have each recovered between roughly 22% and 30%.

AI's Physical Backbone: Real Estate

Chips require physical space to operate, and that is where real estate investment trusts, or REITs, come in. These trusts lease climate-controlled facilities equipped with the power generators and high-speed connectivity necessary for continuous, round-the-clock operations.

Patrick Wilson, a portfolio manager on the real estate securities team at CenterSquare Investment Management, described REITs as a sound investment, pointing to the industry's transition from the AI training phase to the inference phase—the stage at which models shift from being trained to being actively deployed for everyday queries. While initial AI model training occurred in large, rural data centers selected for cheap land and affordable power, day-to-day model operation works best in facilities near major population centers, where shorter geographic distances reduce latency—the time delay between a request and a response. Because established REITs already control the limited, "carrier-dense" real estate in these urban centers, they are uniquely positioned to capture this next wave of demand.

REITs also offer tax efficiency for ordinary investors. "REITs are only taxed once, provided that they satisfy IRS tax rules. So, if they pay out 90% of their taxable net income into a dividend, there's no corporate level tax that they pay," Wilson explained.

To capture this opportunity, Wilson pointed to two REIT leaders: Equinix (EQIX) and Digital Realty (DLR). Equinix operates as a major landlord for internet and cloud computing infrastructure, running data centers that host servers for thousands of companies ranging from Fortune 500 firms to major cloud providers such as AWS and Google Cloud. Digital Realty has spent more than two decades acquiring and constructing similar facilities, capitalizing on sustained demand for secure, climate-controlled server space that continues to outstrip supply. Wilson noted that this supply-demand imbalance has enabled the REIT to raise rents, and its thousands of tenants make it more stable than newer competitors that rely on only a handful of large contracts.

Wilson cautioned, however, that investment risks remain. Data center REITs are vulnerable to rising interest rates, which increase borrowing costs and pressure property valuations. They also rely on maintaining premium rents, meaning a flood of new competitors could undermine their pricing power. Geographic constraints add another layer of complexity: while data centers near metropolitan areas are more valuable, suitable land in those locations is scarce. "They need a lot of land, and that's not necessarily found in midtown Manhattan," he said.

Nevertheless, Wilson considers REITs a safer investment compared to newer, heavily indebted AI infrastructure firms. If AI demand slows or short-term customers depart, those companies remain saddled with long-term obligations and substantial interest payments, making them considerably more fragile than established REITs.

Utility Upside: Powering the Boom

The enormous electricity requirements of AI data centers have fundamentally reshaped the energy sector. Andrew Bischof, utilities analyst at financial services firm Morningstar, observed that for decades, annual U.S. electricity demand was essentially flat, growing at just 0% to 0.5%. AI's power-hungry nature has dramatically altered that outlook, transforming utilities from historically low-growth yield stocks into vehicles with meaningful upside potential.

"You're now seeing more growth-oriented investors coming to utilities because they can provide that six to eight and sometimes 10% annualized growth over the five-year forecast," Bischof said.

One company well positioned to benefit is American Electric Power (AEP), a major public utility holding company headquartered in Ohio that distributes electricity to more than five million customers across 11 states, according to its website. Anticipating a substantial surge in data center-driven electricity demand, AEP plans to invest $78 billion between 2026 and 2030 to expand the necessary infrastructure. Morningstar is optimistic about this buildout, citing AEP's strong track record of completing committed projects, and expects the investment to translate into 9% average annual earnings growth through 2030—a robust figure for a utility.

Not all utility stocks, however, offer the same profile. Bischof noted that in certain regions—particularly the Mid-Atlantic—the transition toward supporting data centers has driven up power prices and raised concerns that these increased costs are being passed directly to retail customers.

As a result, Bischof recommended additional utilities beyond AEP, including DTE Energy (DTE), Alliant Energy (LNT), and Evergy (EVRG)—companies located in regions where regulators and communities are broadly aligned on the data center boom.

For a more contrarian approach to the power dimension of the AI buildout, David Trainer, CEO of investment research firm New Constructs, suggests looking beyond utilities entirely. He argues that while alternative and green energy sources are expanding, they still cannot deliver the high-intensity power required by massive AI data centers, making fossil fuels essential to meeting this demand. Trainer therefore views "cheap" traditional energy stocks as a rare opportunity to capture the infrastructure boom at value prices, pointing to refiners such as Valero (VLO) and HF Sinclair (DINO), which he believes the market is pricing as though their profits are headed for a permanent 30% to 40% decline.

Cooling the Boom: Thermal Management

Another investment angle lies in the hardware that prevents AI data centers from overheating. The massive electricity consumption of these facilities generates intense heat that conventional air cooling systems cannot adequately manage.

Nick Lieb, industrials analyst at Morningstar, argues that this makes cooling and power equipment a fundamentally different type of AI trade. While companies like Nvidia release new flagship GPUs every year or two, the core technology underlying data center infrastructure evolves far more gradually, offering investors a more stable alternative.

Vertiv (VRT), a long-standing provider of precision cooling and power systems, is Lieb's primary example of a company built to capitalize on this trend.

"Vertiv owns the original data center cooling brand called Liebert, which actually invented the original computer room air handling unit… in the 1960s. That base technology is still quite relevant today [and] sells a lot of computer room air handling units," Lieb explained.

Lieb cautioned, however, that this stability is counterbalanced by significant concentration risk: more than 80% of Vertiv's revenue is generated directly from the data center market. Any slowdown or disruption in spending by the major hyperscalers could leave Vertiv investors exposed.

For investors seeking more limited exposure to the infrastructure boom, Lieb pointed to power management company Eaton (ETN). Only about a quarter of Eaton's sales are directly tied to data centers; the remainder comes from commercial structures and other segments of the broader electrical grid. This diversification provides Eaton with a stronger hedge in the event that data center growth decelerates. Lieb added that Eaton is uniquely positioned to benefit from a U.S. electrical grid he describes as "old and decrepit" and long overdue for modernization. The company produces essential hardware—specifically transformers and switchgear—required to support both new data center demand and this critical national infrastructure upgrade.

Fragile Foundations: Risks on the Horizon

Despite the success of the AI narrative, some observers caution that the current surge in capital spending may not ultimately deliver the returns many anticipate.

Lieb views the sheer scale of current spending as a potential source of long-term risk, noting that hyperscalers have reached a point where the cost of building new data centers is beginning to "exceed the cash that they generate," forcing tech giants to increasingly turn to debt markets and equity issuance to raise funds.

Trainer described the current market as a "crowded trade," making it nearly impossible to identify an obvious stock that is not already overvalued. He warns that many of the large firms like Amazon and Microsoft currently "bragging" about their massive capital expenditures "can't afford to stay in the race" at their current spending rates—a competitive contest he contends many participants lack the balance sheets to sustain.

Yet the scale and potential impact of the buildout are difficult to dismiss. Mowrey emphasized that the data center boom is significant not only in dollar terms but also in its broader economic implications. In his view, it is laying the foundation for a technological shift that could reach nearly every sector of the economy.

"AI adoption is still in its early innings… I think we're just scratching the surface," he said. "The broad enterprise application across healthcare, manufacturing, and financials—that's going to take years to filter in."