IT Infrastructure & Cloud
Calculating Monthly Cloud Costs and Preventing Bill Overruns
A cloud bill is made of several components that grow in different ways. This guide maps the components, how to estimate the monthly figure, and the controls that keep the bill from ballooning.
SEO and Content Specialist
9 min read
Short answer: a monthly cloud bill consists of six main components: compute, storage, outbound data transfer, managed services, licenses, and support plans. Estimate it with the provider’s official calculator, then put four controls in place from day one: budgets and alerts, tags on every resource, rightsizing to match the real load, and usage commitments only for stable workloads. A budget sends a warning, it does not stop usage.
Many companies move to the cloud with a monthly cloud cost estimate that looks reasonable, then receive a much larger bill a few months later. The cause is rarely one big thing. It is usually a pile of small ones: a test server left running, a disk left behind after a server was deleted, data transfer that was never counted, or a server size copied from the old server. This guide covers where the numbers on the bill come from, how to estimate them before migrating, and the controls that keep them readable.
Six components of a monthly cloud bill
Each provider uses different product names, but the bill lines can be grouped the same way.
| Component | Examples | How it grows |
|---|---|---|
| Compute | Virtual machines, containers, functions | Per hour or per second while running, according to machine size |
| Storage | Machine disks, object storage, snapshots, backups | Per capacity per month, plus the number of read and write requests |
| Data transfer | Data out to the internet, between zones, between regions | Per volume of data. On Google Cloud, inbound data is not charged, while outbound and inter-zone data is |
| Managed services | Managed databases, load balancers, NAT, message queues | A mix of hourly charges and the volume processed |
| Licenses | Windows Server, SQL Server, marketplace software | Included in the machine price or brought with licenses you already own |
| Support | The provider’s technical support plan | The basic plan is included, higher tiers are billed separately |
Two components are most often missed in the first estimate: outbound data transfer (for example an application that sends large files to users, or backups copied out of the cloud) and leftover resources that are still billed even when unused, such as disks not attached to any machine and public IP addresses that are reserved but not used.
Licenses also need a separate check. Azure documentation, for example, states that the virtual machine reservation discount does not apply to Windows and SQL software costs. Those license costs are still calculated on their own.
How to estimate before migrating
All three major providers offer an official calculator: AWS Pricing Calculator, Google Cloud Pricing Calculator, and Azure Pricing Calculator. A calculator only computes what you enter, so the quality of the estimate depends on the data you prepare. AWS documentation also notes that its calculator results do not include taxes.
Steps you can follow:
- Real usage data from the current servers: average and peak CPU, memory, disk capacity, and monthly data growth. The size of the old server is not a reference, because it was usually bought with spare capacity for several years.
- Running hours: production servers run all month, but test and development servers often only need to run during working hours.
- Outbound data transfer volume: how much data is sent to users, to the office, and to backup locations outside the cloud.
- Managed services in use: databases, load balancers, NAT, and monitoring each have their own line.
- Licenses and support: note which are included in the machine price and which you bring yourself.
- Other environments: test and staging environments are often forgotten, even though they can match a large part of the production cost.
The result is best presented as a range (normal usage and peak usage), not a single number. A single-number estimate is almost certain to miss, and the gap is only questioned when the bill arrives.
Budgets and alerts: an alarm, not a brake
A budget is a cost limit you set for a period, usually monthly, together with alerts when usage approaches or passes it. It is the first control to put in place, with one important note: budgets generally do not stop usage.
| Provider | What the budget does | Note from the official documentation |
|---|---|---|
| AWS Budgets | Alerts on actual and forecasted cost, by email or Amazon SNS. Actions can be attached, for example an IAM policy that prevents new resources from being created | Data is updated up to three times a day, and there is a delay between usage and the alert |
| Google Cloud Budgets | Alerts at percentage thresholds, 50%, 90%, and 100% by default, also on forecasted cost. Notifications by email or Pub/Sub | An alerts-only budget does not set a hard cap on spending |
| Azure Budgets | Alerts on actual and forecasted cost, can call an action group for automation | Resources are not affected and consumption is not stopped. Cost data is typically available within 8 to 24 hours |
Because of the data delay, set several thresholds, for example an early alert on forecasted cost and an alert on actual cost. Send alerts to the budget owner and the technical team at the same time. Alerts that only reach the cloud admin’s mailbox are often read after the bill is out.
Tags: knowing who owns which resource
Tags (labels on Google Cloud) are key and value pairs attached to resources, for example environment = production or application = erp. Without tags, the bill only answers “how much”, not “for what and whose”.
Three things to know before designing a tagging scheme:
- On AWS, cost allocation tags must be activated in the Billing console before they appear in Cost Explorer and the cost allocation report, and they can take up to 24 hours to appear.
- On Azure, resources do not inherit tags from the resource group or subscription. Azure Policy is used to inherit or enforce them.
