Colocation sounds simple until you start comparing providers, contracts, rack densities, cloud options, network routes and cooling models. Then "we just need somewhere to put the servers" quickly becomes less of a plan and more of a procurement trap with a spreadsheet.
At its simplest, colocation means housing your own servers, storage and networking equipment inside a professionally managed data centre. You still own and control the hardware. The provider supplies the building, power, cooling, physical security, connectivity options and operational environment around it.
For many organisations, colocation sits between on-premises infrastructure and public cloud. You keep the control and predictability of owning hardware, but you avoid the cost and complexity of running a data centre yourself.
This guide explains what colocation means, how it works, what it costs in the UK, how it compares with cloud and on-premises infrastructure, and how to choose the right provider in 2026.
What Does Colocation Mean?
Colocation, often shortened to "colo", is the practice of placing your own IT hardware inside a third-party data centre.
You rent space, power and supporting infrastructure, but you retain ownership of the servers. That is the main difference between colocation and cloud. In cloud, you rent virtual resources from someone else's platform. In colocation, you bring your own physical equipment and run it in a controlled facility.
Think of it as commercial property for IT infrastructure. You bring the equipment. The provider supplies the building, electricity, backup power, cooling, security, connectivity options and physical environment.
Colocation can suit a business with a few units of equipment, a full rack, multiple racks, a private cage, or a high-density compute deployment. The right setup depends on workload, power draw, compliance needs, connectivity and growth plans.
It is useful for organisations that want more control than public cloud, better resilience than an office server room, predictable infrastructure planning, physical ownership of hardware, secure rack space, carrier access and room to scale without building their own data centre.
"First-time buyers often underestimate the importance of proper rack configuration and the role of PDUs in managing power distribution. It's crucial to understand that optimal setup not only enhances efficiency but also mitigates risks related to power outages and equipment failures"
Gavin Lester, Carbon-Z
Colocation Vs Cloud Vs On-Premises At A Glance
Before going deeper, it helps to compare the three main infrastructure models.
| Factor | Colocation | Public Cloud | On-Premises |
|---|---|---|---|
| Who owns the hardware? | You own the servers | Cloud provider owns the platform | You own the servers |
| Who manages the facility? | Colocation provider | Cloud provider | You |
| Cost model | Rack, power, connectivity and support charges | Usage-based virtual resources | Capital expenditure plus facilities cost |
| Scalability | Scales by space, power and hardware planning | Scales quickly through virtual resources | Limited by your building and facilities |
| Control | High hardware and network control | Lower physical control, high platform flexibility | Full control, full responsibility |
| Security responsibility | Shared between you and the provider | Shared between you and cloud provider | Mostly yours |
| Best fit | Owned hardware, predictable workloads, compliance, and hybrid infrastructure | Variable workloads, rapid scaling, managed platforms | Legacy systems, local control, specialist internal needs |
There is no single winner. Public cloud can be excellent for flexible workloads and rapid development. On-premises infrastructure can still make sense for specialist local systems. Colocation is often the middle ground when you want control, resilience and better facilities without running the building yourself.
For a deeper comparison of hosting models, our guide to colocation vs cloud looks at where different workloads belong.
How Does Colocation Work?
Colocation usually follows a clear process. The details vary between providers, but most projects move through the same core stages.
The first step is understanding what you need to house. This includes the number of servers, switches and storage arrays, rack units, power draw, growth plans, cooling requirements, connectivity, access needs and support expectations.
For standard enterprise hardware, this may be straightforward. For GPU, AI or HPC workloads, the review needs to go deeper because rack density and cooling can change the design completely.
Once the requirements are understood, the provider confirms the commercial model. This may be based on rack space, power allocation, committed kilowatts, bandwidth, cross-connects, private cage space, smart hands, or a mixture of these.
You should also agree on service level agreements, installation responsibilities, monitoring responsibilities, access process, contract length, exit process and additional charges. This is where vague assumptions need to leave the room quietly.
Your equipment is delivered to the data centre and installed in the agreed rack, cage, suite or hall. Depending on the agreement, installation may be carried out by your own engineers, the provider's engineers, or both.
The work normally includes racking hardware, connecting power, installing network links, labelling equipment, testing boot sequences and checking remote access.
