Data centre valuation: The billion-dollar question

In our fast-paced digital world, data centres are the backbone of modern infrastructure. They support everything from share trading and banking, to streaming services, weather apps, artificial intelligence, cloud computing, and the Internet of Things.

In the past two years, data centres have caught the eye of investors seeking returns and sector exposure. Yet, high capital costs, technical intricacies, and regulatory challenges pose significant barriers to entry. 

Consequently, we’ve seen major corporate transactions, often overshadowing individual property trades. This trend continues despite ongoing discussions in the property sector about tapping into this asset class.

Mandala Partners¹ forecasts that by 2030, Australia will have more than double the internet-connected devices compared to 2024, driving the need for an additional 1,750 MW of data centre capacity. AusNet² estimates each megawatt costs $9–14 million to commission, suggesting over $20 billion of investment, with Mandala projecting up to $26 billion.

This translates to a need for at least 17 new hyperscale centres in Australia over the next five years. While major operators are acquiring land, planning approvals, grid connections, and development delays will slow progress. Suitable land is scarce, especially in constrained markets like Sydney.

Examples of recently acquired land for data centre development are detailed below:

^{Source: Greenstreet News, MSCI Real Capital Analytics}

Currently, data centre business valuation and takeover transaction pricing incorporates future growth, development pipelines, intellectual property, workforces, systems and processes as well as the underlying property and its significant plant and equipment. 

For investors, lenders, and tax teams, isolating the value of tangible assets (land, buildings, plant and equipment) from the broader business is essential to understand returns and determine taxation obligations.

Data centres in Australia are concentrated around cities, with Sydney and Melbourne hosting most facilities. Proximity to large population centres supports low latency performance for most customer use cases, in particular providing access to skilled labour, established power grids and superior telecommunications infrastructure. Urban clustering also supports redundancy and disaster recovery through multiple nearby facilities. While AI training is less latency‑sensitive, everyday applications still benefit from urban proximity.

From a valuation and tax perspective, site selection is integral to value. For taxable Australian real property (TARP), site goodwill should be included where location (access to power, network connectivity and proximity to customers) creates an advantage that attaches to the site.

For emerging AI and cloud operations, Australia offers several advantages: comparatively abundant land relative to dense Asian cities, a stable policy environment, strong renewable energy generation and the potential for competitive energy costs as the grid transitions. These factors support the case for expansion, though the country faces competition to become an Asia‑Pacific hub.

Various investors are becoming increasingly interested in this sector, but barriers exist, including the need for specialised technical expertise and experience in their design, security, customer relationships and operations.

The following funds have property development arms active in this market:

• Goodman – Worldwide data centre strategy and large industrial land bank, including proposed 90MW build at Artarmon.
• HMC Capital – Purchase of Global Switch Australia at Ultimo for $1.937bn and acquisition of the ‘iseek’ platform for $400m, leading to DigiCo REIT. 
• IFM (previously ISPT) – 170MW proposed data centre at Julius Avenue in North Ryde.

However, despite being flagged as an emerging trend, local real estate investors’ participation has been slower than the enthusiasm suggests. Reasons include:

• Land supply dynamics: In land‑constrained markets such as inner/western Sydney, operators may need to partner with landowners to secure suitable sites. In regions with more industrial land, like Melbourne, operators can acquire sites outright and control their design and delivery processes.
• Planning frameworks and community priorities: Dedicated state processes exist (e.g., NSW SSD pathways for data centres, and Victoria’s Development Facilitation Program), yet urban clusters like Macquarie Park illustrate tensions with housing priorities and the potential for pushback, including around employment opportunities.
• Application pipeline versus real demand: Operators may approach multiple landowners and progress several applications in parallel, creating more development applications than the market will ultimately absorb.
• Operating models and risk allocation: Build‑and‑operate, powered‑shell, and build‑to‑spec variants allocate capex, delivery, and operational risk differently, and not all developers are set up for the associated technical and operational complexities.
• Market structure and competition: Global players are active, and while some have advanced strategies overseas, Australian activity in certain cases remains at earlier stages. Individual assets continue to be planned and marketed (e.g., proposals noted for South Sydney), but conversion to operating stock is uneven.
• Energy transition and long‑term power: Securing scalable, reliable, and increasingly renewable power is central to feasibility. Longer term shifts in the generation mix, including future technologies, could reshape location and cost dynamics.
• Significant upfront capital costs: These often reach into the billions and combined with ramp-up and vacancy risk, together with complexities in tax structuring, make target returns difficult to achieve for traditional property investors.

These constraints suggest participation will remain selective and partnership‑led in the near term, with broader property-level investment more likely as capabilities and relationships deepen and power and planning constraints ease.

Leasing benchmarks are limited because most facilities are owner operated. Where agreements do exist, it can be difficult to separate the value of fixtures and fittings, including significant plant and equipment, from the contractual payments, reducing their usefulness as pure rent evidence.

Hyperscale and colocation facilities are typically occupied under service agreements or contractual licences (occasionally registered leases, though less common in Australia). Charges usually comprise:

Service Level Agreements apply, with credits reducing charges if power, cooling, or other service levels are not met.

