Data Centre REITs and Digital Infrastructure: The AI-Driven Property Play

Data centres are the physical backbone of the digital economy. Every search query, streaming video, AI inference, and cloud application runs through a facility that houses servers, networking equipment, and the power and cooling infrastructure required to keep them running. As artificial intelligence workloads surge, data centre demand has grown at a pace that has outstripped supply in most major markets, turning digital infrastructure into one of the most sought-after property subsectors in the world.

Why Data Centres Are Real Estate

A data centre is, at its core, a building — one that requires planning consent, construction, power grid connection, and ongoing maintenance. What makes data centres different from conventional warehouses or offices is the density of power consumption and the criticality of the operations inside them.

Modern hyperscale data centres consume 50–200 megawatts of power — enough for a medium-sized town. Purpose-built cooling systems, redundant power supplies (N+1 or 2N architecture), raised floors or under-floor cooling, and physical security infrastructure all add capital cost. A single hyperscale campus can represent a capital investment of £1–5 billion.

For investors, this means high barriers to entry, long-duration asset lives, and sticky tenants — hyperscalers and enterprise cloud users do not move data centre locations lightly. Lease terms are typically 10–20 years with built-in rent escalators.

The AI Compute Demand Surge

The deployment of large language models and generative AI has materially changed the data centre demand outlook. Training a large AI model requires thousands of GPUs running continuously for weeks or months; inference (serving responses) requires consistent GPU capacity at scale. NVIDIA's data centre revenue exceeded $40 billion in FY2024 alone — a rough proxy for the extraordinary investment flowing into AI compute hardware.

The International Energy Agency estimated in 2024 that data centres accounted for approximately 1–1.5% of global electricity consumption, with that share projected to roughly double by 2030 driven primarily by AI. In the UK, National Grid and NESO have flagged data centres as among the fastest-growing sources of grid demand.

This surge has created a capacity crunch. Grid connection queues in the London-Slough corridor, northern Virginia, Dublin, and Frankfurt stretch to five to ten years in some cases — a genuine constraint on supply growth that supports rental values for existing capacity.

The Power Constraint Problem

Power is the binding constraint on data centre growth in most developed markets. Land is available; construction is straightforward; but securing a grid connection of sufficient capacity has become an obstacle measured in years rather than months.

UK Power Networks, National Grid Electricity Distribution, and other distribution network operators have faced a surge in connection requests from hyperscalers and data centre operators. Projects that applied in 2022 may not receive connections until 2030 or later. Some operators have responded by investigating on-site generation — combined heat and power, small modular reactors, and direct renewable power purchase agreements — as ways to reduce grid dependency.

This dynamic has several investment implications. Existing data centres with established grid connections carry significant scarcity value. New entrants face high barriers. And energy transition infrastructure (renewable energy, battery storage, grid investment) is increasingly linked to data centre growth as both driver and enabler.

Hyperscaler Capex and the REIT Model

The very largest data centre users — Microsoft, Google, Amazon (AWS), Meta — have been investing directly in data centre construction rather than relying entirely on colocation REITs. Microsoft announced plans to invest $80 billion in data centres in FY2025 alone; Google and Amazon have made similar commitments.

This raises a question for REIT investors: if hyperscalers build their own capacity, does that reduce demand for third-party colocation? The answer is nuanced. Hyperscalers do both — they own facilities and they lease from third parties when speed-to-market, location, or flexibility justify it. Colocation REITs retain a significant role for enterprise customers, edge deployments, and markets where hyperscalers have not yet built their own facilities.

Major Listed Data Centre REITs

Equinix (EQIX) is the global market leader in colocation, operating more than 260 International Business Exchange (IBX) data centres across 71 metropolitan areas on six continents. It is the world's largest data centre REIT by market capitalisation and revenues. Equinix's network-dense facilities — many interconnected with the Internet exchange points that route global traffic — have an almost impossible-to-replicate strategic position.

Digital Realty Trust (DLR) is the second-largest, with a broader range of facility types including hyperscale campuses and colocation. Its PlatformDIGITAL ecosystem links facilities globally to offer multi-site deployments.

Iron Mountain (IRM) has pivoted from physical document storage towards data centres, a transformation that has significantly re-rated its equity.

CyrusOne and QTS Realty Trust were both taken private in 2021–2022 by KKR and Blackstone respectively, illustrating the private capital appetite for the sector.

In the UK, purely listed pure-play data centre equities are limited. SEGRO (SGRO) has data centre exposure through its urban logistics and industrial portfolio. Tritax EuroBox (EBOX) has some digital infrastructure exposure in its Continental European portfolio. UK investors typically access the sector through US-listed REITs (subject to withholding tax considerations) or through diversified digital infrastructure ETFs.

Cooling Technology and Carbon Intensity

Data centres are power-intensive and water-intensive. Conventional air-cooled data centres use evaporative cooling towers that consume significant water. In water-stressed regions, this is increasingly a social licence issue.

Liquid cooling technologies — direct liquid cooling (DLC), where coolant is circulated directly to chips — and immersion cooling, where servers are submerged in dielectric fluid, are gaining traction. These approaches can be three to five times more energy-efficient than conventional air cooling, materially improving power usage effectiveness (PUE) ratios.

For AI workloads, where GPUs run at extremely high power densities (400–700W per chip), conventional rack air cooling is often insufficient. AI data centres typically require liquid cooling, making the capital cost higher but the operational efficiency better.

Carbon intensity remains a reputational and regulatory issue. The UK's Climate Change Committee and the EU have both flagged data centre emissions as a growing concern. Many operators have committed to matching consumption with renewable energy (through PPAs or RECs), though the definition and timing of "matching" is disputed.

Investing in Data Centres Beyond REITs

Beyond the pure REIT route, investors can access the data centre theme through:

• Data centre REITs ETF (DTCR): focused ETF available in US-listed format; UCITS equivalents are limited.
• iShares Digitalisation UCITS ETF (DGTL): broad digital infrastructure including data centres.
• Brookfield Infrastructure Partners (BIP): owns data centre assets as part of a broader infrastructure portfolio.
• VanEck Digital Assets Infrastructure UCITS ETF: partial exposure through digital infrastructure equities.

Risks

Valuation. Data centre REITs have historically traded at premium valuations — Equinix often trades at 30–40x EBITDA. A growth slowdown, AI demand disappointment, or interest rate rise can compress multiples significantly.

Power and sustainability constraints. Inability to secure grid connections delays expansion and limits revenue growth. Regulatory action on carbon could impose significant costs.

Technology disruption. Efficiency improvements (lower power per computation, model distillation) could reduce power demand per unit of AI inference, slowing facility expansion requirements.

Geopolitical risk. Data sovereignty laws (EU GDPR, India's DPDP Act, China's Cybersecurity Law) create complexity for global data centre operators and may fragment the market.

The value of investments can fall as well as rise. Yields and valuations can change materially. This guide is educational and does not constitute investment advice. Always seek professional advice appropriate to your circumstances.

How Global Investments Can Help

Digital infrastructure is one of the most significant structural investment themes of the coming decade. Our team can help you assess the appropriate exposure, select between REIT, equity, or fund approaches, and manage concentration risk in a theme that has attracted significant investor interest and — in some vehicles — demanding valuations. Contact us to explore how data centre and digital infrastructure exposure could complement your existing portfolio.