Cloud Cost Optimization: How to Cut Your AWS, Azure & GCP Bill in 2026

Cloud providers make it extremely easy to spin up resources and extremely difficult to notice when you've stopped using them. The result: the average business wastes 32% of its cloud spend every month.

What is FinOps?

FinOps — short for Financial Operations — is the discipline of bringing financial accountability to cloud spending. The FinOps Foundation defines it as "an operational framework and cultural practice which maximizes the business value of cloud by bringing technology, finance, and business together to make data-driven spending decisions."

In plain English: FinOps is the practice of making sure your engineering, finance, and business teams all understand what the cloud costs, why it costs that, and what to do about it.

Before FinOps, cloud billing was typically an engineering concern that finance saw once a month, often with shock. FinOps makes cost visibility a continuous, shared responsibility — with clear ownership, real-time monitoring, and accountability built in.

Where cloud waste actually hides

The Flexera 2025 State of the Cloud Report found that organizations waste an average of 32% of their cloud spend. That waste doesn't come from one place — it accumulates across dozens of small inefficiencies:

The most common waste categories

• Over-provisioned compute: Instances sized for peak load running at average load. The fix — rightsizing — is the single highest-impact optimization for most businesses.
• Idle resources: Dev and test environments left running over weekends and holidays. A dev environment running 24/7 instead of business hours costs 3× more than necessary.
• Orphaned storage: EBS volumes, managed disk snapshots, and S3 buckets from projects that no longer exist. These accumulate invisibly and are often discovered only during an audit.
• Data transfer costs: Egress fees — what you pay to move data out of the cloud — are often the most surprising line item. Architecture decisions made without cost awareness can generate unexpected egress charges at scale.
• Unoptimised storage tiers: Hot storage (expensive) used for data that is accessed infrequently. Moving to cold or archive tiers can cut storage costs by 60–80% for eligible data.
• Licence waste: Enterprise software licences (SQL Server, Windows, Oracle) attached to cloud instances that are underused or duplicated.

Quick wins: savings in the first 30 days

A cloud cost audit almost always surfaces immediate savings that require no architectural changes — just identification and cleanup. Here is what a typical 30-day optimization sprint looks like:

Week 1: Visibility and baseline

Turn on cost allocation tags if they aren't already configured. Without tags, you cannot attribute spend to teams, projects, or environments. Enable AWS Cost Explorer, Azure Cost Management, or GCP Cost Management, and pull 90 days of spend history. Build a baseline: which services are driving the bill, how has spend trended, and where are the biggest line items?

Week 2: Delete the obvious waste

Idle resources are usually the fastest wins. Run the provider's idle resource reports (AWS Trusted Advisor, Azure Advisor, GCP Recommender) and filter for:

• Stopped EC2/VM instances that have been stopped for more than 30 days
• Unattached EBS volumes and managed disks
• Unused Elastic IP addresses and static IPs
• Old snapshots beyond your retention policy
• Unused load balancers

Do not delete without confirmation from the owning team — but most of these are safe to remove or snapshot. This step typically recovers 5–15% of total spend.

Week 3: Rightsize compute

Pull 14 days of CPU and memory utilization data for every running instance. Instances running below 20% average CPU are candidates for downsizing. Instances running below 10% average CPU are almost certainly over-provisioned by at least one instance size. Test the smaller instance type in a staging environment before changing production. This step typically recovers 10–20% of compute spend.

Week 4: Schedule non-production environments

Dev, test, and staging environments do not need to run 24/7. Scheduling them to run only during business hours (Monday–Friday, 8am–7pm) reduces their runtime by ~70%, cutting their cost by the same amount. AWS Instance Scheduler, Azure Automation, and GCP Cloud Scheduler all support this natively.

Reserved Instances and Savings Plans explained

The biggest long-term lever for cloud cost reduction is committing to usage in exchange for a discount. All three major providers offer this in some form.

