There is currently more than 3,800 MW of capacity being built in the U.S., up about 70% from the prior year and another 7,000 MW in various stages of pre-construction.

In the early internet days, the late U.S. Senator Ted Stevens famously used the analogy of a “series of tubes” to explain how the internet worked. Today, investors are trying to understand the “series of tubes” that enable artificial intelligence. Behind every interaction with an AI tool is not only a complex web of digital neural networks, but also a humming physical network of data centers, electricity lines and power plants.

From an investor perspective, after the spectacular run-up in the “Magnificent 7” stocks over the last two years, investors are now asking, are there opportunities to diversify along the value chain? And specifically, in alternative investments?

We think so. Let’s take a closer look at how investors in private energy infrastructure are investing in this trend without the high volatility of the public markets.

A boom in data center building

Even preceding the recent surge in consumer AI, a data center building boom has been underway. The amount of capacity in the U.S. is on track to double over the next decade. There is currently more than 3,800 MW of capacity being built in the U.S., up about 70% from the prior year and another 7,000 MW in various stages of pre-construction1. (Data centers are commonly measured in terms of electrical capacity, as well as the more traditional square footage, underscoring the criticality of energy supply to their operations.)

But electrical grid constraints at a local level are becoming bottlenecks to new construction. AI data centers are massive consumers of electrical power. A ChatGPT query, for example, consumes an estimated 2.9 watt-hours, 10 times more than a Google search2. And as Google and other search engines begin to incorporate AI responses by default, their energy needs will grow. This is particularly a challenge for the hyperscalers looking to scale AI adoption in the coming years. 40% of existing AI data centers are expected to be operationally constrained by power availability by 2027, according to a Gartner study3.

Privately-held power generators are often able to ramp up capacity more quickly than their public counterparts, using funding from investors rather than taxpayer dollars. This is allowing them to take advantage of emerging opportunities to power the latest, most energy-hungry facilities.

These large scale shifts in the energy and data landscape are creating compelling opportunities for investment and value creation in the “tubes” of the internet, particularly in alternative markets.

AI is evolving rapidly and creating winners and losers in the public equity markets. But increased electricity demand is a theme that will remain relevant into the coming years. Private infrastructure investment offers opportunities to gain long-term exposure to these favorable trends with lower volatility than public markets or venture capital.

GTA Slide 47

1 North America Data Center Trends H1 2024, CBRE, August 19, 2024.
2 Powering Intelligence, 2024 White Paper, May 2024, EPRI
3 Gartner Predicts Power Shortages Will Restrict 40% of AI Data Centers By 2027, Gartner, November 12, 2024.
09px241112133702