The real cost of a cloud migration

The bill goes up after the migration. That is what most organisations find at the twelve-month mark, and it surprises them all the more because it contradicts the original investment case. There is nothing abnormal about it.

Why rehosting costs more

A virtual machine sized for an annual peak, migrated as-is, stays sized for that peak — except it is now billed hourly instead of being amortised. Cloud charges you for the elasticity you are not using.

The real gains come from elsewhere: from the ability to shrink outside peak periods, to replace a permanent server with a service consumed per call, or to remove an entire layer in favour of a managed service. None of that happens simply by moving the disk image.

The trade-off to make application by application

• Rehost — fast, no economic gain, justified only by a datacentre exit on a fixed calendar.
• Replatform — replacing infrastructure components with their managed equivalents. Moderate effort, real operational gains.
• Refactor — partial rewrite to exploit elasticity. Significant effort, and the only path to a durable reduction in the bill.
• Retire — the most profitable option, and the most frequently forgotten. A migration is the best moment to notice that an application no longer has users.

What we recommend

Assess the estate before choosing a path, and accept that the answer differs by application. A programme that applies the same treatment to a hundred applications has made a default, not a decision.

And instrument costs from the first wave. Visibility obtained after the fact changes nothing about trade-offs that are already frozen.