Cloud Cost Optimisation: Cutting Your Azure / AWS Bill

Summary: Cloud bills can balloon overnight when they're not actively managed. At SME scale, there are six concrete ways to bring an Azure or AWS bill down: clean up idle resources, right-size compute, buy reserved capacity or a savings plan, optimise storage tiers, automate start/stop on non-production VMs, and control egress traffic. With these six practices, a typical SME can cut its annual cloud bill by 30–50% — no extra investment required, just the discipline to audit regularly.

Cloud adoption in SMEs starts with enthusiasm — "flexible, scalable, modern". Then the first big invoice lands and you hear "but we shut that server down…". Cloud elasticity can be a financial win; mismanaged, it can turn into 2–3× a comparable on-premise bill. Cost control isn't about more investment, licences, or consultants — it's a maintenance discipline.

In this article we walk through concrete steps for cutting Azure and AWS bills at SME scale. Target audience: IT managers, finance leads, and decision-makers who want to turn "why is our cloud bill so high?" into a systematic answer.

Why an SME Cloud Bill Typically Bloats

Bills usually grow for these reasons:

1. Forgotten Resources

• A VM spun up "for testing", forgotten 6 months ago
• An unused database created by a former developer
• Detached disks (not deleted), billed by the hour
• Old snapshots (every snapshot is extra storage)
• Public IP allocations from years back

2. Wrong Sizing

• 4 vCPU / 16 GB RAM provisioned where 1 vCPU / 2 GB would do
• Premium SSD bought where Standard is enough
• Production-scale instances used in a development environment

3. No Reserved Capacity

• You've been running the same VM on pay-as-you-go for 3 years
• A 1-year or 3-year reservation could give 30–72% off

4. Egress Traffic

• Cloud → internet data egress is one of the most expensive line items
• Cross-region traffic (e.g. Europe → US)
• Serving large files without a CDN

5. Lack of Automation

• Development VMs run 24/7 (in use only 8 hours)
• Test environments running through the night and weekend

Step 1: Clean Up Idle Resources

The first — and quickest — savings area.

In Azure

# Stopped (not deallocated) VMs — still billing
Get-AzVM -Status | Where-Object {$_.PowerState -eq "VM stopped"}

# Detached managed disks
Get-AzDisk | Where-Object {$_.DiskState -eq "Unattached"}

# Old snapshots
Get-AzSnapshot | Where-Object {$_.TimeCreated -lt (Get-Date).AddMonths(-3)}

In AWS

• EC2 instances: stopped instances still bill for EBS
• EBS volumes: detached volumes pay full price
• Old AMIs / snapshots: accumulate monthly
• Elastic IPs: unassociated IPs bill by the hour
• CloudWatch Logs: old retention

# Detached EBS
aws ec2 describe-volumes --filters Name=status,Values=available

# Unassociated EIP
aws ec2 describe-addresses --query "Addresses[?AssociationId==null]"

The Routine

• Monthly scan as a habit
• Anything idle for 90+ days gets deleted
• Take a snapshot first, then delete (so you can roll back)

Step 2: Right-Sizing

VMs sitting under 20% CPU/RAM are oversized.

Azure Advisor

The built-in recommendation engine in the Azure portal:

• "Right-size or shutdown underutilized virtual machines"
• 14 days under 5% CPU = a recommendation
• One click to switch to a smaller SKU

AWS Compute Optimizer

AWS's equivalent built-in tool:

• 14 days of metric analysis
• Suggested instance type
• Estimated savings

Manual Right-Sizing

• Collect 1–2 months of metrics (CloudWatch, Azure Monitor)
• Did CPU peak over 80% even once?
• Did RAM peak over 80% even once?
• If the answer is no, drop down one tier

Typical Savings

25–40% savings from right-sizing are common in SME environments.

Step 3: Reserved Capacity or Savings Plans

Reserved-capacity discounts for steady-state workloads.

Azure Reserved Instances (RI)

• 1-year or 3-year commitment
• Available for VMs, SQL DB, Cosmos DB, Storage
• 1 year: ~40% off
• 3 years: ~55–60% off
• Upfront or monthly payment options

Azure Savings Plan

• More flexible (instance-type agnostic)
• For the compute category
• 1 or 3 years
• Up to ~65% off

AWS Reserved Instances + Savings Plans

Similar structure:

• EC2 Reserved Instance: tied to a specific instance type
• Compute Savings Plan: flexible across EC2 + Fargate + Lambda
• EC2 Instance Savings Plan: family-based, region-based
• 1 year: ~40%, 3 years: ~72% off

SME Strategy

• Production VMs: 3-year Savings Plan / Reserved Instance — 50–60% off
• Dev/test: pay-as-you-go (and shut down when not in use)
• Volatile workloads: pay-as-you-go (a reservation here is risky)

A typical split is 80% steady production and 20% variable workload.