- On Google Cloud, budgets can be scoped by label at the billing account level, one label per budget.
A practical minimum scheme: environment (production, staging, test), application or system, owner (the team or person responsible), and cost center. Set it from the first resource you create. Tagging hundreds of resources later is much harder. Do not put sensitive data in tags, because tags are stored as plain text and appear in cost reports.
Rightsizing: matching size to the real load
Rightsizing means adjusting resource size to measured usage, not to the initial estimate. The most common pattern after a lift-and-shift migration is a cloud server that copies the size of the old server, while its CPU rarely reaches half. The pattern is discussed in What Is Lift-and-Shift in Cloud Migration?.
Providers offer tools for this. AWS Compute Optimizer, for example, analyzes the last 14 days of usage metrics by default to recommend sizes and flag idle resources. Recommendations still need context: a month-end close or a sales season can push the load far above a 14-day average.
A safe approach: reduce the size by one step, monitor for one business cycle, then repeat if there is still headroom. Turn off test environments outside working hours, and delete disks, snapshots, and IP addresses that are no longer attached to anything.
Usage commitments: only for stable workloads
All three providers offer lower prices if you commit to using a certain amount of resources for one or three years.
| Provider | Form of commitment | What to watch |
|---|---|---|
| AWS | Savings Plans (an hourly compute usage commitment) and Reserved Instances | Savings of up to 72% on compute according to AWS documentation. Prices are locked for the term |
| Google Cloud | Committed use discounts, resource-based or spend-based | A commitment cannot be cancelled after purchase, and is billed even when unused |
| Azure | Reservations | Up to 72% off pay-as-you-go prices. The discount applies to the bill and does not change the resources |
The practical rule: buy commitments after rightsizing, not before. Locking an oversized resource for three years means paying for idle capacity at a discounted price. Commit only to a base load that has been stable for several months, and leave the rest on pay-as-you-go.
Monthly review: who reads the bill
The controls above only work if someone reads the results. A short monthly review needs to answer five questions:
- How does this month’s bill compare with the budget and last month?
- Which component grew the most, and for which application or environment?
- Which resources are untagged or have no clear owner?
- Which rightsizing recommendations or idle resources have not been acted on?
- Are the commitments already purchased actually being used?
Appoint one budget owner from the business side, not only from the technical team. Decisions such as “do we still need this test environment” are business decisions.
How Legia works on this
In the Cloud Migration & Architecture service, the monthly bill estimate is prepared in the design stage, together with the architecture document, before anything is moved. Cost controls are set up from the start: tags on every resource, budget limits, usage alerts, and resource sizes based on the real load. While the project runs there is a monthly cost review with savings recommendations that can be acted on right away, and after handover the review can continue through a management subscription. The cloud provider’s bill is paid directly in your company’s name, separately from Legia’s service fee. The choice of how to move each system is discussed in Rehost, Replatform, or Refactor, and other services in this pillar are in IT Infrastructure & Cloud.
Frequently asked questions
Can a cloud budget stop usage when its limit is reached?
Generally not. Budgets on AWS, Google Cloud, and Azure mainly send alerts. Stopping or restricting usage has to be set up separately with automated actions, for example AWS Budgets actions or an Azure action group, and must be tested so it does not shut down production systems.
Why does the cloud bill rise when the number of users stays the same?
The most common causes are leftover resources (disks, snapshots, IP addresses, test environments), growth in data and backups, and increasing outbound data transfer. Tags and a monthly review help find which line is growing.
When should we buy Savings Plans, reservations, or committed use discounts?
After resource sizes have been adjusted and the base usage has been stable for several months. A one- or three-year commitment locks in an amount of usage, and on Google Cloud a commitment cannot be cancelled after purchase.
Is the estimate from the provider’s calculator final?
No. The calculator computes the configuration you enter and does not include taxes. Real usage, data transfer, and additional environments often make the bill differ from the estimate, so compare the estimate with the actual bill in the first months.
Sources
- AWS, Managing your costs with AWS Budgetsdocs.aws.amazon.com
- AWS, What is AWS Pricing Calculator?docs.aws.amazon.com
- AWS, Organizing and tracking costs using cost allocation tagsdocs.aws.amazon.com
- AWS, What is AWS Compute Optimizer?docs.aws.amazon.com
- AWS, What are Savings Plans?docs.aws.amazon.com
- Google Cloud, Create, edit, or delete budgets and budget alertsdocs.cloud.google.com
- Google Cloud, Network pricingcloud.google.com
- Google Cloud, Committed use discountsdocs.cloud.google.com
- Google Cloud Pricing Calculatorcloud.google.com
- Microsoft, Tutorial: Create and manage budgetslearn.microsoft.com
- Microsoft, Use tags to organize your Azure resourceslearn.microsoft.com
- Microsoft, What are Azure Reservations?learn.microsoft.com
- Azure Pricing Calculatorazure.microsoft.com

