Cloud migration has been the dominant IT conversation for most of the last decade. By now most Australian SMBs have moved something to the cloud — email is usually the first — but many still run a mix of cloud and on-premises, sometimes intentionally, sometimes because it never got around to being cleaned up.

This post is an honest framework for deciding what belongs in the cloud, what’s fine where it is, and how to plan a migration if you decide to do one.

The question isn’t “should we move to cloud”

The question is “which workloads belong where, and why?”

Cloud is not universally cheaper, faster, or more reliable than on-premises. For some workloads it is; for others, it isn’t. The defining question is what each workload actually needs and what you currently have.

What cloud genuinely does better

For most SMBs, several capabilities are categorically better in the cloud:

Email and collaboration. Microsoft 365 replaced on-premises Exchange for almost every reason that matters. Better uptime, better spam filtering, global access, no hardware refresh cycle, continuous feature updates. If you’re still running on-prem Exchange in 2026, the migration should already be underway.

File sharing and collaboration tools. SharePoint and OneDrive replace traditional file servers with better access, better versioning, better collaboration, and better backup. Classic Windows file servers are in long-term decline.

Backup off-site copies. Cloud backup is a great offsite target. Cheaper and more reliable than tape rotation or dedicated colo space.

Development and test environments. Elastic scaling — spin up for a project, tear down afterwards — is exactly what cloud does well.

Disaster recovery. Warm or cold DR in cloud costs a fraction of a second physical site. For most SMBs, cloud DR is the only practical way to actually have DR.

Public-facing websites and APIs. Static sites on CDNs, serverless APIs, managed databases — dramatically simpler than running this on your own hardware.

What cloud does less clearly better

Some workloads have genuine pros and cons:

Line-of-business apps. If the vendor has a SaaS version, it’s often the right move. If they don’t, running their on-prem version in a cloud VM is usually more expensive than running it on-premises — you pay for elasticity you don’t use.

Databases. Managed cloud databases (Azure SQL, RDS) are excellent but not cheap. For small databases with predictable load, a well-managed on-prem or VM-hosted database can be cheaper.

Large file storage (video, imaging, CAD). Cloud storage is cheap for data at rest; egress (reading it back) isn’t. If you edit 200GB of video daily, you probably want that close to the editors.

Low-latency workstation storage. Network latency to the cloud makes it a poor fit for applications that stream lots of data to/from a workstation constantly. Revit, Premiere, large CAD workflows often still work better locally.

What cloud is typically worse for

Honestly:

Highly predictable, steady workloads you already own the hardware for. If you have a paid-for server doing a steady job and it works, moving it to cloud may not save money for years. The migration cost is real.

Legacy applications that don’t cope with network latency. Some old client-server apps (early 2000s-era line-of-business tools) fall apart when the database is behind a few milliseconds of network hop.

Very high-throughput workloads where egress cost dominates. Cloud is cheap until you move data out of it.

Regulated environments with specific on-prem requirements. Some healthcare, defence, and government environments have rules that make pure-cloud difficult. Hybrid is common here.

The practical migration framework

For most Australian SMBs, a sensible migration roadmap looks like this:

Phase 1 (if not already done): Move email and collaboration

Email to Microsoft 365 or Google Workspace. OneDrive and SharePoint to replace file servers (or at least start the migration). Teams or Meet for collaboration.

This is the easiest win. It usually happens over a few months and delivers immediate improvements in reliability and staff experience.

Phase 2: Cloud backup

Even if primary workloads stay on-prem, backup belongs in cloud (or at least has a cloud offsite copy). Veeam, Backblaze B2, Wasabi, Azure Blob — all work.

Phase 3: SaaS the easy ones

Anything you can replace with a SaaS equivalent should be evaluated. Accounting (Xero, MYOB), CRM (HubSpot, Salesforce, Dynamics), HR (Employment Hero, BambooHR), ticketing (Jira, Zendesk). Most SMBs end up with 20-40 SaaS subscriptions and no more self-hosted equivalents.

Phase 4: Lift-and-shift the servers that make sense

For workloads that need to stay virtualised but don’t have SaaS alternatives, lift-and-shift to Azure or AWS is often the right move when the existing hardware is due for refresh. Don’t migrate just to migrate; time it with hardware lifecycle.

Good candidates: remote access gateways, domain controllers, print servers, small databases, specific line-of-business apps.

Phase 5: Re-architect the ones worth re-architecting

A few workloads benefit from genuine cloud-native redesign — replacing a VM with serverless, moving a database to a managed service, using object storage instead of file storage. These are project-level investments justified by specific business case.

What stays on-prem or hybrid

Not everything needs to move:

The manufacturing line’s control PCs stay near the line
The video editing suite’s storage stays near the editors
The POS system in the shop stays in the shop
Sometimes the domain controller stays local (for fast logon during internet outages)

A hybrid model, done deliberately, is fine and often the right answer.

Common cloud migration mistakes

Lift-and-shift everything without rationalisation. Moving a bloated on-prem environment straight to cloud just means paying more for the same bloat. Migration is an opportunity to retire systems you stopped using three years ago.

Underestimating egress and data transfer. Pay attention to the cost model. Compute, storage, and egress are separately priced. Big egress surprises are a common cause of “cloud is more expensive than expected.”

Skipping the networking thinking. Hybrid connectivity (VPN, ExpressRoute, Direct Connect) and identity (Entra ID Connect, federation) are where most of the hard work lives. Plan these first, not last.

Migrating without a FinOps plan. Cloud spend grows fast without governance. Set up tagging, budgets, and regular cost reviews from day one.

Assuming it’s “set and forget.” Cloud environments need ongoing management — patching, security, cost optimisation, identity governance. You’re swapping one operational burden for another, not eliminating it.

The bottom line

The goal isn’t to “go to cloud.” The goal is for every workload to be in the right place — based on its real requirements, its real cost, and the real capability of your team.

For most Australian SMBs, that ends up being ~80% cloud (M365, SaaS, backup, DR, some migrated servers) and ~20% on-prem or edge (specific local workloads that genuinely benefit from being close to users or equipment).

If you’d like to work through what that mix should look like for your business — including a workload-by-workload assessment of what to move, what to keep, and what to retire — get in touch. We run cloud readiness assessments as a fixed-fee engagement with a clear roadmap and cost model at the end.