Data centre real estate has been one of the highest-growth commercial property segments globally for the past decade and remains among the most actively bid asset classes in Australia. The combination of cloud computing demand, AI-driven workload growth, the post-2020 acceleration of digital infrastructure spending, and limited supply of power-connected sites has driven both cap rate compression and material development activity.
For private investors, direct data centre real estate exposure is constrained: most institutional-grade stock is held by REITs (Goodman, Stockland, Charter Hall), global infrastructure funds, or specialist data centre REITs (NextDC, AirTrunk, the global Digital Realty and Equinix). The available private-investor access points are smaller-format colocation facilities, edge data centres, and specific syndicated investments.
Data centre real estate is more infrastructure than property. Power supply, cooling, fibre connectivity, and operator covenant are the principal value drivers. The building is the wrapper; the operational complex inside is the asset.
The Three Principal Data Centre Types
Hyperscale
Purpose-built facilities for single-tenant occupancy by major cloud providers (AWS, Microsoft Azure, Google Cloud, Oracle, Meta). Typically 30 to 200+ MW power capacity, very long-WALE leases, and institutional-scale capital requirements. Largely inaccessible to private investors at the asset level.
Colocation
Multi-tenant facilities where multiple enterprise and SME customers lease cabinet, cage, or partial-room space. Operators include NextDC, AirTrunk (acquired by Blackstone), Equinix, Digital Realty Sydney/Melbourne. Some private-investor access via secondary market or syndicated investments.
Edge data centres
Smaller facilities (typically 1 to 5 MW) located closer to end users, supporting low-latency applications. Smaller capital requirements; some private-investor accessibility as the segment grows.
1 The Power Question
Power supply is the principal constraint on data centre development. Three components matter:
Connection capacity
Local grid capacity at the site. Major grid upgrades for new data centre developments are common; the timeline for connection can be 3 to 7 years for large new facilities. Existing operational facilities with established connections have substantial value purely from the power connection.
Cost
Electricity is the largest operating expense at a data centre. Wholesale electricity pricing trajectory directly affects operator economics.
Source
Renewables-sourced power is increasingly required by hyperscale tenants. Power purchase agreements (PPAs) with solar and wind generators, on-site solar, and green energy certificates are all common structures. Facilities without a credible renewable pathway are increasingly disadvantaged in winning hyperscale leases.
2 Lease Structures
Hyperscale leases
Long-WALE (15 to 30 years), triple-net, with strong creditworthy tenants. Rent is typically per-megawatt or per-room basis, with fixed annual increases. The tenant operates the IT environment; the landlord provides power, cooling, and connectivity infrastructure.
Colocation leases
Shorter contracts (1 to 5 years for individual customers, with the underlying facility lease longer). The operator manages the multi-tenant environment; the property investment is in the building and infrastructure.
3 Building Specification
Data centre buildings are specialised infrastructure rather than standard commercial buildings:
- Power infrastructure. Substation, distribution, redundancy (typically N+1 minimum, often 2N for hyperscale), UPS systems, backup generators.
- Cooling. Chillers, cooling towers, hot/cold aisle containment, increasingly liquid cooling for high-density compute.
- Connectivity. Fibre entry points, redundant fibre paths, on-site cross-connect infrastructure.
- Floor and structure. Raised floor (older facilities), slab-on-grade with overhead distribution (modern), high static load capacity.
- Security. Multi-layer physical access control, biometric, surveillance, often facility certification.
4 The Power Usage Effectiveness Question
Power Usage Effectiveness (PUE) measures total facility energy consumption divided by IT equipment energy consumption. Lower PUE = more efficient cooling. Modern hyperscale facilities target PUE of 1.2 to 1.3; legacy enterprise facilities can be 1.6 to 2.0.
A lower PUE means more usable IT capacity per megawatt of grid connection, which is a fundamental economic driver. Facilities with structurally high PUE will either lose tenants or require costly cooling upgrades over the lease term.
5 The AI Workload Effect
AI training and inference workloads have driven a substantial increase in compute density per rack. Traditional data centres designed for 5 to 10 kW per rack are being upgraded or rebuilt for 30 to 100 kW per rack. The implications:
- Cooling infrastructure must support higher heat loads.
- Power density per square metre increases.
- Legacy facilities require capex to support AI workloads.
- Greenfield development for AI-specific facilities is accelerating.
6 Buyer-Side Access Pathways
Direct asset acquisition
Generally restricted to institutional-grade buyers. Mid-market colocation and edge data centre acquisitions are possible at private-investor scale but require specialist DD.
Listed REIT exposure
NextDC (NXT.ASX), Goodman Group (GMG.ASX, partial data centre exposure), Charter Hall industrial trust (CHC). Listed exposure provides liquidity but introduces market-price volatility uncorrelated with the underlying asset performance.
Syndicated investment
Some property syndicators offer fractional ownership in data centre assets. Specialist research before commitment.
Private credit
Lending to data centre developers via private credit funds. Different risk profile from equity ownership.
7 Yields and Pricing
Hyperscale-tenanted data centres trade at among the tightest yields in the global commercial property market. Australian institutional data centre yields are competitive with Singapore and Tokyo equivalents. Colocation and edge data centre yields trade wider, reflecting operating complexity and lower covenant.
Frequently Asked Questions
Can a private investor realistically buy a data centre?
Direct asset acquisition is constrained by minimum ticket size (typically $50 million+) and the specialist DD required. Edge data centres and smaller colocation facilities are more accessible. Syndicated and listed exposures provide alternative access.
How does the AI workload affect existing assets?
Legacy facilities require capex to support higher-density compute. The implementation cost and timeline depend on the building specification; modern facilities have more capacity for upgrade than legacy.
Is the segment supply-constrained?
Yes, principally by power supply. Greenfield development cycles of 5 to 7 years mean supply lags demand materially. The supply constraint is the structural reason for sustained yield compression.
What about the environmental cost?
Data centre energy consumption is substantial and growing. Operator commitments to renewable power, water-efficient cooling, and heat reuse are increasingly material to tenant decisions. The transition to renewable-sourced operations is a major capital and operational program.