AI experts weigh in on whether the fad makes sense or just an overbuilt trend for institutional investors
AI data centers are seemingly becoming the hottest frontier in infrastructure investment as Canada’s so-called Maple 8 pension giants are betting big on them.
According to a report from The Logic, there’s an estimated US$6.6 billion invested in public data center and digital infrastructure firms among Canada’s big pension funds, up from just $2.3 billion before the rise of ChatGPT.
If the trend is real, is it sustainable for the long haul?
“Data is the new commodity,” said Carl Elia, vice president and director of global infrastructure investments at TD Asset Management. “It’s really being recognized as an essential infrastructure asset class and something that’s needed. There’s tremendous growth in it as well.”
While the first wave of investment focused on massive hyperscale facilities, Elia noted a shift toward more targeted, localized development. He believes the rapid expansion of AI has supercharged demand for data processing, as everything from cars to thermostats generates vast volumes of data.
Elia believes AI is more than just a tech fad but a structural shift reshaping the global economy as they’re still seeing infill across markets. Elia pointed to smaller, co-located data centers and tower assets being built closer to population centers to handle surging data demands from everything from smartphones to thermostats.
He said the appeal of AI data centers ties back to the fundamental strengths of infrastructure investing - steady income, inflation protection, and portfolio diversification.
“What you’re getting from these assets is long-term, steady cash flow,” he said.
However, prior to the AI boom, many infrastructure managers were starting to question the viability of further investment in data centers.
“It was more common for us to hear from managers that their data centers were looking a bit overbuilt and the returns are a little bit low,” said Bradley Hough, recalling that changed abruptly with the rise of generative AI, particularly the launch of ChatGPT, which sparked a dramatic surge in demand.
“It was really the emergence of AI that changed that picture in a positive way,” said Hough, consultant at PBI Actuarial Consultants.
But he warned the sector’s growth isn’t without complications, particularly when it comes to power. For two decades, electricity demand in the US had been trending downward, thanks to energy efficiency gains like LED lighting and flat-screen TVs. But that has since gone away with the emergence of data centers and AI.
Additionally, the surge in electricity usage, Hough added, is reshaping not just digital infrastructure but the entire utility and power generation sector, as the surge in electricity demand driven by AI data centers is forcing infrastructure managers to rethink not just how much power is needed, but how it's produced.
“The limiting factor on some of these data centers could be the supply of power, rather than the ability to build data centers,” he said. “They’re carefully not just looking at how much electricity but how they need to generate it, whether it’s renewables or gas-powered turbines,” he said.
For now, he underscored many managers are diversifying their exposure cautiously, often demanding higher returns to compensate for emerging technology risk.
Meanwhile, Elia emphasized that as digital infrastructure grows, it’s offering the same structural benefits traditionally found in sectors like transportation or utilities. He emphasized that many of these assets come with contracted or inflation-linked revenues, which help insulate portfolios from market volatility.
“Usually, you’ve got a long-term contract with a high-quality counterparty,” he said. “You’re very sure of the cash flows that you get.”
Hough agrees the sector has clear appeal for institutional portfolios, asserting that it’s pretty common to see some form of digital infrastructure in most infrastructure portfolios.
Hough noted that while AI data centers have grabbed the spotlight, other forms of digital infrastructure like fiber optics and communication towers are also gaining traction in institutional portfolios.
“Fiber’s been present in portfolios,” he said, adding that “communication towers are one of the traditional ones people forget about.”
He emphasized that each sub-sector carries its own set of risks, especially given how new and fast-moving the space is. Pension funds have also been cautious, often demanding higher returns to offset the uncertainties.
“These sectors are new and still emerging,” Hough said. “As long as you’re aware and you price those appropriately, you can build those into your allocation.”
Elia acknowledged, however, that some fiber networks have drawn criticism due to weaker business models, particularly those with a retail focus and less reliable revenue streams.
“Some fiber networks have gotten some bad publicity being more retail focused,” he said, noting that this can make them less secure as infrastructure assets.
Still, he sees plenty of upside in the space, adding that the US is seeing a rise in smaller-scale infrastructure like localized towers, especially in high-growth areas like the Sun Belt.
The broader takeaway, he emphasized, is that “the vast use of data means that the infrastructure needs to be there for all population centers,” driving continued expansion across digital asset classes.
Meanwhile, Hussein Allidina, managing director and head of commodities at TDAM, pointed to a related, and potentially more limiting, constraint on commodities, noting there’s a “magnitude and material number of commodities that are required to build the data centers. You need a tremendous amount of copper, aluminum and fuel sources, like nuclear, natural gas, coal, wind,” he said.
But the scale of demand is testing the limits of available technology and the supply chain.
“Gas turbines have a four-year wait,” Elia noted, explaining that even the most responsive energy sources aren’t able to keep pace in the short term.
Allidina argued that power generation has seen little meaningful investment in decades, and that supply is now “inelastic”, even natural gas, one of the quickest to ramp up, faces long lead times.
“That whole process is inflationary in nature,” he said.
For Canada’s Maple 8, investing in US-based infrastructure may continue to offer the best access to scale as Elia doesn’t see Canada matching the US in this sector.
“I don’t foresee Canada having the same scale, even on a per capita basis,” he said. “The US is positioned in terms of the diversity of its economy, the scale of the businesses that are there, the scale of its government activities. They’re a leader in a lot of those things.”
