In this article
Cloud providers charge for what you use, and it is easy to use more than you need. Servers get left on, old data piles up on expensive storage, and nobody looks at the bill until it is a surprise. The five checks below find the most common waste. They work the same way on AWS, Azure and Google Cloud, although each provider uses different names.
Take care with every change. Ask your engineer or cloud partner to make it, take a backup first, and change one thing at a time. The aim is a smaller bill, not a broken system.
First, make the bill readable
You cannot cut what you cannot see. Do two things before the checks.
Add tags. A tag (called a label in Google Cloud) is a short note attached to a resource, such as 'project: website' or 'owner: sales'. AWS says activated cost allocation tags let you organise and track costs by tag, and Google Cloud says labels let you break down your billed charges. Once everything is tagged, you can see which project or team spends what.
Set a budget with alerts. A budget is a monthly amount you choose, and the provider warns you as spending approaches it. Azure says its budgets do not stop your resources or your consumption, and Google says an alerts-only budget does not cap usage. AWS Budgets can also trigger an action, such as a rule that blocks new resources, but AWS warns that a notice can arrive after the charge is made. Treat a budget as an early warning, not a safety net.
Check 1: Switch off what nobody uses
Idle resources are the easiest saving. Typical examples are servers for a project that ended, test systems left running over weekends, and copies of databases nobody opens. Ask who owns each one. If nobody does, stop it first, wait a few weeks, and delete it only after taking a backup.
The providers help you find these. AWS Compute Optimizer identifies idle resources. Azure Advisor recommends shutting down virtual machines (rented servers) that were not used at all in the last seven days. Google Cloud gives idle VM recommendations.
Check 2: Match server size to real use
This is called right-sizing. Servers are often chosen large to be safe, and then never use their full power. Look at how much processing power and memory each one uses over several weeks. If it uses only a small part of what it has, a smaller size may do the same job for less. Leave room for your busiest days, and test after the change.
AWS Compute Optimizer gives rightsizing recommendations, Azure Advisor recommends resizing underused virtual machines, and Google Cloud gives machine type recommendations. Azure notes that its savings figures use retail prices, so the real saving may be lower.
Check 3: Move old data to cheaper storage
Storage costs less when you read it less often. Each provider offers tiers (also called classes): hot ones for data you use daily, and cooler ones for data you rarely touch. AWS has S3 Standard-IA, the Glacier classes and S3 Intelligent-Tiering, which moves objects between tiers by itself for a small monitoring fee. Azure Blob Storage has hot, cool, cold and archive tiers. Google Cloud Storage has Standard, Nearline, Coldline and Archive classes, plus Autoclass, which moves data automatically.
Cooler tiers have trade-offs. They charge more to read the data, they have minimum storage periods (Azure lists 30, 90 and 180 days for cool, cold and archive), and archived data can take hours to get back. Do not archive anything you may need quickly.
Check 4: Watch data transfer
Data moving into a cloud is often free, but data moving out costs money, and so does data moving between regions (the cloud's locations). On its bandwidth page, Azure lists data in as free and charges for data out after a monthly free allowance. AWS also lists a monthly free allowance for data out to the internet, with charges above it. Big downloads, backups copied to another region and services that talk across regions can add up quietly.
Look for data transfer lines on your bill. Ask your engineer to keep services that talk to each other in the same region.
Check 5: Pay less for steady use with a commitment
If a workload will run all year, you can often get a lower price by promising to use it. AWS has Savings Plans, Azure has savings plans, and Google Cloud has committed use discounts. Each is tied to a one-year or three-year term.
The catch is that you are committed. Azure says savings plan purchases cannot be cancelled or refunded, and Google says you are billed a monthly fee for the whole commitment period. So do checks 1 to 4 first, and commit only to the steady base you are sure to use. Committing before you remove the waste locks the waste in.
Make it a habit
Once a month, check the budget alerts, look at the biggest lines on the bill, and find an owner for any cost that has no tag. Small, regular checks stop the waste from coming back.
What to do next
If you want a second pair of eyes on your cloud setup, send us your requirement. We reply with a written proposal showing scope, timeline and quote, at no cost until you agree. Every system we deliver has monitoring, tests and human checks.
Sources, checked on 19 October 2026: AWS documentation (AWS Budgets, cost allocation tags, Compute Optimizer, S3 storage classes, Savings Plans, EC2 on-demand pricing); Microsoft Learn (Cost Management budgets, Azure Advisor cost recommendations, blob access tiers, savings plans, bandwidth pricing); Google Cloud documentation (Cloud Billing budgets, labels, idle VM recommendations, storage classes, committed use discounts).