Step 4: Storage-Tier Optimisation

Not everything needs to live on Premium SSD.

Azure Storage Tiers

Tier	Access	Cost	Use
Premium SSD	Instant	High	Production DBs, IO-heavy
Standard SSD	Instant	Medium	Web servers, office apps
Standard HDD	Instant	Low	Low-IO, backup
Cool Blob	Instant (read cost higher)	Low	30+ days untouched
Archive	Hours (rehydration)	Very low	Annual archive

AWS Storage Tiers

• gp3 / gp2: general purpose
• io2: high IOPS
• st1: throughput-oriented (large files)
• sc1: cold access
• S3 Standard / Intelligent-Tiering / Glacier: object storage

Lifecycle Policy

Automatic tier transitions:

• New file → Standard
• After 30 days → Cool / IA
• After 90 days → Archive

This setup runs without manual care once the policy is defined.

Step 5: Automated Start / Stop

Development and test environments don't need to run 24/7.

The Scenario

• Working hours: 09:00–19:00 (10 hours)
• Weekdays: 5 days
• Instead of 24/7 → 50 hours/week = ~30% runtime
• Savings: ~70%

Implementation

Azure Automation:

• Scheduled start/stop via Runbook
• Tag-based selection (using an "auto-shutdown" tag)

AWS:

• AWS Instance Scheduler
• Lambda + EventBridge

Tag strategy:

• auto-shutdown: 19:00
• auto-start: 09:00
• environment: dev

A Typical SME Gain

10 dev/test VMs × ~70% savings = a significant line item.

Step 6: Egress (Outbound) Traffic

Cloud → internet data transfer is one of the most expensive items.

Cost Ladder

Direction	Cost
Internet → Cloud (ingress)	Free
Within the same region	Free or near-free
Cross-region	Medium
Cross-cloud	High
Cloud → Internet (egress)	High

Optimisation

• Use a CDN: CloudFront, Azure CDN — cheap for large static content
• Region choice: the same region as your customers (a Europe region if you serve from Türkiye)
• Cross-region replication: is it really necessary?
• Compression: HTTP gzip enabled
• Bandwidth alerts: catch unexpected spikes

A Common Mistake

A customer portal pushing gigabytes daily, no CDN: high monthly bill. A CDN cuts that by up to 80%.

FinOps Discipline

FinOps (Financial Operations) principles for continuous optimisation.

Three Phases

1. Inform: who's spending what, right now?
2. Optimize: where are the savings opportunities?
3. Operate: ongoing monitoring and action

Tag Strategy

Mandatory tags on every resource:

• environment: prod / dev / test
• owner: team-or-person
• project: project-name
• cost-center: accounting-code

These tags drive monthly reporting: which project spent what.

Monthly Cost Review

Even at SME scale, 30 minutes a month for a cost review:

• Month-over-month delta
• Anomalies (surprise spikes)
• Action list
• Budget target

Budget Alerts

The single most important safeguard against surprise bills.

Azure

• Cost Management → Budgets
• Monthly limit + 50% / 80% / 100% alerts
• Email + Action Group webhook

AWS

• AWS Budgets
• Per-service budgets
• Email + SNS topic

Practical for SMEs

• Total monthly budget + 80% alert
• A separate budget for test environments (a hard cap)
• Anomaly detection (Cost Anomaly Detection service)

What Yamanlar Bilişim Offers

Our cloud-cost support areas at SME scale:

• Analysis of your current cloud bill
• Right-sizing recommendations
• Reserved Instance / Savings Plan strategy
• Lifecycle policy design
• Automated start / stop rollout
• Tag strategy and budget alerts
• Monthly cost-review service
• KVKK-aware cloud-region decisions

Frequently Asked Questions

• Azure: Azure Advisor + Microsoft Cost Management; third-party CloudHealth, Spot.io
• AWS: AWS Compute Optimizer + Trusted Advisor; third-party Cloudability, Apptio
• Multi-cloud: Spot.io, FinOut, Vantage

For SMEs the built-in tools (Advisor, Trusted Advisor) are usually enough; once you tally the savings, you typically find a third-party tool isn't worth the extra cost.

Conclusion

Cloud cost optimisation doesn't require a dedicated FinOps team — it's a steady maintenance discipline. Cleaning up idle resources, right-sizing, reserved capacity, storage-tier optimisation, automated start / stop, and egress control are the six basics that take most SMEs to 30–50% annual savings. None of them require investment; they require habits.

Yamanlar Bilişim provides cloud-bill audits, optimisation recommendations, and monthly review services sized to your needs — turning your cloud bill from a source of surprises into a manageable, reportable line item.