Reserved Instances (RIs)

You commit to a specific instance type, operating system, and region for 1 or 3 years. In exchange, you get 30–60% off on-demand pricing. The trade-off is inflexibility — if your needs change, you are still paying for what you committed to. Most providers now offer convertible RIs that allow some changes.

Savings Plans (AWS) / Azure Reservations / GCP Committed Use

More flexible than RIs. Instead of committing to a specific instance, you commit to a spend amount per hour (e.g., "$5/hour of compute"). The discount applies across instance families, sizes, and regions within the commitment scope. This is the right choice for most businesses because it accommodates change without full rigidity.

The golden rule for commitments: Only commit to your baseline usage — the minimum you will definitely need no matter what. Cover peaks with on-demand or spot instances. Over-committing is worse than under-committing.

AWS, Azure, and GCP: provider-specific optimization tips

Building governance that keeps savings permanent

The problem with one-time cloud audits is that the waste comes back. Without governance — policies, tagging, budgets, and accountability — engineers spin up resources and forget them, and the bill creeps back up within 6 months.

Mandatory tagging

Require every cloud resource to have at minimum: an owner tag (email or team name), an environment tag (production, staging, dev), and a project/cost-centre tag. Enforce this via cloud policy (AWS Service Control Policies, Azure Policy, GCP Organization Policy) so untagged resources can't be created. Without tags, there is no accountability.

Budget alerts

Set monthly budgets for every team and project, with alerts at 80% and 100% of budget. When a team gets an alert, they investigate. This one practice alone prevents most budget overruns before they happen.

Regular rightsizing reviews

Run rightsizing reports monthly, not once. Your instance needs change as traffic patterns evolve. What was correctly sized six months ago may now be over-provisioned as you've optimized your application.

A designated FinOps owner

Someone needs to own the cloud bill. This does not have to be a full-time role — a part-time FinOps practice with a designated owner who reviews spend weekly and runs quarterly optimization sprints is sufficient for most SMBs. Without an owner, nothing gets done.

Canadian considerations for cloud cost

Canadian businesses have some specific factors that affect their cloud cost strategy:

Data residency and PIPEDA compliance

PIPEDA and provincial privacy laws (particularly Quebec's Law 25) may require that certain personal data remain in Canada. AWS (Canada Central — Montreal), Azure (Canada Central — Toronto, Canada East — Quebec City), and GCP (Montreal, Toronto) all offer Canadian regions. However, Canadian regions are sometimes more expensive than US regions, and not all services are available in Canadian regions — which affects your architecture choices and cost.

Currency and FX risk

All three major cloud providers bill in USD. A weakening Canadian dollar means your cloud bill effectively rises without any change in usage. Building FX risk into your cloud budget and monitoring it monthly prevents unexpected overruns. Some large enterprises hedge this risk; most SMBs simply account for it in their planning.

SR&ED eligibility

Cloud computing costs that directly support qualifying scientific research and experimental development (SR&ED) work may be eligible for SR&ED tax credits. This includes cloud compute used for software development, AI model training, and system testing. Documenting cloud spend by project and qualifying activity is worth doing if you have active SR&ED claims.

The cloud cost audit checklist

Use this checklist to structure your first cloud cost audit:

Visibility

• Cost allocation tags are enabled and enforced on all resources
• Cost Explorer / Cost Management is configured and reviewed monthly
• Budget alerts are set for all major accounts and projects
• 90-day spend trend is documented and understood

Waste elimination

• Idle instances (stopped for 30+ days) identified and reviewed
• Unattached storage volumes and orphaned snapshots catalogued
• Unused load balancers and static IPs identified
• Non-production environments scheduled to shut down out of hours

Rightsizing

• 14-day CPU and memory utilization pulled for all instances
• Instances below 20% average CPU flagged for rightsizing review
• Rightsizing tested in staging before production change
• Database instances reviewed for over-provisioning

Commitments

• Baseline usage identified (minimum steady-state compute)
• Savings Plans / Reserved Instances coverage calculated
• Commitment recommendations reviewed in provider console
• Commitment renewal dates calendared