Connectivity is one of the biggest reasons organisations choose colocation. A good facility gives you access to carriers, private network options, internet services and cloud connectivity.
Cross-connects are physical or logical connections between your rack and another network provider, carrier, cloud platform or customer environment inside the facility. They are easy to overlook in early cost planning, but they can become a major part of the design.
Before a production cutover, test power, connectivity, monitoring, access, backup processes and failover routes. For a migration, the cutover plan should include rollback steps. Traditional wisdom applies: test before you trust.
After go-live, the provider maintains the facility infrastructure. You continue to manage your hardware, operating systems, applications and data, unless you have agreed managed services.
Smart hands support may include remote reboots, cable changes, drive swaps, hardware replacement, visual checks, access escorting, delivery handling and rack audits.
What Does A Colocation Data Centre Provide?
A colocation facility provides the environment your hardware needs to run reliably.
This includes utility power, UPS systems, backup generation, rack power distribution and monitoring. The exact resilience model should be made clear before you sign. For high-density deployments, power often becomes the central pricing and design factor.
Servers generate heat under load. Standard environments often use air cooling with hot and cold aisle containment. If your workload sits within standard rack density and cooling requirements, air-cooled colocation may be the right fit.
High-density workloads may need liquid cooling, direct-to-chip cooling or immersion cooling. If you are planning GPU clusters, AI training hardware or other dense compute, the cooling conversation should happen much earlier.
Colocation security usually includes controlled access, CCTV, visitor logging, locked racks, private cages or private suites, and site-level procedures. This matters because the equipment may support critical applications, regulated data or customer services.
A colocation site should provide practical connectivity options, not just a single internet feed. This may include carrier access, dark fibre, cloud connectivity, private networks, cross-connects and resilient routing.
Good support is also about practical access, clear escalation routes, documented procedures and knowing who is allowed to touch what.
Types Of Colocation Services
Colocation is not one fixed product. Buyers usually choose from several service levels.
Rack units suit smaller deployments where you only need a small amount of space. A full rack gives you a dedicated cabinet with defined power and connectivity. These options are common for network equipment, backup systems, SaaS platforms, managed service providers and predictable enterprise infrastructure.
Private cages and halls give you a physically separated space inside the facility. They are useful when you need tighter access control, larger deployments, customer segregation or a clearer compliance boundary.
High-density colocation is designed for workloads that draw more power and produce more heat than standard racks. These include GPU clusters, AI training, rendering, simulation and HPC workloads.
For these requirements, HPC colocation may be more suitable than conventional rack space because the facility needs to support higher density, different cooling routes and more complex service access.
Liquid cooling becomes important when air cooling is no longer practical for the workload. Direct-to-chip cooling removes heat at the component level using cold plates and a coolant loop, while immersion cooling places hardware in dielectric fluid to remove heat from the system environment.
The key point is not to pick the shiniest cooling option. It is to match the cooling route to the hardware, rack density and maintenance model.
Why Organisations Choose Colocation
Organisations usually move into colocation for a practical reason. It is rarely because someone woke up with a poetic love of racks.
Common triggers include an office server room reaching its limit, rising power and cooling costs, unpredictable cloud spend, compliance needs, limited on-site connectivity, resilience concerns, hardware refresh plans, disaster recovery requirements and fewer facilities responsibilities for internal IT teams.
Colocation works well when the organisation wants to keep ownership of hardware but no longer wants to own the building-level problem.
"The most common trigger for organisations moving from on-premises or cloud solutions to colocation is the need for enhanced control over their infrastructure and greater reliability. As businesses scale, they often realise that colocation offers a balance of performance, security, and cost-effectiveness that on-premises solutions struggle to match."
Gavin Lester, Carbon-Z
Who Uses Colocation? Common Use Cases
Colocation is used across many sectors because it solves a broad infrastructure problem: how to run owned hardware in a better environment.
Smaller organisations may start with a quarter rack, half rack or full rack. The attraction is better resilience, better connectivity and better physical security than an office comms room.
Financial services organisations often need predictable latency, strong controls, resilient infrastructure and clear physical location of systems. Colocation can support those needs while allowing firms to retain control over hardware and network design.
Healthcare organisations and suppliers may need secure, resilient environments for systems that support clinical operations, patient services, imaging, research or administrative platforms. Data protection and continuity planning are usually central to the decision.
