Investors pile into AI infrastructure bets beyond chip giants

The second-wave AI trade is moving beyond chips

The newest round of institutional buying shows a clear shift in the artificial intelligence trade: investors are not just chasing the most obvious chip winners, they are building exposure to the infrastructure that keeps AI running. In first-quarter filings, more than 4,000 institutional investors added to existing positions or started new ones in a basket of nine AI infrastructure-related companies, while only 146 institutions sold holdings in that group.

That imbalance matters because it suggests the money is still flowing into the AI buildout even as investors become more selective about the best-known megacap names. The pattern points to confidence in the ecosystem behind AI, not just in the headline companies that dominate market discussion.

Where the capital is going

The money is concentrating in the less glamorous but essential layers of the AI economy: compute, networking, power, storage and data-center capacity. Those categories benefit when companies, cloud providers and governments keep spending on training and deploying AI systems, because every new model needs chips, servers, racks, cooling, bandwidth and electricity.

The institutions buying those names were not just backing semiconductor demand. They were also putting money into data-center companies and utilities, which underlines a broader reality of the AI boom: it is not only a software story. It is an electricity story, a real-estate story and an industrial-capacity story.

Among the companies cited in the filings were Oracle, Arista Networks and Vertiv Holdings Co., along with other infrastructure-linked names such as Digital Realty. The mix shows how investors are spreading exposure across the physical stack that supports the AI economy rather than leaning on one narrow part of it.

Why infrastructure looks more attractive than crowded AI leaders

The attraction is partly valuation and partly diversification. The best-known AI stocks have become heavily owned, closely watched and in some cases expensive, which raises the bar for future gains. By contrast, infrastructure companies can benefit from the same secular theme while offering a different entry point into the trade.

That distinction is important for large investors managing risk. A chip leader may depend on one or two product cycles, while a network gear maker, power systems supplier or data-center operator can benefit from a longer wave of capital spending. If AI deployment broadens across industries, those companies can capture demand from many customers at once.

There is also a market-structure argument. When a theme becomes crowded, investors often search for the suppliers, landlords and service providers that earn revenue regardless of whether a specific model or application becomes a breakout hit. In the AI space, that means betting on the plumbing of the system: the equipment, the power and the buildings.

What the filings actually show

The data come from nearly 6,000 hedge funds, pension funds, college endowments and other asset managers that file Form 13F disclosures with the U.S. Securities and Exchange Commission. Those filings are designed to make large institutional holdings more publicly visible, but they are an imperfect window into live positioning.

Form 13F reports are due within 45 days after the end of each calendar quarter, and they disclose only long positions in Section 13(f) securities. They do not show short positions, and they do not capture trades made after the quarter ends. In other words, the market saw a March 31 snapshot only after the filings were submitted by mid-morning on May 15.

That lag matters when sentiment can shift quickly. A portfolio manager who added AI infrastructure in March may have trimmed it later in the spring, just as another manager may have increased exposure after the quarter closed. The filings still reveal a broad institutional preference, but they do not provide a real-time position report.

The company-level signals behind the theme

The company commentary helps explain why infrastructure names are drawing interest. Oracle said it is becoming a go-to place for AI workloads and reported remaining performance obligations above $455 billion in its latest quarter. That figure signals a deep pipeline of contracted business and suggests cloud demand tied to AI remains substantial.

Arista Networks described itself in first-quarter 2026 results as serving large AI and data-center environments. That positioning places the company at the center of networking demand, where faster data movement and lower latency are critical when AI systems scale.

Vertiv offered perhaps the clearest operating proof point. The company reported first-quarter 2026 net sales of $2.65 billion, up 30% from a year earlier, and said Americas organic sales rose 44% on strong data-center demand. For investors, that kind of growth supports the idea that AI spending is flowing into the unsexy but indispensable equipment that keeps data centers powered and cooled.

What happens if AI spending cools

The bullish case for infrastructure depends on continued capital expenditure. If AI spending slows, these stocks could feel the pressure quickly, especially the ones most directly tied to new data-center builds, power upgrades and networking rollouts. The same companies that benefit from a wave of construction can lose momentum if customers pause or delay projects.

Still, the impact would not be identical across the group. A company with broad enterprise demand, a large installed base or recurring service revenue may prove more resilient than a pure play on fresh buildouts. Oracle’s backlog, Arista’s scale in data-center networking and Vertiv’s exposure to critical power and cooling systems all suggest different degrees of protection if spending normalizes.

The bigger question is whether AI investment is still in the early buildout stage or moving into a more disciplined phase. If the industry keeps adding servers, cooling and utility capacity, infrastructure names may continue to outperform as the second-wave trade deepens. If capital spending fades, the market will quickly separate durable infrastructure franchises from the more cyclical beneficiaries of the boom.

For now, the filings show that institutional investors are still backing the machinery behind AI. That is a more cautious, and arguably more durable, way to own the trend.