An organisation rehosts its estate to cloud infrastructure and finds costs higher and reliability no better. What did rehosting not deliver, and what should follow?

What rehosting does and does not deliver

Delivers: exit from ageing hardware or a data-centre contract, on a predictable timeline, with low application risk. That is a legitimate objective and rehosting achieves it well.

Does not deliver: cost reduction, reliability improvement, or elasticity. Those come from how applications are built, not from where they run.

Why costs go up

1. Sized for peak, billed continuously. On-premises hardware sized for peak is a sunk cost; the same sizing in cloud is a continuous bill for capacity that is idle most of the time.

2. No elasticity to exploit. An application that cannot scale horizontally, or that requires manual intervention to scale, runs at peak size permanently.

3. Managed services not adopted. Self-managed databases, queues and load balancers on cloud instances cost more than the equivalent managed services and retain all the operational burden.

4. Network costs discovered. Cross-zone and cross-region transfer, NAT gateway data processing, and egress — none of which existed as line items on-premises, and all of which are shaped by an architecture that was designed when networking was free.

5. No attribution or governance, so nothing pushes back on the growth.

Why reliability does not improve

The application's failure modes are unchanged. A single-instance application with no redundancy and manual failover has exactly the same availability on new infrastructure — and may have slightly worse, because the underlying instances are less reliable individually than dedicated hardware, with the compensating availability coming from redundancy the application does not use.

What should follow

Rehost is a first step, not an outcome. The sequence that delivers:

1. Attribution and cost visibility, so the growth has an owner. Nothing else works without this.
2. Rightsizing against percentile utilisation with explicit failure headroom — not against averages.
3. Replatform the highest-cost components onto managed services, which removes both cost and operational burden.
4. Address elasticity for the applications with variable demand, which is where the cloud's economic model actually applies.
5. Redundancy and failover for the applications whose availability matters, which is an application change rather than an infrastructure one.

The framing to establish beforehand

Rehosting buys time and an infrastructure exit. It does not buy cloud benefits. Organisations that present it as a cloud transformation set an expectation that the next budget cycle will contradict — and the disappointment is frequently what stops the follow-on work that would have delivered the benefits.