Colocation vs Hyperscale: Which Data Centre Model Is Right?

Two Models, One Decision

The data centre industry operates on two fundamentally different models: colocation (multi-tenant facilities where space, power, and cooling are leased) and hyperscale (single-tenant purpose-built campuses). Understanding the differences is essential for making the right infrastructure decision.

Colocation: Shared Infrastructure

You rent rack space, power, and cooling. The provider handles the building. You handle your hardware. The trade-off: you get speed to market (weeks, not years) and OpEx pricing, but you lose control over cooling configuration and power density.

**Advantages:** - **Speed to market**: Deploy in weeks rather than years. Pre-built infrastructure eliminates construction risk. - **Capital efficiency**: Operating expense model rather than capital expenditure. No upfront construction costs. - **Interconnection**: Carrier-neutral facilities provide access to dozens or hundreds of network providers, enabling cost-effective connectivity. - **Geographic flexibility**: Lease space in multiple markets without building in each one. - **Expertise included**: The provider handles physical security, power management, cooling, and facility maintenance.

**Limitations:** - **Limited customisation**: Standard power densities and cooling configurations may not match specialised workloads. - **Shared resources**: Power and cooling capacity is shared among tenants. Peak demand from one tenant can affect others. - **Cost at scale**: Per-kW pricing makes colocation expensive at very large scale (100+ MW). - **Dependency**: Reliance on the provider for uptime, maintenance, and capacity planning.

**Typical customers:** Enterprise IT, SaaS companies, financial services, content delivery, managed service providers.

**Cost range:** $100-$200/kW/month depending on market, redundancy, and contract term.

Hyperscale: Purpose-Built Campuses

Hyperscale data centres are single-tenant facilities designed for one organisation's specific requirements. They are typically built at massive scale (50-500+ MW) with custom power and cooling architectures.

**Advantages:** - **Full control**: Every aspect of the facility is designed to the operator's specifications. - **Cost efficiency at scale**: Custom design and bulk procurement dramatically reduce per-kW costs at 100+ MW scale. - **Optimised design**: Custom cooling (including liquid cooling for AI workloads), custom power distribution, and purpose-built network architectures. - **Capacity planning**: No competition for shared resources. Predictable power and cooling availability.

**Limitations:** - **Massive capital requirement**: Construction costs of $8-18M/MW mean a 100MW facility costs $800M-$1.8B. - **Long timelines**: 18-36 months from ground-breaking to operational, plus 6-24 months for permitting and design. - **Operational burden**: The operator is responsible for all facility management, maintenance, and staffing. - **Location risk**: A single-site investment creates geographic concentration risk.

**Typical customers:** Cloud providers (AWS, Azure, Google), social media platforms, AI training companies, large enterprises with 50+ MW requirements.

**Cost range:** $8-18M/MW construction cost; operating costs of $30-60/kW/month (self-operated).

The Hybrid Model

Most large enterprises now use both models simultaneously. Colocation for interconnection-heavy workloads and geographic distribution, hyperscale for bulk compute and AI training.

The $40 billion Aligned Data Centers acquisition by a consortium including Microsoft, NVIDIA, and xAI exemplifies this hybrid approach - hyperscale tenants acquiring purpose-built capacity while maintaining colocation presence for edge and interconnection.

Decision Framework

How to Evaluate

Start with the Score Tool to assess any property for data centre viability. For detailed colocation vs hyperscale analysis tailored to your requirements, contact our advisory team.