Private cloud vs public cloud sits near the top of every serious infrastructure conversation in Australian organisations right now. The question feels deceptively simple: run workloads on shared, hyperscaler infrastructure or on dedicated hardware you control? In practice, the answer hinges on data sensitivity, compliance obligations, cost structure, and the operational maturity of your team. Getting it wrong means either overspending on privacy you did not need or exposing data that should never have left your control.
What private cloud actually means
Private cloud is infrastructure provisioned for a single organisation, either on-premises or hosted by a third party in a dedicated environment. The key characteristic is exclusivity: compute, storage, and networking resources are not shared with other tenants. That isolation is what organisations pay a premium for, and it is what drives most of the legitimate use cases.
On-premises private cloud typically means VMware-based or OpenStack environments running in your own data centre or a co-location facility. Hosted private cloud puts that dedicated hardware under a managed service provider's roof, often with physical security and connectivity guarantees that a standard colocation arrangement does not provide. Azure Local represents a newer pattern: cloud-native management tooling applied to dedicated on-premises hardware, giving organisations familiar hyperscaler interfaces without the shared tenancy model.
What public cloud actually delivers
Public cloud means AWS, Microsoft Azure, and Google Cloud Platform running shared, multi-tenant infrastructure at global scale. You consume compute and storage on demand, paying for what you use rather than provisioning capacity upfront. The economics are compelling for variable workloads: you scale out during peak demand and shrink back when it passes, without carrying idle infrastructure on your balance sheet.
All three major hyperscalers operate Australian regions. AWS has Sydney and Melbourne. Azure runs East Australia (Sydney), Southeast Australia (Melbourne), and Australia Central (Canberra). GCP operates Sydney and Melbourne. For most Australian organisations, public cloud delivers low enough latency for the majority of workloads, and the hyperscalers have invested heavily in local compliance certifications to address data residency concerns. Our guide to Australian data residency covers what those certifications actually guarantee and where the gaps remain.
The cases where private cloud still wins
Several scenarios genuinely favour private cloud, and conflating them with general preference or vague security anxiety leads to poor decisions.
- Regulated data that must not leave a specific boundary. Healthcare workloads covered by the My Health Records Act, certain Defence contracts under ITAR or the Australian Government's PROTECTED classification, and some financial sector obligations under APRA CPS 234 create hard requirements around where data can reside and who can access the underlying infrastructure. Public cloud's shared tenancy model, even with logical isolation, does not satisfy some of these requirements.
- Predictable, high-utilisation workloads. Public cloud's on-demand pricing model favours variable workloads. If your databases run at 80 to 90 percent utilisation around the clock, owning the hardware often costs less over a three to five year period than paying hyperscaler rates. Reserved instances and savings plans narrow the gap, but rarely close it for continuously saturated workloads.
- Latency-sensitive applications with strict SLAs. Industrial control systems, real-time trading infrastructure, and some edge manufacturing environments require sub-millisecond latency that even a local cloud region cannot reliably guarantee once network hops are factored in. Private infrastructure co-located with the application's physical dependencies solves this cleanly.
- Sovereignty requirements. Federal and state government agencies operating at certain classification levels face explicit mandates about where data is processed and stored, and about what foreign entities can access that infrastructure under law. This is the driver behind Australia's sovereign cloud programs run by providers including AWS Secret Region, AUCloud, and Macquarie Government.
The cases where public cloud wins
For the majority of Australian businesses and a large share of government workloads, public cloud is the right default. Here is why.
- Development and test environments. Spinning up and tearing down environments on demand is where public cloud's elasticity shines most clearly. The cost of provisioning private infrastructure for non-production workloads is almost never justified.
- Applications with unpredictable or seasonal demand. Retailers handling Christmas traffic spikes, media streaming services, and SaaS platforms experiencing growth cannot easily pre-provision private infrastructure at the right size. Public cloud absorbs that variability without waste.
- Speed to market and managed services. Hyperscalers offer managed databases, AI/ML platforms, container orchestration, and hundreds of other services that would take years and significant headcount to build privately. For most organisations, the opportunity cost of replicating those services on-premises is prohibitive.
- Disaster recovery and backup. Using public cloud as a recovery target for on-premises or private cloud workloads is one of the most cost-effective DR patterns available. You pay for storage rather than idle standby compute, and hyperscaler availability zones give you geographic separation without maintaining a second data centre.
The hybrid reality most Australian organisations are living
The private vs public framing is somewhat artificial because most mid-to-large Australian organisations operate both. The practical question is not which one to choose but how to divide workloads between them and how to connect the two environments reliably. A common pattern places sensitive production databases and regulated workloads on private infrastructure while running application tiers, analytics, and customer-facing services on public cloud.
Network connectivity between private and public environments is where many teams underestimate complexity. Direct Connect (AWS), ExpressRoute (Azure), and Cloud Interconnect (GCP) all provide dedicated private connectivity from Australian data centres to hyperscaler regions, but the configuration, BGP management, and redundancy requirements are non-trivial. Teams frequently discover this late. For a detailed look at where Australian teams go wrong with cloud networking design, see our coverage of common cloud networking mistakes in Australia.
Cost: the number that changes the conversation
Private cloud carries high upfront capital expenditure and ongoing operational costs: hardware refreshes every three to five years, power, cooling, facilities, and the engineering team to manage it. Public cloud shifts that to operational expenditure, which looks attractive until workloads are large and stable enough that hyperscaler rates exceed the amortised cost of ownership.
A useful rule of thumb is to model total cost of ownership over five years for any workload you are deciding on. Include hardware depreciation, rack space, power, staffing, and the cost of maintaining currency with security patches and hardware support contracts on the private side. Compare that against committed use discounts, reserved instances, and realistic utilisation assumptions on the public side. The crossover point is more often around 60 to 70 percent sustained utilisation than most teams expect.
Making the call: a practical framework
Running through four questions in order covers most real-world decisions.
- Is there a hard regulatory or classification requirement? If yes, that requirement usually prescribes the answer. Confirm it with your legal and compliance team before assuming public cloud cannot satisfy it, because the certified regions and sovereign cloud programs have expanded significantly.
- What is the utilisation profile? Continuous high utilisation favours private. Variable or growth-stage workloads favour public.
- What is the team's operational capacity? Private cloud demands skilled engineers to manage infrastructure, security patching, and hardware lifecycle. If that capacity does not exist in-house, the true cost of private cloud includes building or buying it.
- What is the acceptable blast radius of a provider failure? Public cloud outages affect many tenants simultaneously but are resolved by large engineering teams quickly. Private cloud failures are your problem entirely. Both models require a tested recovery plan.
Neither private nor public cloud is inherently more secure, more expensive, or more reliable. Those properties depend on how each environment is designed, operated, and governed. The choice is a function of workload characteristics, team capability, and regulatory context, not ideology.