Open Cost Explorer or Azure Cost analysis. Group last month's spend by usage type, sort descending, and look at the top ten lines. In most accounts we review, at least a couple of them are things nobody can explain: storage attached to instances that were deleted months ago, a test environment running every weekend, snapshots from a migration that finished last year.
That's the first place to reduce cloud costs, and it doesn't need a re-architecture. The big early savings come from waste that's easy to find once someone actually looks: idle resources, oversized instances, the wrong storage tier, and non-production running around the clock.
Here are ten quick wins across AWS and Azure that a small team can work through in about 30 days, in an order that stops you locking in waste.
How cloud bills creep up
Creating a resource takes a minute. Forgetting it takes no effort at all. Each team optimises for shipping, and nobody owns the total. The FinOps Foundation's State of FinOps 2026 report still has workload optimisation and waste reduction as the single top current priority for practitioners, which tells you how universal the problem is.
Both clouds now ship free tools that flag most of this for you. The work is acting on the findings, in the right order, without breaking production.
10 quick wins to reduce cloud costs in 30 days
Go roughly in order. Microsoft's guidance on savings plans and reservations says it in five words: "Discounts reduce rates, not waste." Clean up before you commit.
Week 1: get visibility and delete the obvious
- Set budgets and anomaly alerts. On Azure, budgets alert on actual or forecast cost, but Microsoft is explicit that resources aren't affected and consumption isn't stopped. On AWS, Cost Anomaly Detection uses machine learning to flag unusual spend and ranks root causes by service, account, Region and usage type. Expect up to 24 hours' lag, since it works from Cost Explorer data. Turn both on day one.
- Delete idle and orphaned resources. Unattached volumes and disks, old snapshots, unused load balancers, idle NAT gateways, and stopped instances still paying for storage.
aws ec2 describe-volumes --filters Name=status,Values=availablelists unattached EBS volumes; on Azure, filteraz disk liston adiskStateofUnattached. Azure Advisor also recommends shutting down VMs that weren't used at all over the last seven days, judged on CPU and outbound network. - Find an owner for everything else. No owner, no excuse to keep it. Even a basic
ownertag makes the rest of the month far easier. Our guide to tagging and showback for small teams covers the minimum set.
Week 2: right-size compute
- Work through the right-sizing recommendations. AWS Compute Optimizer is free at its default settings once you opt in, looks at 14 days of CloudWatch metrics, and covers EC2, Auto Scaling groups, EBS, Lambda, RDS and Aurora, NAT gateways and more. Azure Advisor does the same for VMs and scale sets. Start with non-production. Move to production with a rollback plan and a check on memory, not just CPU.
- Move to cheaper instance families. AWS states that Graviton-based instances cost up to 20% less than comparable x86 instances. Most Linux containers, interpreted languages and managed databases such as RDS move across with little or no code change. Native dependencies and older Docker images built only for amd64 are the usual blockers, so test first.
Week 3: storage and schedules
- Switch EBS gp2 volumes to gp3. AWS lists gp3 at up to 20% lower price per GB than gp2, with a baseline of 3,000 IOPS and 125 MB/s at any size. The change is an online
modify-volumecall with no detach. Check bigger gp2 volumes first: gp2's IOPS and throughput scale with size, so a large gp2 volume can outperform gp3's baseline and you may need to provision extra IOPS or throughput on gp3. Compare the volume's CloudWatch metrics before you switch. - Tier object storage by access pattern. S3 Intelligent-Tiering moves objects not accessed for 30 consecutive days to an Infrequent Access tier (up to 40% saving) and to Archive Instant Access after 90 days (up to 68%). Objects under 128 KB aren't tiered, so it does little for buckets full of tiny files. On Azure, the blob access tiers cool, cold and archive have minimum retention periods of 30, 90 and 180 days, with prorated early deletion charges, so only push data down if you'll keep it that long.
- Schedule non-production. Dev and test rarely need to run overnight or at weekends. Instance Scheduler on AWS uses resource tags and Lambda to start and stop EC2 and RDS instances on a schedule. Azure VMs have a built-in auto-shutdown setting, but the time zone defaults to UTC. Set it to AUS Eastern (or your local zone), or your 7pm shutdown happens at 5am or 6am Sydney time, depending on daylight saving.
Week 4: change how you pay
- Use Spot for work that can be interrupted. AWS quotes savings of up to 90% off On-Demand for Spot Instances, and Azure quotes up to 90% off pay-as-you-go for Spot VMs. Good fits: CI runners, batch jobs, stateless workers. Never your only database.
- Commit to the cleaned-up baseline. Only after the first eight steps. AWS Savings Plans offer up to 66% off On-Demand for Compute Savings Plans and up to 72% for EC2 Instance Savings Plans. Azure reservations offer up to 72% off pay-as-you-go. Our Savings Plans vs Reserved Instances guide explains how much to commit.
