Colocation vs On-Premises Infrastructure: Should You Upgrade or Move?
A practical framework for comparing control, resilience, cost, capacity and operational responsibility on a like-for-like basis.
If your data centre is running out of power or cooling, or the lease is coming up for renewal, you need to decide how much to invest in the current site and whether it still fits your plans. Colocation is worth considering when demand is uncertain, you need higher rack densities or running a facility is becoming difficult to justify. But keeping infrastructure on premises can still make sense when you need control of the building, have a large, stable load and can fund and operate the required resilience.
Whether the review starts with a capacity problem, an audit finding or a new workload, the first step is to work out what each option would take to meet the same requirements. Below, we work through cost, capacity, timing and operating responsibilities to help you assess where an upgrade is enough, where colocation offers more value and what to check before committing.
Define what the infrastructure needs to deliver
Define the capacity, rack density, resilience, security, connectivity and support you will need over the planning period. Record what is certain, what may change and what happens if demand is higher or lower than forecast. This prevents a short-term facilities problem from dictating a long-term infrastructure decision.
Also separate mandatory constraints from preferences. Equipment may need to remain close to operational technology, within a particular location or under an organisation’s direct physical control. Where those constraints genuinely apply, they can narrow the options before cost modelling begins.
What it takes to build or upgrade
The visible business case usually covers construction and equipment. A resilient facility also needs electrical and cooling architecture suited to the required availability, fire detection and suppression, physical access controls, environmental monitoring, maintenance arrangements and tested operating procedures. The exact design of an on-premises data centre depends on the target service level and the existing building.
These capabilities create ongoing responsibilities as well as capital costs. The organisation must assign facility accountability, maintain supplier contracts, test systems, manage changes and provide an out-of-hours response. A technically sound design will not deliver its intended resilience if maintenance and operating discipline decline over time.
What colocation covers and what you still manage
Colocation transfers responsibility for the contracted facility layer, which can include power, cooling, physical security and associated maintenance. The provider shares the fixed facility investment across customers. The customer remains responsible for its equipment, operating systems, configurations, applications and any obligations not transferred by the contract.
The contract therefore matters as much as the building. Buyers should examine service levels, maintenance provisions, access arrangements, capacity expansion, remote-hands support, metering, change processes and exit terms. Colocation reduces facility work, but it does not remove infrastructure governance.
Compare both options against the same requirements
| Decision factor | Build or upgrade on-premises data centre | Colocation |
|---|---|---|
| Control | Direct control of the building, systems and operating model | Control of customer equipment; facility control defined by contract |
| Cost profile | Capital investment plus continuing facility operating costs | Recurring contracted charges plus migration and customer-side operating costs |
| Resilience economics | The organisation funds the full design, redundancy and maintenance burden | Facility investment is shared, subject to the contracted service and design |
| Time to capacity | Assessment, design, approvals, procurement, construction and commissioning | Provider selection, due diligence, contracting, connectivity and migration |
| Skills and accountability | Facility engineering and operations must be staffed or contracted | Provider operates the facility layer; customer retains IT and vendor governance |
| Density and growth | Constrained by site power, cooling, floor loading and upgrade feasibility | Constrained by available capacity, facility design and contract terms |
| Change risk | Committed capital may be difficult to repurpose | Flexibility depends on expansion rights, minimum commitments and exit terms |
When building or upgrading can be the better choice
On-premises data centre investment can be justified when the organisation needs direct control of the site, when workloads must remain physically close to operations, or when an existing facility can meet the target standard through proportionate upgrades. It may also suit a large, stable requirement that can use the fixed investment efficiently and support a dedicated operating model.
The case should still account for lifecycle renewal, staffing, testing and future density. Ownership provides control, but it also leaves the organisation with the risk that the building or design becomes a constraint before the investment has delivered its expected value.
