The discipline of AI exposure

Artificial intelligence (AI) workloads have introduced a new category of demand into the data centre sector, distinct from cloud/enterprise computing that shaped growth in the prior decade. Large AI models require far higher densities of computing power, concentrated in specialised, GPU-intensive facilities that consume electricity at record scale. Access to power is critical, creating a genuine opportunity for investors in real assets. The scale of capital required to facilitate the buildout is enormous, calling on public and private, debt and equity capital markets.

Private equity, project financing or unlisted infrastructure funds typically demand long lock-up periods, high minimums and limited asset-level visibility, at a time when technology and its power requirements are continually evolving.

We believe public markets offer a more accessible and sustainable path into the AI-related data centre investment theme. Listed property and listed infrastructure provide differentiated exposure to the value chain, with each benefiting from distinct demand drivers and risk profiles.

Listed property can provide exposure to data centres via vehicles that typically own the physical facilities and network ecosystem (not the chips and servers).

Listed infrastructure sits further upstream, in the power generation, transmission and grid assets that AI data centres depend on to operate, with returns driven more by regulated or contracted revenue and long-duration capacity investment.

Given how tightly AI compute growth is now bound to power availability, the case for holding both forms of exposure, rather than treating them as interchangeable, is arguably stronger here than for digital infrastructure in general. This paper sets out that case.

The benefit of options

The AI capex spend could prove transformational to the economy, whilst creating significant demand for the hard assets that enable it. Yet many of the high-profile hyperscale data center facilities being developed by private equity newcomers to the industry do not necessarily offer investors the best risk-adjusted returns.

Instead, we believe exposure to the digitalisation theme is best accessed through listed data centre platforms and/or the regulated electric utilities.

Property play

Hyperscale facilities can be highly attractive development opportunities. Purpose-built campuses leased to investment-grade hyperscalers can generate double-digit yields on cost for the developer, supported by long duration (15-20 year) leases featuring contracted cash flows, embedded annual escalators (2%+), and tenants whose credit quality is among the strongest globally. 

In many respects, these assets resemble contracted infrastructure investments, with predictable earnings streams and limited near-term leasing risk. Equally important, hyperscale developments are among the few real asset opportunities capable of absorbing tens of billions of dollars of capital within a single project or campus, making them particularly attractive to large scale institutional investors seeking both deployment capacity and predictable cash flows.

However, residual value remains the key underwriting consideration. Unlike traditional real estate, hyperscale facilities are frequently bespoke assets designed around the requirements of a single customer, and many are increasingly being developed in secondary or power-abundant locations rather than established data centre hubs. While the contracted cash flows may be sufficient to justify the initial investment, the value of the asset beyond the lease term is inherently less certain. 

Technological change, evolving power and cooling requirements, and the challenge of re-leasing a highly specialized facility create a wide dispersion of potential terminal value outcomes.

As shown in the table below, developers can achieve satisfactory IRR’s (7-13% depending on different terminal value assumptions), assuming 9.5% initial yield on development cost with 2.5% annual escalators over the course of a 15-year lease.

However, an investor acquiring the same data centre from the aforementioned developer at a lower initial yield (circa 6.0% as is often assumed by industry participants¹) face much lower returns (sub 4% IRR) if the residual value is 30% or less of their purchase price. In our view, the prudent investor should make conservative assumptions regarding the terminal value of such assets given the risk associated with their relatively remote location, bespoke design, re-leasing prospects, technological relevance, and alternative use value.

We believe those residual value risks are less acute when investing through a listed data centre REIT. Leading global data centre platforms such as Equinix (EQIX) and Digital Realty (DLR) own diversified portfolios across major global data centre markets, serving thousands of customers rather than a single hyperscale tenant.

