By Jim Wright, fund manager, Premier Miton Global Infrastructure Income fund
One of the most attractive aspects of the listed infrastructure sector is it can provide investors with exposure to themes that are multi-year, and in many cases, multi-decade in duration.
This became particularly clear to us last year when, amid an economic and political backdrop that at times seemed to change by the day, we were able to rely on fundamental long-term structural trends. Chief among these is the continued electrification of Western markets, which implies ongoing growth in electricity demand. This growth requires sustained investment in networks and in all types of power generation, including natural gas, renewable energy, storage and, over time, new nuclear capacity.
Artificial intelligence, which to date has been closely linked with technology stocks, also has clear implications for listed infrastructure, as rising electricity demand is being augmented by the rapid growth in AI computing power and data centres.
There is no doubt that a great deal of hype and exuberance surrounds AI, which arguably represents an accelerator of upward trends in energy demand rather than a fundamental departure from them. That said, plans for new data centre capacity are very real and are already beginning to impact actual investment decisions across the regulated utility sector. Capital has been committed by hyperscale investors, there is a significant buildout of data centre capacity underway in the US, and these data centres require a reliable and resilient power supply to operate.
The question of how much additional generation will be required, and where this power will come from, is central to the successful rollout of AI across advanced economies.
Powering the digital economy
Data centres may seek to access power from captive sources, either through new-build capacity or the reallocation of existing generation, using behind-the-meter solutions. Alternatively, they may connect to local electricity grids and draw power alongside existing customers through front-of-the-meter solutions.
On-site power requirements have attracted a number of innovative approaches, often involving technologies that are unproven at scale and have long lead times, such as small modular nuclear reactors. It is plausible that a material proportion of new data centre capacity will ultimately include some onsite generation capability.
See also: HSBC AM’s Corona: The secular trends driving infrastructure
However, it is increasingly apparent that grid-based solutions are preferred by hyperscale operators. The reality is the grid offers the insurance of multiple sources of power, which is particularly important given data centres are 24-hour users of electricity and must avoid supply disruption at all costs.
For regulated utilities, the opportunity to benefit from increased load growth is clear, and there is now tangible evidence of this in capital expenditure plans and forecasts for rate base growth through to the end of the decade.
Balancing growth with affordability
Grid-based solutions do, however, create understandable concerns around affordability, focused on two related issues.
First, whether increased power demand from data centres in markets where reserve margins are tight could lead to higher electricity prices for existing industrial, commercial and residential customers. Second, whether the investment required in grids and new generation to support data centre growth will be spread across the broader customer base, increasing costs for customers who do not receive corresponding benefits.
Encouragingly for investors, a number of utilities have begun to develop strategies aimed at separating the costs of incremental data centre growth and protecting existing ratepayers.
See also: Watts it worth: Why smarter infrastructure will power the AI economy
In Indiana, NiSource has established a commercial partnership with Amazon within a new framework that effectively ring-fences the investments required to power Amazon’s data centre. This partnership is expected to deliver approximately $1bn of cost savings to existing customers over the 15-year life of the contract, while also providing access to low-cost energy and improved system reliability.
Across its service territory, American Electric Power is guiding to average residential customer rate increases of approximately 3.5% annually over the forecast period, with costs expected to be borne by other customer classes that are driving the incremental investment. In Wisconsin, WEC Energy has developed a specific regulatory model for data centre demand through its Very Large Customer tariff, which is explicitly intended to meet the needs of very large customers while protecting other customers and shareholders.
AI is reinforcing a long-term growth trend
It is reasonable to assume data centre customers will be prepared to pay their fair share of the investment required in power generation. Where this investment is delivered through front-of-the-meter solutions from regulated utilities, tensions are likely to arise.
However, where utilities can establish structures that separate the specific costs of generation and transmission associated with data centre growth, they are likely to be successful, even under increased scrutiny from regulators and politicians with a strong focus on affordability.
The scale of change required to support AI, and how this will ultimately affect consumers, remains open to debate. What is clear is AI-related electricity demand is already influencing investment and regulation across the utility sector. Rather than representing a fundamental shift, AI reinforces the long-term electrification trends already underway and provides another durable source of growth for utilities and infrastructure investors alike.
