On-premise or cloud: a decision, not a trend
The cloud migration conversation usually starts from the wrong end — with the platform rather than with the workload.
What actually decides it
Predictability of load. Steady, constant workloads are cheaper on hardware you own. Spiky or seasonal ones are cheaper rented. A server running at a flat sixty percent all year is the classic case for owning it.
Data gravity. If a workload processes large volumes of local data, moving it to the cloud means paying to move that data twice, continuously.
Latency. Design software, machine control and video editing all suffer over a connection. The office next to the data is sometimes the right answer.
Regulation and residency. Some contracts and some regimes specify where data physically sits. That narrows the options before any technical discussion.
Who maintains it. Owned hardware requires someone to patch, monitor and replace it. If that person does not exist, the true cost of on-premise is higher than the invoice suggests.
Where the cloud clearly wins
- Email and collaboration — running your own mail server is rarely defensible now
- Disaster recovery — an off-site copy without a second building
- Anything with an unpredictable growth curve
- Test environments that exist for two weeks at a time
Where owning still wins
- Large stable storage volumes, where monthly rental compounds badly
- Workloads pinned to specific hardware or licensing
- Sites where connectivity is genuinely unreliable
- Cases where five-year total cost, honestly calculated, favours the purchase
The hybrid most companies land on
Identity, email and collaboration in the cloud. File and application servers local, backed up off-site. Disaster recovery cloud-based. That shape is common because it follows the logic above rather than a slogan.
Do the five-year calculation including electricity, cooling, replacement and the labour to run it. Then decide per workload, not per company.
We work through this in IT consulting.