For much of the last several years, the AI investment story has revolved around computing power. That makes sense. Advanced semiconductors sit at the center of the AI ecosystem, and demand for them has been extraordinary. But as the buildout accelerates, the story is becoming broader.
A data center is ultimately a physical asset. It needs electricity, cooling, networking, land, skilled labor, electrical equipment, financing, and access to a power grid capable of supporting enormous incremental loads. Those systems do not scale at software speed, and that mismatch may become one of the defining characteristics of the next phase of the AI investment cycle.
The early AI story was largely about the technology itself. Increasingly, the next chapter is about everything required to deploy that technology at scale.
Demand is moving faster than infrastructure
There is little evidence today that demand is the primary problem. Eaton reported that first quarter 2026 data center orders increased approximately 240% from a year earlier, while Electrical Americas orders were up 42% on a rolling twelve month basis. The contrast offers a useful glimpse into how unusually strong the data center infrastructure cycle has become.
But orders and energized capacity are two very different things.
Switchgear has to be manufactured. Transformers have to be delivered. Transmission capacity has to exist. Cooling systems must be installed. Utility connections need to be secured, and projects still have to move through permitting, construction, and labor constraints.
In other words, one of the most advanced areas of the digital economy is becoming increasingly dependent on a very traditional industrial supply chain.
That changes the investment conversation. The question is no longer simply who can design the fastest processor or produce the most advanced compute. It is also who can supply the infrastructure that allows that compute to be deployed.
AI is becoming an energy and infrastructure story
Power may prove to be one of the most consequential pieces of that equation.
The International Energy Agency expects U.S. electricity demand to rise meaningfully in the years ahead, with data center expansion accounting for a substantial portion of that growth. As the scale and density of AI computing increase, the demand reaches well beyond the technology sector itself.
Utilities, power generation, transmission, electrical equipment, natural gas infrastructure, storage, cooling, and grid modernization are all becoming part of the same ecosystem.
The AI investment landscape already extends well beyond the companies producing compute. Increasingly, it includes the businesses and industries supplying the infrastructure required to make that compute possible.
This is where we think the second order effects become particularly interesting. If power availability becomes a constraint, the response is not simply to wait for the grid to catch up. The companies driving this buildout have enormous financial resources and strong economic incentives to find alternatives.
We are already seeing greater interest in direct power procurement, onsite generation, advanced cooling systems, modular construction, storage, and other approaches designed to work around traditional infrastructure limitations. Longer term, the search for reliable, scalable power is also helping renew interest in technologies such as nuclear energy.
That is a topic worthy of its own discussion, but it illustrates a larger point: constraints do not necessarily stop innovation. In many cases, they redirect it.
The scale of the buildout is changing the financing story too
Infrastructure is only one constraint. Capital is another.
The five largest U.S. hyperscalers are spending at an extraordinary pace, and the scale of that investment is becoming large enough that financing the AI buildout is turning into a capital markets story in its own right.
J.P. Morgan estimates that the five largest U.S. hyperscalers will spend approximately $697 billion in capital expenditures during 2026. Epoch AI has separately modeled aggregate capital spending across Microsoft, Amazon, Alphabet, Meta, and Oracle growing materially faster than operating cash flow, with the two trends approaching a crossover during 2026.
That does not imply financial distress among these companies. These remain some of the most profitable and well capitalized businesses in the world.
It does, however, highlight the extraordinary amount of capital required to build the next generation of AI infrastructure.
As the investment requirements rise, corporate debt, project financing, joint ventures, private capital, and other financing structures are becoming more important. For investors, that broadens the implications of AI once again. The buildout may increasingly influence not only technology earnings, but also credit markets, interest rates, infrastructure spending, and the allocation of capital throughout the economy.
The bottlenecks are real, but so is the capacity to solve them
It would be easy to look at power shortages, equipment backlogs, labor constraints, permitting challenges, and rising capital requirements and conclude that the AI buildout is running into trouble.
We think the more interesting conclusion is more nuanced.
The bottlenecks are real, and they may influence how quickly planned capacity actually comes online. But the resiliency of the companies behind this investment should not be underestimated.
These are businesses with enormous balance sheets, considerable technical expertise, and a powerful incentive to solve the problems standing between them and future capacity. When one path becomes constrained, they have the ability to pursue another.
That makes the bottleneck itself potentially important from an investment perspective.
Scarcity can create pricing power. If demand for electrical equipment, power generation, cooling, networking, or specialized infrastructure continues to grow faster than supply, companies controlling critical capabilities may benefit disproportionately. At the same time, the effort to work around those shortages can create entirely new markets and investment themes.
None of that means every company associated with AI infrastructure will succeed. Valuation still matters. Execution matters. Competition matters. And large capital investment cycles rarely progress in a straight line.
But it does suggest that some of the most important investment implications of AI may increasingly be found outside the companies receiving most of the attention today.
What we are watching
We remain constructive on the long term AI buildout, but we believe the next phase may look different from the first.
The early story was largely about who could produce the most advanced compute. The next phase may be defined not only by who develops the most powerful technology, but by who can solve the physical and financial constraints required to deploy it at scale.
The bottlenecks are real. But so is the ability of the largest hyperscalers and infrastructure providers to innovate around them.
That is why we think the AI race is increasingly becoming a story of infrastructure, capital, and execution, not simply computing power.
And that may be where some of the most interesting developments come next.
Sources: Eaton Corporation, International Energy Agency, J.P. Morgan, Epoch AI.
Securities and Advisory Services offered through LPL Financial, a Registered Investment Adviser. Member FINRA/SIPC. www.finra.org, www.sipc.org.