How much does cloud migration cost? (and when not to move)

"We're moving to the cloud and paying less" is the most repeated sales line of the last decade. And it's false half the time. The cloud isn't cheaper: it's more flexible, and flexibility is billed. What changes isn't the price, it's who decides when it goes up: before, your server provider; now, the person who leaves an environment running over the weekend.

In this article you'll find the three types of migration with their real ranges, the invoice line items nobody warns you about in month thirteen and — most useful of all — the cases where migrating will cost you more than staying put. Because those exist too, and almost nobody talks about them.

The myth that keeps costing money: the cloud is always cheaper

The honest comparison isn't "own server versus cloud". It's "own server with its electricity, its cooling, its hardware every five years and the person maintaining it" versus "cloud with its variable bill, its licences, its data egress and the person governing it". Once you include both columns in full, the result depends on three things: size, load stability and who looks at the invoice every month.

For a company with one small, stable, fully depreciated server, the cloud can cost more from month one. For a business with peaks —a store that multiplies traffic during a campaign, a portal with seasonal queries— the cloud wins by a mile, because you pay for what you use and not for what you might use. And for everything in between, the answer is a word nobody wants to hear: it depends.

The three types of migration (and what each one costs)

The word "migration" hides three completely different projects. Confusing them is the number one cause of budgets that double halfway through.

Rehost: lift and shift

Your current server is brought up on a cloud virtual machine, without touching the application. It's the fastest and cheapest, and also the one that makes the worst use of the cloud: you pay for machines that still behave as if they were in your office. A small server can land at $30-150 per month; an environment with a database and a replica, $200-800. It's a valid option when the goal is leaving hardware behind, not optimising.

Replatform: swap pieces without rebuilding

The application stays the same, but the database moves to a managed service, and so do backups, email or files. This is where real time savings start: you stop patching the database engine by hand. The project usually costs $2,000-12,000 depending on the number of systems, and the monthly bill drops or stabilises because the server that only existed for that disappears.

Redesign: actually use the cloud

The application is rewritten to scale on its own and be billed by usage. It's the most expensive project —from $15,000 and no ceiling— and it only makes sense if the business grows unpredictably or the current system can't cope. Doing it "for modernity" without that reason is the most expensive way of fixing nothing.

Real ranges: from a website to an ERP

This table summarises what the project costs and what it costs afterwards, which is the part that almost never appears in the proposal. Indicative figures in dollars, excluding taxes.

The month-thirteen invoice: what nobody warns you about

The first month almost always goes well. The problem arrives when the staging environment is still running, the logs have been growing for a year and someone discovers that getting data out costs money. These are the line items that eat the savings:

• Data egress: moving information out of the provider is billed. If your business exports a lot —reports, video, copies to another site— that number can exceed the server's.
• Storage that only grows: photos, logs, backups and old versions. Nobody deletes anything because nobody knows what can be deleted.
• Hourly licences: Windows Server, SQL Server and some tools are billed by usage time. A machine running with no traffic still bills.
• Backups and recovery: backup isn't free, and neither is disaster. Having a backup without ever testing a restore is having an illusion with an invoice.
• Provider support: the basic plan doesn't answer when what matters goes down. The plan that does answer is billed separately.
• The person watching the invoice: if nobody governs spend, the cloud grows on its own. It's the cost almost nobody budgets and the one that saves the most.

The cost of not governing the cloud

There's a whole discipline —FinOps— built around a very simple idea: someone has to review the invoice every month with the numbers in front of them. Tagging resources by project, switching off what isn't used, reserving capacity for what is stable and setting spend alerts. For small companies that translates into two hours a month and savings that usually beat 15% of the bill. It isn't magic: it's that nobody was looking.

When NOT to migrate (and what to do instead)

There are four situations where migration is a bad idea, and they're worth saying before signing a project:

1. The server is fully depreciated, stable and you have someone to run it. Replacing something that works with a variable bill isn't modernising, it's moving house.
2. The software vendor doesn't support the cloud. Some ERPs, workshop tools and industry systems only run on-premises. Migrating them means losing support.
3. Leaving costs more than the saving. If exporting your data or changing providers has a high price, the flexibility they sell you doesn't exist.
4. There's nobody to operate it. The cloud doesn't administer itself. Without someone watching access, backups and costs, risk goes up more than spending goes down.

What to do instead

There's an intermediate option that solves 80% of cases: leave operations where they are and take to the cloud only what adds value —backups, a disaster recovery environment, email—. You pay little, you gain peace of mind and you don't touch what works. That hybrid is, for many small businesses, the most profitable decision of all.

Conclusion: migrate what moves, not what can move

The cloud is an excellent tool for what changes size and a bad idea for what has been the same for five years. The right question isn't "when do we move?", it's "which part of our IT gains something by moving?". If the answer is none, the most profitable project is not doing it.

At TakeYourDesign we run migrations with numbers before and after: cloud services, IT infrastructure and verified backups. If you want to know what it would cost in your case, tell us what you have running today and we'll give you a range with no obligation. And if what you need is the step-by-step process, it's in the cloud migration guide for businesses.

Frequently asked questions

How much does cloud migration cost?

$300-1,500 for a website, $1,500-6,000 for two to four servers and $5,000-40,000 for an ERP, plus the monthly bill afterwards.

Is the cloud cheaper than an own server?

Not always. With a small, stable, fully depreciated server it can cost more; with traffic peaks the cloud usually wins.

What is data egress cost?

What the provider charges for moving data out of its platform. If you export a lot, it can exceed the cost of the server.

When is it NOT worth migrating?

When the server is depreciated and stable, when the software vendor doesn't support the cloud, or when there's nobody to govern the bill. See cloud services.

What is FinOps and what is it for?

Reviewing the bill every month with a method: tagging resources, switching off what isn't used and reserving capacity for what's stable. It usually saves over 15%.

How do I know what it would cost in my case?

Tell us what you have running today and we'll give you a range with before-and-after numbers, with no obligation.