Own datacenter, hosting, or cloud: count speed of change, not servers
How to move the infrastructure conversation from hardware costs to speed of response, SLA, and clear ownership of responsibility.
When a company starts talking about infrastructure, the conversation almost always starts with hardware. How much does a server cost? How many racks do we need? Is the bandwidth sufficient? These are concrete questions, easy to put in a spreadsheet.
The problem is that the most expensive decisions get made at exactly this level - and that is why they cost so much later. Hardware is easy to price. But the business is not paying for hardware.
What an infrastructure choice actually decides
Behind the choice between your own datacenter, hosting, and cloud are three separate questions that usually get collapsed into one.
The first is total cost of ownership. This includes not just equipment, but rent, cooling, power, redundancy, a refresh cycle every three or four years, and the salaries of the people maintaining all of it.
The second is SLA and accountability. Who is responsible for availability, and what happens when something goes down at 3 a.m. on a Sunday? In your own datacenter, the responsibility is entirely yours. With hosting, it is partly delegated. In the cloud, it is delegated differently, with explicitly written terms.
The third is speed of change. How long does it take to add capacity, launch a new service, test a hypothesis, or roll back a failed deployment? This is the question that gets skipped most often.
Why speed matters more than it seems
An owned datacenter is fixed infrastructure. You buy for peak load and on average use 20 to 30 percent of what you bought. Time to bring a new server online is days to weeks. That is acceptable when the business changes slowly and predictably.
When the business changes quickly, that constraint becomes a strategic brake. Every delay in launching a new service, every queue for resource allocation, is lost time - not just lost money.
Hosting removes some of the operational burden but usually gives no elasticity. You rent fixed capacity, and scaling it up quickly is not straightforward.
Cloud changes the economics precisely through elasticity. You pay for what you use. Resources appear in minutes, not weeks. That is not a sales pitch - it is a change in operating model.
Where cloud is not the answer
Cloud is not a universal solution. There are situations where it loses or does not apply.
Regulatory requirements. In some industries, data must physically reside in a specific location or under specific control. There may simply be no choice.
Stable, predictable load. If your workload is constant and does not change year over year, the economics of a well-tuned physical setup can be better than a cloud with variable monthly bills.
Specialised hardware. Some workloads require equipment that is either unavailable in the cloud or priced disproportionately.
Missing competencies. Cloud does not simplify the operating model by itself. It moves the complexity somewhere else. Without people who know how to manage it, the costs often exceed what was expected.
How responsibility is distributed
This is probably the most underestimated part of the decision.
In your own datacenter, you own everything - and bear full responsibility for everything. You decide when to replace a disk that has started throwing errors. You decide how to build redundancy. You are responsible for physical security.
In a public cloud, the provider takes ownership of the physical layer and part of the platform - but does not take responsibility for what you build on that platform. Misconfigured access controls, open ports, unencrypted storage - that is your zone, not theirs.
Understanding that boundary matters more than picking a provider.
Questions worth asking before deciding
Before choosing an infrastructure model, I recommend answering a few concrete questions:
- How often did your infrastructure requirements change in the past year?
- How long does it currently take to bring a new service or environment online?
- Who in your organisation can manage the chosen platform, and what happens when that person leaves?
- What are your actual RTO and RPO targets - and does your current infrastructure meet them?
- What portion of your current spending is infrastructure itself, and what portion is the IT staff wrapped around it?
The answers to these questions produce a picture that does not fit in a price comparison table. But it is exactly that picture which determines which solution is right for a specific business at a specific moment.