Their assets are generally located in Tier 1 markets with rich connectivity where demand is deepest, replacement costs are highest, and barriers to entry are most acute. Importantly, many facilities support retail colocation and interconnection ecosystems, particularly in the case of EQIX, creating network effects and customer stickiness that extend beyond the underlying real estate. This diversification reduces exposure to any individual tenant², workload, or technology cycle, while improving the fungibility of assets should customer requirements shift over time. In our view, ownership of scaled and highly connected platforms in strategic markets provides a more durable risk-adjusted avenue to access secular growth in data centres compared with concentrated exposure to individual hyperscale facilities in pioneer locations.

Power play

Power has been a widely publicised constraint of the AI build-out stemming from the electricity-intensive nature of these facilities. Meeting this new demand requires targeted and substantial investment in generation, transmission and distribution infrastructure.

While this would naturally raise the same risks and questions from investors around residual value, the regulatory framework in which monopoly electric utilities operate within provides meaningful protections around long-term returns and asset values. As such, the monopoly electric utilities undertaking this investment are where the more durable, regulator-underwritten economics reside.

U.S. listed utility, Entergy (ETR), illustrates why the regulated electric utility is an attractive way to access AI-driven power demand. ETR is the monopoly utility supplying power to Louisiana where Meta’s Hyperion data centre is being developed. ETR plans to build ten new natural gas-fired generation plants, grid-scale batteries and renewables, with potential nuclear expansion beyond that. This investment is deployed into the regulated asset base, where the return is set by the utility regulator rather than negotiated customer-by-customer. Entergy Louisiana has an authorised 9.7% return on equity, equivalent to a 6.95% after-tax WACC, and begins accruing that return during construction.

This regulatory underpinning becomes crucial given the scale of the deployment, estimated to be at least US$20bn across generation, transmission and storage. Moreover, ETR has secured minimum payment contracts of 15 to 20 years, guaranteed by Meta and unaffected by any future change in the site’s ownership or financing. These minimum payments cover capital costs, operating expenses and ETR’s approved return for shareholders and debtholders, protecting local customers from bearing these facility-specific costs should the site draw less power than anticipated. 

That said, while AI-related load growth is additive to ETR’s already robust growth profile, the success of any single facility is not existential to the utility’s defensive earnings base.

Two Paths, S(ai)me Opportunity

The scale of investment required to meet AI compute demand is not in question. What remains less certain is the route investors take to capture the best risk-adjusted returns. Direct or unlisted exposure to AI data centres, whether through project financing, private equity or specialist real asset funds, concentrates risk in ways that are easy to underestimate: single-tenant dependency, construction and technology obsolescence risk, illiquidity, and limited ability to reprice or exit as conditions change. These are not reasons to avoid the theme, but rather, reasons to carefully consider how that exposure is applied.

Listed property and listed infrastructure offer two distinct, complementary paths into this growth, each avoiding many of the risks that concentrated, illiquid exposure to data centres typically carries. 

Listed property spreads risk across a portfolio of tenants, locations and lease structures rather than a single facility or operator, while retaining daily liquidity and market pricing. Furthermore, Equinix and Digital Realty provide networking infrastructure critical to the operation of the internet, which should endure regardless of the success of AI.

Listed infrastructure, in turn, provides exposure to the power and grid assets that AI compute depends on, often underpinned by regulated or contracted revenue that offers a degree of earnings visibility uncommon in direct real asset investment.

Neither approach eliminates risk entirely, but each addresses a different part of it, and together they allow a portfolio to participate in the structural expansion of AI infrastructure without concentrating that participation in a single point of failure.

As this theme continues to develop, the quality of execution, in asset selection, portfolio construction and ongoing risk management, will matter as much as the decision to seek exposure in the first place. 

Investors who approach AI data centre growth through a disciplined combination of listed property and listed infrastructure assets are better placed to capture its potential upside while retaining the liquidity, transparency and diversification that experienced investors rightly expect.

Managed Fund

Resolution Capital Real Assets Fund

Australian Shares

View

ETF

Resolution Capital Global Listed Infrastructure Fund - Active ETF (RIIF)

Alternative Assets

View