1.4 Public, private, hybrid and multi-cloud
Describes the cloud landscape as of August 2026
What this is and why it exists
Where the cloud runs is a separate question from what it serves you. Public, private, hybrid and multi-cloud are answers to "whose building, whose rules" — and most requirements force none of them, which is itself worth knowing before someone's slide deck forces one for you.
The vocabulary
- Public cloud — the provider's shared infrastructure, many customers pooled, pay as you go.
- Private cloud — cloud-style self-service built on infrastructure dedicated to one organisation.
- On-premises — your hardware in your building; private cloud's blunter cousin.
- Community cloud — shared infrastructure for several organisations with one shared concern, such as a group of banks or hospitals.
- Hybrid cloud — one workload estate spanning private and public, connected.
- Multi-cloud — using more than one public provider on purpose.
The mental model
Public cloud economics rest on one assumption: pooling strangers smooths demand, so shared hardware runs hotter and cheaper than dedicated hardware. Every other model is a deliberate retreat from that assumption, paid for in money or effort, and each retreat needs a reason.
Good reasons exist. Private and on-premises earn their cost when law requires data in your custody, when latency requires the machine beside the factory line, or when steady heavy load makes owning cheaper. Community cloud pools organisations that share a regulator — real, but rare in practice, because getting rivals to share infrastructure is politics before it is engineering. Hybrid buys a migration path or keeps one anchored system at home — and its hidden line item is connectivity: the link between halves has a cost, a latency and a failure mode, and it is now part of every incident. Multi-cloud buys survival of one provider's failure and negotiating leverage, and its honest price is portability: you build to the lowest common feature set, run two skill sets, and test two of everything.
When no requirement forces a model, the default that needs no justification is public.
What you should now be able to explain or do
For a stated requirement — a data-residency law, a factory-floor latency budget, a board's fear of lock-in — name the deployment model it forces, or say honestly that it forces none. Name the hidden cost of hybrid and the honest cost of multi-cloud.
Check yourself
What assumption makes public cloud cheap?
Pooling — many customers' peaks and troughs smooth each other out, so shared hardware runs at higher utilisation than any single customer could manage.
Name two requirements that genuinely force private or on-premises.
Law that requires data in your own custody, and latency that requires the machine physically beside the equipment. (Steady heavy load is a third, on economics.)
What is the hidden cost of hybrid?
The connection between the halves — its money, its latency and its failure modes become part of every workload that spans it.
What does multi-cloud actually cost?
Portability: lowest-common-denominator designs, doubled skills and tooling, and twice the testing — paid continuously for insurance you hope never to use.
Go deeper
We haven't checked most of these for screen reader use yet.
Back to Public, private, hybrid and multi-cloud: work through the checklist