Research workloads can involve storage-heavy systems, HPC clusters or specialist equipment. Colocation can help when internal facilities cannot support the required power, cooling or physical space.
Online platforms often need reliable performance, predictable infrastructure cost and resilient connectivity. Public sector and regulated organisations may need resilience, physical security, auditability and careful supplier governance. Colocation can support hybrid strategies where some systems remain on owned hardware while other workloads move to cloud.
Colocation cost depends on space, power, cooling, connectivity, support and contract terms. There is no single price that applies to every deployment because a lightly loaded rack and a high-density GPU rack place very different demands on the facility.
That is why buyers should be careful when comparing headline rack prices. A cheaper quote may not include enough power, enough cooling headroom, the right connectivity options or the support needed during installation and operation.
The main cost drivers are rack space, committed power, actual power consumption, cooling density, connectivity, cross-connects, private cages, smart hands support, installation support, contract length, access requirements and future expansion capacity.
| Pricing Model | How It Works | Best Fit |
|---|---|---|
| Per unit | You pay for a small amount of rack space | Small deployments, network equipment, backup systems or test environments |
| Per rack | You pay for a dedicated cabinet with agreed power and connectivity | Standard enterprise infrastructure |
| Per kilowatt | Pricing is based on committed or used power | High-density workloads where energy and cooling are the main cost drivers |
| Hybrid | Space, power, connectivity and support are priced together | Larger, mixed or more complex deployments |
Power is often one of the biggest variables. A modest rack is a different proposition from a high-density rack that needs advanced cooling and careful electrical planning. Connectivity can also change the final cost, so cross-connects, carrier choice, cloud connectivity and resilient network routes should all be reviewed before comparing providers.
A useful quote should explain rack allocation, power model, cooling assumptions, bandwidth, cross-connect charges, installation fees, smart hands rates, access policy, overage charges, expansion options and exit terms.
If your workload is high-density, GPU-heavy or likely to grow, a power assessment should come before final pricing. Otherwise, you may compare quotes that look similar on paper but are built around very different power and cooling assumptions.
Colocation Security, Compliance And Data Protection
Colocation does not remove your compliance responsibilities. It changes where some of the physical and operational controls sit.
You are still responsible for your data, systems, applications, encryption, access controls and governance. The provider is responsible for the facility controls they provide, such as physical security, power resilience, cooling and site access.
The ICO's guidance on controllers and processors is useful when reviewing data protection roles. For many colocation arrangements, you should be clear about who controls the data, who processes data, and what access the provider's staff may have to equipment or systems.
You should also check ISO or equivalent certifications, access control, visitor management, CCTV retention, incident reporting, data protection terms, subcontractor controls, physical segregation, secure disposal and audit rights.
The UK Government's data centres factsheet says data centres were designated as Critical National Infrastructure in 2024 and describes them as critical to economic activity and public services.
How To Choose A UK Colocation Provider
A good colocation provider should make the technical reality clear, not hide it behind comforting words like "enterprise-grade". That phrase has done a lot of miles.
| Factor | What To Ask |
|---|---|
| Power | What power density can the site support now and at refresh? |
| Cooling | Is the cooling model suitable for your actual rack load? |
| Connectivity | Which carriers, cloud routes and cross-connect options are available? |
| Security | What physical controls, access logs and certifications are in place? |
| Support | What smart hands tasks are included, charged or excluded? |
| Scalability | Can you add space, power or density without moving site? |
| Resilience | What redundancy model applies to power and cooling? |
| Contract | How are exit, expansion, overage and service changes handled? |
| Location | Does the site match latency, access and power needs? |
| Transparency | Are all recurring and one-off charges clearly explained? |
Location still matters, but not every workload needs the same site profile. Some buyers may prioritise carrier access and latency through London colocation , while others may compare options such as Swindon colocation , where power flexibility and room to scale are part of the decision.
"When evaluating colocation providers buyers should look beyond headline rack prices and consider factors such as service reliability, customer support, scalability options, and the overall security of the facility. The true value lies in the ability to meet long-term needs and ensure seamless operations."
Gavin Lester, Carbon-Z
How To Migrate To Colocation
Migration is where good planning pays for itself. The goal is not simply to move hardware. It is to move risk out of the process.
