The rise of the AI infrastructure asset class

A new investment category is emerging at the intersection of artificial intelligence and infrastructure, attracting billions in capital from some of the world's largest investors. Major players like BlackRock, Microsoft, and KKR are racing to fund the physical backbone needed to power AI: from data centers to power plants to transmission lines.

The scale of investment needed is staggering. U.S. electricity demand is expected to include up to 128 gigawatts of new capacity in 2029, according to a recent estimate, while some predict that building needed data centers will require up to $900 billion over the next ten years. Others forecast that scaling AI and cloud infrastructure in the U.S. could cost $1 trillion by 2030.

The emerging opportunity is attracting the attention of major investors, who are evolving their investment strategies to make the most of it. And they’re coming from a broad spectrum of investment backgrounds, according to Pankaj Sachdeva, a senior partner at McKinsey who focuses on data centers and AI infrastructure. 

“Whether you are a real estate investor, an infrastructure investor, a credit fund, or a private equity investor, you're looking at data centers as an asset class and asking whether you should be investing in it,” he told Latitude Media. “Even some of the investors that were not that interested [two years ago] are now starting to look at it from an opportunity standpoint.” 

In September, BlackRock, Global Infrastructure Partners, Microsoft, and MGX got the ball rolling with the Global AI Infrastructure Investment Partnership, a $30 billion partnership to “invest in data centers and supporting power infrastructure.” One month later, KKR and Energy Capital Partners upped the stakes with a $50 billion strategic partnership “to fund data center, power, and grid infrastructure in the U.S. and globally.”

In a recent Brookfield Asset Management earnings call, president Connor Teskey said that the firm has been mulling a dedicated AI infrastructure fund, “leaning towards more of the infrastructure side of AI as opposed to the more private equity or growth side of it.” In the same call, he listed “AI infrastructure” as a key area of focus alongside the energy transition and private credit. 

The repeated use of the term “AI infrastructure” begs the question: are we witnessing the rise of a new AI infrastructure asset class? 

“We're seeing some early indicators of that,” Sachdeva said. “There are many parts of this AI value chain that look like infrastructure investment, with a specific yield and a high level of certainty around it.”

Is it infrastructure? 

Tania Tsoneva, head of infrastructure research at CBRE Investment Management, told Latitude Media that, as the AI boom takes off, it’s important to remember that the infrastructure asset class is generally a defensive, inflation-linked strategy within investors’ portfolios. For an investment to be considered infrastructure, it needs to be an essential service, with a predictable cash flow, high barriers to entry, and inflation linkage. 

"Generative AI provides a significant incremental boost to digital infrastructure and power generation in certain regions,” she said. “However, not all generative AI investments qualify as infrastructure, and not all clean energy investments will be specifically allocated for generative AI purposes."

For instance, it’s becoming more common for data centers to have long contracts with hyperscalers with good credit qualities, which makes them reliable, infrastructure-like investments. 

Meanwhile, chip manufacturers and generative AI startups belong to the same AI value chain, but, as less predictable investments, they likely have a different risk profile, and don’t fit within the emerging infrastructure asset class.

Mapping out the risk 

It’s still early days. One PwC analysis found that AI could contribute more than $15 trillion to the global economy in 2030, via a combination of increased productivity and production-side effects. But as Sachdeva pointed out, that growth will largely depend on investments made in the coming years in physical infrastructure like data centers, as well as servers and chips.

Today, individual investors are still making their risk assessments, and deciding how to approach the value chain. There are three kinds of risk investors are wrapping their heads around, Sachdeva said: 

1. The uncertainty of the predicted growth (whether or not the economic potential is there and the infrastructure will be needed); 
2. The profile and reliability of the end-customers (“This is increasingly becoming a customer-concentrated market with a few hyperscalers that are very committed to investing in this ecosystem… for the foreseeable future, he said.”); and 
3. The technology risk (“What if we have a breakthrough in terms of power generation that allows us to create all this power at a lower cost than it takes today?”).

 How individual investors assess these risks will determine how they decide to invest in the AI value chain, and whether they consider it an infrastructure investment. 

Ultimately, Sachdeva added, it comes down to how deeply they’ve reviewed the risks, what kinds of returns they’re looking for, and their appetite for risk — especially in a booming but uncertain market. “I think that in the end, the answer for each investor turns out to be different,” he said.