Glossary
Private cloud
A private cloud is infrastructure dedicated to one organization and operated with the self-service, pooled capacity and metering of a public cloud. The operating model, not the hardware or the building, makes it a cloud; without self-service and metering it is a conventional data center.
Two tests separate it from a renamed data center
The first test is how capacity is obtained. If teams file a ticket and wait for an administrator to allocate storage, the environment fails it. In a private cloud, requests go through a portal or API and automation provisions capacity from a shared pool with no person in the path.
The second test is metering. Usage is recorded per tenant and shown back or charged back, so consumption has a visible cost to the team that causes it. Virtualization alone passes neither test: it pools hardware but still runs on requests to administrators.
What runs inside
| Component | Role | Covered under |
|---|---|---|
| Shared pool | Capacity from many servers combined and allocated on demand | Storage pooling |
| Tenants | Departments or customers with their own quotas, access rules and keys | Multi-tenant storage |
| Provisioning in code | Templates and APIs that replace tickets | Cloud automation |
| Growth by servers | Capacity added as standard nodes join the pool | Scale-out storage |
Storage is usually an S3-compatible object layer for unstructured data, alongside block and file services drawn from the same pool and metered the same way.
Capacity planning moves to the owner
A public cloud hides the supply side. In a private cloud the organization owns it, and new hardware can take weeks or months to arrive after an order. Self-service removes the human delay, so demand can now outrun deliveries. A headroom buffer bought ahead of the demand curve is the usual answer. Without it, self-service requests fail because the pool is full, and the platform team becomes the bottleneck it was built to remove.
Why do organizations build one?
Three reasons recur. Location and operator control, where data must stay in a given country or under a given operator, described under data residency. Cost of leaving, because egress fees make large, frequently read datasets expensive to hold in a public cloud. And proximity, since AI and analytics workloads run best with compute next to the data. A private cloud is also the owned half of most hybrid cloud designs, and one environment among several in a multi-cloud estate. Who hosts the building matters less than who controls the infrastructure: a dedicated environment in a third party's facility still counts.