Making sure portfolios aren’t overly concentrated in big tech companies might be sensible from a valuation perspective, but it’s not the same as removing AI risk, which is now beginning to permeate credit and real assets too, according to a paper from Mercer.  
 
“Beneath what might look like a well-diversified top-down portfolio, these accelerating megatrends create what we might call ‘stealth concentration risk’,” says the paper, titled ‘Rumours of my demise have been greatly exaggerated’: The future of private markets’. 
 
“It is therefore critical that asset owners understand the types of risk they hold in their portfolio and where they hold it. Without integrated risk oversight, investors may hold duplicated exposures across diversified buckets. This is why we believe that managing investment risk will increasingly rely on more than top-down diversification.”  
 
Mercer says that the AI thematic and the risk that comes with it is making its way into real assets through surging demand for electricity – which benefits funds exposed to utilities, transmission networks, renewables and natural gas generation – as well as the growth in data centres and e-commerce logistics hubs. 
 
“These investments, though labelled as real assets, are partly technology exposed due to AI-related demand and partly energy and climate infrastructure exposed due to power intensity and location dependence,” the report says.  
 
“AI and data centre exposure might also be present in private equity due to exposure to high-growth companies operating within the AI sector. Indeed, we are already seeing some cross-over in markets, with core, open-ended infrastructure funds acquiring assets from closed-end funds in private equity, real estate, and infrastructure. Without integrated risk oversight, investors may end up holding duplicated exposures across ‘diversified’ buckets.”  
 
At the 2025 Fiduciary Investors Symposium in Healesville, Victoria, a number of asset owners said that they are scrutinising their portfolios to determine where they might have exposure to common risks as AI-related themes drive explosive growth across a number of different asset classes. Justin Pascoe, head of portfolio construction and execution at Cbus, said that fund had undertaken an exercise where the fund’s investment committee looked at the total portfolio exposure to AI.  

“They said we’re not exposed through cash, but there is some in the private credit part of the portfolio, property, infrastructure, global equities – and we’ve also got certain companies here in Australia that are exposed to that theme,” Pascoe said.  
 
“If you’re an active manager you might be underweight the Mag Seven, but in the belly of your portfolio – the real asset part of the portfolio – you’re probably overweight AI. So you need to ask ‘where is the best risk-adjusted way to play it?’ rather than just saying ‘I’ve got this bucket in real assets that I have to fill’.”  
 
At the same event, William Scott, head of real assets at Commonwealth Superannuation Corporation, said that, when considering common risks across their portfolio, asset owners needed to take a discerning view of data centres and their exposure to AI and growth.  
 
“Your exposure to AI growth and equity could be as low as triple net lease of 15 years for a cold shell to Microsoft. Microsoft is AAA-rated credit, so you’re probably pretty confident of that investment reaching target return. In terms of obsolescence, you’ve rented a cold shell so you’re in a good position to repurpose that for other things if you want.   
 
“But that goes all the way to the other end of the spectrum, where you’re a start-up data centre developer, looking to get land and power and make money by flipping that quickly to someone who will then contract it. That’s super exposed, because even a change in the rate of growth of AI demand will dramatically affect the value of that pipeline. There’s a really big range of exposures to the AI risk you can get even within data centres.”