List every server, switch, storage device, firewall, cable, application dependency and support owner. Check which systems are physical, virtual, critical, legacy or due for refresh.
Measure current and expected power draw, rack units, storage growth, network throughput and cooling needs. Do not rely only on nameplate ratings or old asset records.
Decide whether you need shared rack space, a full rack, multiple racks, a cage, a private suite or high-density capacity. Confirm power, cooling, network and access requirements.
Connectivity often decides whether a migration succeeds smoothly. Plan internet, private circuits, cloud links, VPNs, firewalls, cross-connects and failover routes before the move.
Choose a cutover window based on business impact. Plan for backups, rollback, staff availability, supplier access, testing and stakeholder communication.
Rack the hardware, connect power, complete cabling, check labels, test remote access, validate monitoring and confirm failover. Do not leave documentation until after go-live, when everyone is tired and pretending not to be.
The first 90 days should be used to check power draw, cooling stability, alerting, support tickets, access requests and growth assumptions. This is where you find out whether the design matches reality.
UK Colocation Market And Trends In 2026
The UK colocation market is being shaped by three linked pressures: demand, power and density.
Grand View Research estimates that the UK data centre colocation market generated around USD 4.01bn in revenue in 2024 and forecasts it to reach around USD 11.25bn by 2030. Converted into sterling, that places the 2024 market at roughly £3.2bn, depending on the exchange rate used. The exact figure varies by source and methodology, but the direction of travel is clear: UK colocation demand is growing quickly.
CBRE's UK data centres outlook for 2026 forecasts 180MW of new supply in the London area in 2026, following a record 193MW of new supply in 2025. This points to continued expansion, but not necessarily easy access to the right capacity for every workload.
Government policy is also shifting. The UK Government's AI Growth Zones programme is designed to unlock investment in AI-enabled data centres and supporting infrastructure by improving access to power and providing planning support. Separate government guidance says timely grid connections are one of the biggest blockers for AI Growth Zones, which is why power access now sits at the centre of infrastructure planning.
For buyers, this creates a more complicated market. Capacity is growing, but not evenly. Power availability can matter more than traditional location prestige. AI and GPU workloads are changing rack density expectations. Liquid cooling is moving from niche to practical requirement, and regional sites may become more attractive where power and cooling are better aligned.
In 2026, colocation decisions should not be made on rack price alone. The better question is whether the provider can support the workload you are bringing today and the hardware you are likely to bring next.
How Carbon-Z Supports Colocation Buyers
Carbon-Z supports a range of colocation requirements, from standard enterprise workloads through to high-density compute.
The important point is that different workloads should not be forced into the same infrastructure model. A standard storage rack and a GPU training rack may both sit in a data centre, but they place very different demands on power, cooling, cabling and service access.
If you are unsure where your workload fits, the sensible starting point is a technical review of power, cooling, connectivity and growth plans. That helps avoid choosing a contract that suits today's rack count but fails at the next refresh.
How is colocation priced in the UK? +
UK colocation is usually priced by rack units, full racks, committed power, cross-connects, support services and contract terms. Higher-density workloads often move towards power-based pricing because energy and cooling become the main design drivers.
Is colocation more secure than cloud? +
It depends on what you mean by secure. Colocation gives you more physical control over owned hardware, while cloud gives you access to managed platform security. In both cases, security is shared between you and the provider.
What SLAs should I expect from a colocation provider? +
You should expect clear SLAs for power availability, physical access, support response, environmental conditions and incident communication. Always check what is included, excluded and how credits or remedies are handled.
Do I need a lot of hardware to use colocation? +
No. Some organisations colocate a few units of equipment, while others use full racks, cages, private halls or high-density pods.
Who owns the servers in a colocation facility? +
You normally own the servers, storage and networking equipment. The provider owns and operates the data centre infrastructure around them.
How does UK data protection apply to colocation? +
You remain responsible for protecting personal data processed by your systems. Review access controls, provider contracts, data processing roles, physical security and incident processes in line with UK GDPR.
How long does a typical colocation contract run? +
Contract terms vary. Smaller deployments may use shorter agreements, while larger or high-density deployments often require longer terms because power, space and cooling capacity must be reserved.