Cloud migration is sold on the promise of removing capital expenditure. That is true, and it is incomplete — the capital cost is replaced by an operating cost that continues indefinitely, and whether the trade is favourable depends entirely on what you are moving and how well it is planned.
Here is how the cost actually breaks down, including the parts that are routinely underestimated.
The three cost buckets
Every migration has a project cost, a destination cost, and a transition cost. Businesses reliably budget the first, roughly estimate the second, and forget the third.
Project cost — the migration itself
Assessment and planning, the migration work, testing, cutover, and post-migration stabilisation. This is one-off and scales with complexity rather than headcount.
For a straightforward move — file shares and email to Microsoft 365 or Google Workspace for a business under 50 people — this is usually a few thousand to low tens of thousands. For anything involving servers, line-of-business applications, or databases with uptime requirements, it rises substantially and the variance widens.
Destination cost — what you pay ongoing
Licensing per user, infrastructure consumption if you are running virtual servers, storage, and egress charges for data leaving the platform. This replaces your hardware refresh cycle, and it is permanent.
Transition cost — running both at once
During cutover you are paying for the old environment and the new one simultaneously. For a phased migration that overlap can run months, and it is the line most frequently missing from budgets entirely.
What drives the number
- Data volume. Moving two terabytes of file data is a different exercise from moving twenty, and bandwidth becomes the constraint. Large transfers sometimes require physical shipping appliances.
- Application compatibility. Software written for on-premise deployment may not run in a cloud environment, may require a different licence, or may have no cloud version. Discovering this late is the most common cause of a stalled project.
- Email complexity. A simple mailbox move is routine. Shared mailboxes, public folders, decades of archive, and legacy calendar permissions are where the hours go.
- Identity. Migrating from on-premise Active Directory to cloud identity, or running a hybrid arrangement, is frequently the most intricate part.
- Downtime tolerance. A migration that can take a weekend costs less than one that cannot interrupt business at all, because zero-downtime approaches require parallel running and rollback capability.
- Compliance. Regulated data adds encryption requirements, audit evidence, and residency constraints.
Costs that surprise people
- Licence changes. Some software licensed perpetually on-premise moves to subscription in the cloud. The ongoing cost can exceed the infrastructure saving.
- Egress charges. Getting data in is free. Getting it out is not — which matters if you later change providers, and is worth understanding before you are dependent.
- Backup, separately. Cloud platforms provide availability, not backup. Microsoft's own shared responsibility model puts data protection on you. Third-party backup for Microsoft 365 is a distinct subscription many businesses discover after migrating.
- Bandwidth. Moving to the cloud shifts load from your LAN to your internet connection. Businesses on a connection adequate for email frequently need an upgrade, plus a secondary line for resilience.
- Retraining and support. A visible spike in helpdesk tickets for several weeks after cutover is normal and should be budgeted.
- Over-provisioning. Cloud resources sized like physical servers — for peak load, permanently — is the single largest source of ongoing waste. Right-sizing after migration typically reclaims a meaningful share of the monthly bill.
The comparison people get wrong
"Cloud is cheaper than servers" is not reliably true, and the honest comparison requires counting what the on-premise option actually costs.
Against a server you own, count: the hardware itself amortised over its life, replacement at end of life, the operating system and licensing, power and cooling, physical space, backup infrastructure, the maintenance labour, and the business risk of a hardware failure with a lead time on replacement parts.
Counted fully, cloud is frequently comparable rather than dramatically cheaper — and the real argument for it is usually not cost. It is that capacity becomes elastic, resilience is achievable without buying a second set of everything, remote work stops being a special case, and you stop spending capital in lumps every five years.
Businesses that migrate purely to reduce cost are sometimes disappointed. Those that migrate for flexibility and continuity generally are not.
How to keep it predictable
- Insist on a discovery phase before any fixed quote. Anyone pricing a migration without inventorying your applications and data is guessing.
- Get the ongoing monthly cost modelled, not just the project cost. Ask for the figure at 12 and 36 months.
- Budget the overlap period explicitly.
- Identify the applications that cannot move early. There are usually one or two, and they determine whether you end up hybrid.
- Plan a right-sizing review 60 to 90 days after cutover. Initial provisioning is always generous, and this review usually pays for itself.
- Confirm who owns the tenant and the data. It should be your business, not your provider.