Tim Humphreys: Why essential infrastructure is the true power behind the AI revolution

For professionals only.
Capital at risk.
In an article first published on Wealth DFM, Tim Humphreys highlights how power companies can provide a ‘picks and shovels’ route to investing in the growth of artificial intelligence.
Richland Parish occupies around 570 square miles of rural Louisiana. Founded in the late 1860s, it was named in recognition of its suitability for farming. Traditionally, it has specialised in producing corn, cotton and other crops.
All things considered, the area is unlikely to be quite so reliant on agriculture in the future. Richland Parish will soon be the home of a new “titan cluster” – an enormous data centre designed for the next stage of the AI revolution.
Nicknamed Hyperion and set to be erected alongside a stretch of Highway 183, the state-of-the-art facility was conceived largely in secret by Meta. It will create thousands of jobs and is expected to be fully online by 2032.
Meta CEO Mark Zuckerberg celebrated the project’s go-ahead by posting an image in which Hyperion is superimposed onto an aerial photograph of Manhattan. It shows the centre would be large enough to blanket around a fifth of New York’s most expensive real estate.
Yet this jaw-dropping snapshot doesn’t serve merely as evidence of Hyperion’s monumental scale. In its own way, it also underlines the magnitude of the AI-fuelled transformation now taking place all around us.
In my opinion, investors could do worse than to sear the pic into their collective memory. Why? Because it’s hard to find a more striking illustration of how the AI build-out is adding to the appeal of global essential infrastructure.
The HALO effect
The narrative around AI routinely focuses on the quest to achieve superintelligence and agentic automation. As a result, news flow and investment thinking alike tend to be dominated by a familiar handful of tech giants and chipmakers.
In fact, the usual suspects constitute only a fraction of this far-reaching story. For instance, it’s well worth noting that the march of AI has also propelled us to the brink of an arguably unprecedented step change in energy demand.
This has already helped bring about what has become known as the HALO framework. HALO stands for “heavy assets, low obsolescence”, and it’s difficult to overstate the likely significance of this combination in shaping global growth over the years and decades to come.
Despite their huge relevance to the AI boom, heavy assets aren’t the stuff of software, code, algorithms and intellectual property. They’re instead the stuff of pipelines, railways, telecommunications towers, battery storage, cooling equipment, major industrial machinery and the like.
Why are such assets likely to enjoy low obsolescence in the age of AI? The straightforward answer is that they should remain essential. Even today’s game-changing technological advances don’t mean that electricity no longer needs to flow, that goods no longer need to move or that physical systems no longer need to be maintained.
In my view, the key word here is “power”. It’s the sheer power required by the spread of AI and the proliferation of data centres, along with all the associated essential infrastructure, which is likely to underpin some of the brightest investment opportunities as radical disruption continues to remould almost every aspect of day-to-day life.
A classic “pick-and-shovels” approach
So is the lesson simply that everyone should invest in data centres? Not exactly. Although their criticality to AI’s rise is beyond question, they might not meet the strictest definitions of essential infrastructure.
This is because data centres operate in an extremely competitive environment. The barriers to entry aren’t particularly high in this arena, contracts are normally short, and users can switch providers in response to pricing and capacity. At least relatively speaking, obsolescence is a possibility.
The HALO ideal is much more likely to be found upstream. Electricity grids, transmission networks, utilities – these are the kinds of asset that can allow investors to profit from AI without embracing an undue level of risk.
Take the US’s biggest renewable electricity producer, NextEra, which we see as one of the most direct beneficiaries of the AI build-out. Its new contracts include a 25-year agreement with Google to recommission a retired nuclear plant that will support the company’s AI infrastructure in the Iowa region.
Ultimately, the HALO approach represents a classic “picks-and-shovels” play. There’s no great need to worry about which hyperscalers will eventually win the AI race, because these assets are going to prove pivotal in any event.
The businesses that truly enable the AI revolution may never have the headline-grabbing impact of the awesome edifice that will one day loom over Highway 183 in Richland Parish. But that doesn’t mean their long-term investment attractions should be overlooked.
Tim Humphreys is co-manager of the IFSL Marlborough Global Essential Infrastructure Fund.
Find out more about our Global Essential Infrastructure Fund
This article is provided for general information purposes only and should not be construed as personal financial advice to invest in any fund or product. These are the investment manager’s views at the time of writing and should not be construed as investment advice. The opinions expressed are correct at time of writing and may be subject to change. Capital is at risk. The value and income from investments can go down as well as up and are not guaranteed. An investor may get back significantly less than they invest. Past performance is not a reliable indicator of current or future performance and should not be the sole factor considered when selecting funds.