When colocation may be the better fit
Colocation becomes more attractive when the required resilience would be expensive to duplicate for a modest load, when capacity may change materially, or when the existing site cannot support the target rack density. It can also be useful when the deadline is incompatible with a major building project, provided migration, connectivity and contracting can be completed safely.
These advantages are conditional. Available power and cooling, expansion rights, network options and service commitments vary by facility and contract. Due diligence should test the specific offer rather than rely on portfolio-level descriptions.
Build a like-for-like business case
A fair comparison applies the same requirements to every option. Do not compare the current cost of a basic server room with the price of a facility designed to deliver a higher level of resilience. First establish the target standard, then price the work needed for each option to meet it.
- Capital works, design, approvals, equipment, commissioning and lifecycle replacement
- Facility power and cooling, including the energy used by supporting systems
- Floor space valued consistently rather than treated differently between options
- Maintenance contracts, testing, compliance activity and facility management time
- Connectivity, migration, parallel running and decommissioning
- Colocation charges, metering, support services, growth provisions and contractual indexation
- The cost and operational effect of likely growth, contraction or a technology refresh
- Material outage and concentration risks, stated as scenarios rather than false precision
Model the options over the same period and use consistent assumptions for load, energy price, growth and service level. Run sensitivity cases where forecasts are uncertain. You will not be able to put a precise value on every risk. The aim is to make assumptions visible so they can be tested, rather than letting them quietly favour one option.
Questions that test the recommendation
- What event triggered the review, and does it require a strategic move or a smaller immediate fix?
- What capacity, rack density and resilience are required now and over the planning period?
- Which location, control, security or regulatory constraints are mandatory?
- What single points of failure exist in the current environment?
- What would it cost to bring each option to the same target standard?
- Who will own facility operations, maintenance and out-of-hours response?
- How will each option respond if the forecast is materially wrong?
- What are the migration, connectivity, concentration and exit risks?
- How does finance treat capital expenditure, operating commitments and residual asset value?
- What evidence supports the provider or design claims used in the business case?
Make the operating responsibilities clear
Choose the option that meets your service requirements at a total cost and level of risk your organisation can accept, with room for likely changes in demand. Building or upgrading can make sense when control, scale and operational capability justify the investment. Colocation may offer greater value through shared facility investment, faster access to suitable capacity or reduced facility responsibility. Compare both against the same requirements and document who will be responsible for each part of the service.
Your data centre partner needs to support today’s workloads and leave room for your requirements to change. Macquarie Data Centres provides secure, resilient and scalable data centre solutions for hyperscale, enterprise, government and cloud environments.
Contact us to discuss your requirements, learn more about our team, or request a tour to see our facilities firsthand.
Frequently asked questions
Should we build our own data centre or use colocation?
Build or upgrade when direct site control, scale and operating capability justify the investment. Consider colocation when facility resilience, density or flexibility would be difficult to deliver economically on your own. Test both against the same requirements.
Is colocation cheaper than an on-premises server room?
Not automatically. Existing rooms can appear cheaper when floor space, staff time, maintenance, energy overhead and resilience gaps are omitted. Compare total costs over the same period and at the same target service level.
What does it cost to make an on-premises room resilient?
The cost depends on the building, load and required service level. Scope the electrical, cooling, fire, security, monitoring, maintenance and operating changes needed for the target standard before seeking credible estimates.
Can we upgrade an existing room for AI workloads?
Possibly, but feasibility depends on target rack density, available power, heat rejection, floor loading, water requirements where relevant and the upgrade path. Assess the workload and site together before assuming a retrofit is practical.
Does colocation mean losing control of our infrastructure?
You retain control of your equipment, configuration and applications. Responsibility for the facility layer is shared or transferred according to the contract, so access, maintenance, support and change rights must be clear.
How long does each option take?
The critical path differs. Building or upgrading can involve assessment, design, approvals, procurement, construction and commissioning. Colocation involves due diligence, contracting, connectivity and migration. Use project-specific evidence rather than generic timeframes.