What Cloud Actually Is

Defining characteristics

Cloud computing is on-demand access to shared computing resources over a network. The widely used definition identifies five characteristics:

• On-demand self-service — provision resources without a human transaction.
• Broad network access — reachable over standard networks from varied devices.
• Resource pooling — shared physical infrastructure serving many tenants.
• Rapid elasticity — scale up and down quickly, in some services automatically.
• Measured service — metered consumption and billing.

The economic consequence is the shift from capital expenditure to operating expenditure. You stop buying capacity for the peak you might reach in three years and start paying for what you use — which is cheaper only if consumption is actually managed, and is frequently more expensive when it is not.

Service models

• Infrastructure as a Service (IaaS) — virtual machines, storage, networks. You manage the operating system upward. Most control, most responsibility.
• Platform as a Service (PaaS) — managed runtimes, databases and application services. You manage your application and data; the provider handles the platform beneath.
• Software as a Service (SaaS) — finished applications: email, office suites, CRM, HR systems. You manage configuration, users and your data.
• Serverless / Functions as a Service — code executed on demand with no server management and consumption-based billing that can fall to zero when idle.

A useful way to remember the trade-off: moving up the stack reduces what you have to operate and reduces what you can control. Most organisations run a mixture across all four.

Deployment models

• Public cloud — shared infrastructure from a major provider. Greatest scale and elasticity.
• Private cloud — dedicated infrastructure, on-premises or hosted. Used for regulatory, data residency or latency reasons; retains most of the operational burden.
• Hybrid — both, integrated. The practical reality for most established organisations.
• Multi-cloud — services from more than one provider. Reduces concentration risk and increases complexity, skill requirements and data transfer cost.
• Edge — processing near where data is produced, for latency or bandwidth reasons. Relevant in industrial, remote and field environments where connectivity is limited.

What genuinely changes

• Provisioning time falls from weeks to minutes, which changes how projects are run more than it changes what they cost.
• Capacity planning becomes cost management.
• Infrastructure becomes code, versioned and repeatable.
• The skills required shift from hardware and operating systems toward architecture, automation, security and financial management.
• Mistakes scale instantly. A misconfiguration can expose data globally in seconds — which is why governance matters more, not less.