A cloud business case is usually built on avoided hardware refresh and improved resilience. Both are real. The overrun that follows is also real, and it normally traces back to three early decisions.
The first is lifting workloads without resizing them. On-premise servers were bought for peak load five years into their life. Running the same shape in the cloud pays for capacity nobody uses.
The second is deferring the landing zone. Identity, network segmentation, tagging and budget alerts are far cheaper to establish before workloads arrive than to retrofit across dozens of subscriptions afterwards.
The third is treating cost as a finance topic rather than an engineering one. When teams cannot see the running cost of what they deploy, consumption grows quietly until someone escalates the invoice.
A migration with sizing discipline, a landing zone in place and per-workload cost visibility usually lands close to its business case. One without them rarely does.