In principle, market rent is the amount agreed by informed, willing parties acting at arm’s length after proper marketing and without compulsion. In this emerging asset class, the prevalence of owner‑operators and sale‑and‑leaseback structures can blur evidence of a true market rent.

One way to set rent is by reference to an economic rent. This can be derived using a yield‑on‑cost approach, where net property income is divided by the total cost of developing the land, buildings and installed plant and equipment, or the return a developer would expect on the project. Our research indicates a range of 5.3% to 10.0%, with an average of around 6.7%. Goodman’s annual report³, covering large and complex logistics facilities including data centres, cites a yield-on-cost of 7.5%.

An important crosscheck is affordability. In analogous asset classes such as hotels and certain infrastructure, sustainable rent often sits in the range of 30%–50% of EBITDA, however noting development barriers such as access to power are not relevant.

Using publicly reported EBITDA and total megawatt (MW) outputs of data centre operators we have considered how a market rent range of 30% to 50% of EBITDA compares on a $/kW/month basis, specifically relating to the real estate and installed plant only. 

Market commentary suggests hyperscale charges of $120–$150 per kW per month, typically bundled with services such as connectivity, hardware, maintenance, security, fire protection, and climate control, with power usually passed through. The above analysis supports that for TARP and stamp duty assessments, these service and management components should be excluded to isolate the value of tangible assets (land, buildings, and installed plant and equipment).

We have observed in the market a yield range between 3.50% and 7.04%. Factors such as lease terms, design and useful life of improvements /obsolesce, quality and location are key considerations.

Key sales include:

• NextDC S6: Located in Artarmon, NSW, the property sold as a vacant possession. We note the property sold for $202.4m on an analysed market yield of 6.5%. 
• Telstra Clayton Data Centre: Located in Clayton VIC – The property sold on a yield of 4.2% for $416.7m in 2020 underpinned by the 30-year lease to Telstra and generally tighter market conditions at the time of sale.

Operating data centre sales include:

• The $24bn portfolio sale of global data centre operator AirTrunk. This sale was reported to be a 21x contracted EBITDA multiple. We note that this applies to the entire business operations.
• Global Switch in Ultimo, NSW, to HMC Capital for circa $1.937bn. This property is over 70,000 sqm of GFA, resulting in a rate per sqm of approx. $28,500/sqm. We understand this property sold at a 24.7x EBITDA multiple.  
• Spark, New Zealand NZD $705m sale of 75% of their data centre portfolio representing a 30.8x forward EBITDA multiple. The sale included $236m of property plant and equipment. 

These transactions illustrate that pricing varies with lease security, asset specification and location, and that enterprise‑level multiples should be separated from property‑level yields when deriving valuation benchmarks. Interestingly, on the basis that a 24.7x multiple implies an earnings yield of approximately 4.05%, the scope for a leaseback opportunity appears constrained. This is when benchmarked against prime industrial yields in Sydney, which remain the sharpest in the market at 4.97% - 5.46%⁴ implying no available value arbitrage in disposing of the underlying property asset.

The International Valuation Standards (IVS) made a specific inclusion of ESG/sustainability considerations which came into effect from 31 January 2025. The IVS glossary defines ESG as “The criteria that together establish the framework for assessing the impact of the sustainability and ethical practices, financial performance or operations of a company, asset or liability. ESG comprises three pillars: Environmental, Social and Governance, all of which may collectively impact performance, the wider markets and society.”

ESG will be fundamental to understand in data centre valuation due to evolving technology creating efficiencies in the use of power and cooling, together with customer demand for energy-efficient infrastructure and transparency for their own ESG reporting.

Some facts about data centres that we’ve learned recently from the report by Mandala: Empowering Australia’s Digital Future October 2024¹:

• According to the Australian Energy Market Operator (AEMO), data centres consume 1% of Australia’s total electricity consumption.

• The data centre industry aims to power data centres through renewable energy, driving increased demand for power purchase agreements from renewable energy projects.

• Data centres centralise and specialise in data storage; and without the aggregation in energy-efficient, hyperscale data centres, Australian businesses would consume 67% more energy each year (approximately 2 terawatt hours)

In conversations with several data centre operators, a clear consensus has emerged around improving efficiency and sustainability. Many are focused not only on meeting Australia’s growing demand for capacity, but also on maximising underutilised capacity. To prepare for anticipated regulatory changes, operators are increasingly procuring renewable energy and seeking closer alignment with occupiers’ green energy and digitalisation targets. Over the past two years, investors have likewise sought to balance profitability with sustainability, targeting assets already powered by renewables or with a clear path to be.

It is likely, with possible policy changes and customer focus on ESG features and with ESG now embedded in valuation standards and customer expectations, data centres that can lead on efficiency and credible renewable procurement will realise lower costs, stronger demand and more resilient valuations. 

Australia’s digital economy is accelerating, and demand, capital, and ambition are converging to make data centres a compelling investment, although execution is complex. The specialised nature of the infrastructure, the skills required to build and operate it, and ongoing capacity bottlenecks have created a persistent gap between property and business valuations. To meet tax requirements, property valuations must capture all relevant elements, from location-specific goodwill to ESG attributes of the built form. As the market matures, opportunities will expand, and those who establish the right partnerships and frameworks now will be best positioned to capture returns with confidence.