Why the invoice goes up
On-premise servers are sized for peak and paid for once. In the cloud you rent that peak capacity every hour, including the nights and weekends when it sits idle. A like-for-like rehost of an over-provisioned estate reliably produces a bill higher than the hosting it replaced, and the savings in the business case never materialise.
Where the money actually is
Rightsizing against observed utilisation. Shutting down non-production outside business hours. Commitment coverage on genuinely steady workloads. Storage tiering. Replacing self-managed databases and middleware with managed equivalents so you stop paying engineers to patch them. None of these are migration activities — they are modernisation activities, and they are where the return lives.
Sequencing that works
Build the landing zone with tagging, budgets and guardrails first. Migrate a representative pilot wave and measure real cost against the model. Rightsize immediately rather than "later". Then modernise the workloads where the return justifies the effort, and be honest about the ones where it does not.
The honest business case
We model three scenarios: naive rehost, rehost plus FinOps discipline, and selective modernisation. The first is usually cost-negative, and showing that explicitly is what makes the third credible to a finance director who has been oversold before.