AWS cost optimization and Azure cost optimization side by side
| Quick win | AWS | Azure | Effort |
|---|---|---|---|
| Spend alerts | AWS Budgets, Cost Anomaly Detection | Cost Management budgets | Low |
| Idle resources | Compute Optimizer idle findings | Advisor shutdown recommendations | Low |
| Right-sizing | Compute Optimizer | Advisor resize recommendations | Medium |
| Cheaper compute | Graviton instances | Newer VM series, B-series burstable | Medium |
| Block storage | gp2 to gp3 | Disk SKU review | Low |
| Object storage | S3 Intelligent-Tiering, lifecycle rules | Blob access tiers, lifecycle management | Low |
| Scheduling | Instance Scheduler on AWS | VM auto-shutdown | Low |
| Interruptible work | Spot Instances | Spot VMs | Medium |
| Commitments | Savings Plans, Reserved Instances | Savings plans, reservations | Low effort, high risk if oversized |
Mistakes that cost more than they save
- Buying commitments first. You lock in today's waste for one or three years.
- Taking recommended savings at face value. Microsoft notes that Advisor's estimates use retail rates and ignore existing reservations and savings plans, so real savings can be lower. Occasionally a resize even costs more, if it moves you off a reserved series.
- Right-sizing on CPU alone. Check memory, disk throughput and network before downsizing anything in production. Compute Optimizer only sees EC2 memory if the CloudWatch agent or a supported external source is sending it.
- Archiving data you still read. Cooler tiers charge more to read, and an archived Azure blob can take up to 15 hours to rehydrate.
When you want a second pair of eyes
Our cloud cost optimisation review starts with read-only access to billing and the recommendation tools across your AWS and Azure accounts. In week one we pull the top cost drivers, run Compute Optimizer and Advisor, and check for the usual suspects: orphaned storage, idle NAT gateways, gp2 volumes, non-production running 24/7.
You get a ranked list of savings with the estimated monthly amount, the effort, and the risk of each. Then we make the changes with your team, through your infrastructure as code where you have it, so the fixes survive the next deployment instead of being reverted by Terraform the following week.
We'll also tell you what not to do yet, which is usually buying a three-year commitment in month one. Once the baseline is clean, we model commitment options and set up budgets, anomaly alerts and a monthly showback so costs stay visible after we step back.
Want to know what your bill could look like in 30 days? Request a cloud cost review and we'll show you the quick wins in your own accounts.
Frequently asked questions
How quickly can we reduce cloud costs?
Most of the quick wins here can be done within 30 days. Deleting idle resources, scheduling non-production and switching storage types show up on the next bill. Right-sizing production takes longer because each change needs testing. Commitments should come last, once the baseline is clean, so the discount you lock in is based on what you actually need.
What is the fastest AWS cost optimization win?
For many accounts, it's deleting idle resources and moving EBS volumes from gp2 to gp3, which AWS prices up to 20% lower per GB. Turning on Compute Optimizer is also quick: it's free at its default settings and flags idle and oversized resources across EC2, EBS, Lambda, RDS and other services using 14 days of metrics.
Do Azure budgets stop spending when the limit is reached?
No. Azure Cost Management budgets send alerts when actual or forecast costs cross a threshold, and Microsoft states that resources aren't affected and consumption isn't stopped. For subscription and resource group budgets you can attach an action group that triggers automation, such as a runbook that shuts down non-production VMs, but you have to build that response yourself.
Should we buy Savings Plans or reservations straight away?
No. Clean up first. Commitments lower the rate you pay but do nothing about waste, and they're hard to undo: AWS states that Savings Plans can't be cancelled during the term, and Microsoft states that Azure savings plans can't be cancelled or refunded. Remove idle resources and right-size, then commit to a conservative share of what's left.
Cloud Cost Savings
Request a cloud cost review
Tell us where you're stuck and we'll come back with practical next steps.
Get in touchSources
- FinOps Foundation: State of FinOps 2026 Report
- Microsoft Learn: Decide between a savings plan and a reservation
- Microsoft Learn: Tutorial, create and manage budgets
- AWS: Detecting unusual spend with AWS Cost Anomaly Detection
- Microsoft Learn: Azure Advisor VM and VMSS cost recommendations
- AWS: What is AWS Compute Optimizer?
- AWS: AWS Graviton Processor
- AWS: Amazon EBS General Purpose Volumes
- AWS: Amazon S3 Intelligent-Tiering
- Microsoft Learn: Access tiers for blob data
- AWS: Instance Scheduler on AWS
- Microsoft Learn: Auto-shutdown a VM
- AWS: Amazon EC2 Spot Instances
- Microsoft Azure: Spot Virtual Machines
- AWS: Savings Plans types
- AWS: Compute Savings Plans and Reserved Instances
- Microsoft Learn: What are savings plans?
- Microsoft Learn: What are Azure Reservations?