Artificial intelligence and digital transformation are the hottest themes in infrastructure investing, not only among private investors but also increasingly state governments.
Just this month, Saudi Arabia unveiled one of the biggest state AI infrastructure investments, with plans to buy billions of dollars’ worth of advanced chips from US manufacturers including Nvidia and AMD.
But despite the overwhelming bullishness, some asset owners are wondering when – or if – AI can deliver a miraculous productivity gain and benefit the underlying infrastructure such as data centres.
“I am actually a little bit more – let’s call it sceptical – of AI in the longer-term than I maybe started with,” said Nick Khaw, head of research and co-head of private market at the $38 billion Khazanah Nasional, at the Top1000funds.com Fiduciary Investors Symposium at Harvard University.
“As a macro economist, I think one thing that’s pretty clear, and it’s been documented by economists from down the river at MIT, is that it [AI benefits] really hasn’t shown up in any productivity numbers.
“So either we’re measuring productivity wrongly, which is possible, or it’s not as general purpose as we think.”
With that said, AI and digital transformation is one of the four megatrends Khazanah is monitoring (alongside climate, demographics and de-globalisation). The Malaysian sovereign wealth fund’s philosophy to private markets investment is to look for “something which looks like a headwind today, but it’s a tailwind in the future, because you can buy low and then hopefully get a return later”, Khaw said.
Compared to pure AI technologies such as large language models, Khaw believes robotics is more likely to have a meaningful impact on the real economy particularly in relation to blue collar jobs. But that comes with its own set of potential problems.
“I do worry about data centre usage, but I also wonder for some of these things… if the productivity numbers don’t match up, or if they do and people lose jobs, will the politics push back on something like AI and say, look, too many people are losing jobs. We can’t afford a universal basic income. Maybe let’s stop doing this stuff.”
An interesting dilemma for pension funds in the scenario of new technologies leading to jobs losses is that the workers impacted could be their fund beneficiaries.
When asked how the fund balances its investment and fiduciary needs, Andrew Siwo, head of sustainable investments and climate solutions at New York State Common Retirement Fund, said the fund has set important parameters around workers’ rights protection in portfolio companies.
“I would say more broadly, we do have, at least in private equity, a responsible workforce management policy that addresses our expectations for sound labour management principles,” he said. The policy directs that, for example, the fund’s private equity managers should encourage industry standard wages and benefits and minimise adverse impact on workers when there are mergers and acquisitions.
While opportunities around AI are attractive, Siwo added there are risks and inaccuracies prevalent in AI that are concerning.
“New York State Comptroller, Thomas DiNapoli, the fund’s sole trustee, released an AI audit report recommending the use of governance structures to prevent AI abuse and inaccuracies. Each investment manager that has received capital from us must complete a scorecard to assess material investment risks/opportunities including environmental, social, and governance factors.”
Resilient to volatile trades
Executive director at IFM Investors Adrian Croft is more optimistic about AI and said it is “the most consequential megatrend” for infrastructure investment now. Its most prominent manifestation is data centres, but that booming demand also extends to energy infrastructure due to the substantial need for power.
“There’s certainly a case for a huge amount of potential investment in energy infrastructure, generation and grid enhancement,” he said.
“Renewable energy is going to play a huge part of that… but we think there’s still going to be a role for gas because these data centres do need 24/7 firm capacity.”
But an even longer-term play is fibre networks, which Croft admitted haven’t been the easiest area for equity investments – at least in the US and parts of Europe. This is induced by issues including rising build costs and overbuilding in some markets.
“It has been a tough spot, but the development of large data centres in secondary or emerging data centre markets are all going to need to be connected, so there could be a really good case for more fibre,” Croft said.
While global companies are rushing to establish domestic supply chains amid the uncertain trade environment, Croft believes infrastructure is an asset class resilient to the ongoing impact of localisation.
“If your infrastructure is essential to the community it serves, it’s still going to need to be there. People are still going to keep using energy, water and gas. They’re still going to need to get to work, get to school or get home in the evening. They’re still going to want to communicate and use the internet,” he said.
“But not everything is going to be unscathed… there’s a lot of focus on what’s going to happen with global ports in particular, with volume through US ports going to drop precipitously in coming months. We’ll see.”
With risks come opportunities, though, as Croft believes there could be more demand for local infrastructure like logistics and cold storage.
Granted, “it might not be the most efficient way of doing things if you have to replicate what already works pretty well in various parts of the world,” but it could add more opportunities across the spectrum from core infrastructure, infrastructure adjacencies, to value-add strategies, he said.
Specifically in relation to AI infrastructure, Khazanah’s Khaw said a driver of localisation is data sovereignty concerns, which may prompt companies to keep centres capable of processing data for advanced AI applications domestically.
