Cloud Cost Optimization Strategies Cheat Sheet

Updated 2026-05-25

Next Topic: Cloud Data Transfer and Egress Costs Cheat Sheet

🧠Study flashcards on this topic131 cards · spaced repetition→

Cloud cost optimization is the continuous practice of reducing cloud expenditures while maintaining or improving performance, reliability, and business value across AWS, Azure, Google Cloud, and other cloud platforms. It encompasses strategic approaches ranging from commitment-based discounts and resource rightsizing to automated waste elimination and real-time cost governance. As organizations increasingly adopt multi-cloud and AI workloads in 2026, cost optimization has evolved from periodic reviews to continuous, automated control loops that tie cloud spending directly to unit economics and business outcomes. The most successful strategies combine technical optimization (right-sizing, spot instances, storage tiering, AI/GPU efficiency) with organizational practices (FinOps culture, shift-left cost governance, showback/chargeback) and platform-native tools (AWS Compute Optimizer, Azure Cost Management, GCP Active Assist).

What This Cheat Sheet Covers

This topic spans 19 focused tables and 151 indexed concepts, 131 flashcards. Below is a complete table-by-table outline of this topic, spanning foundational concepts through advanced details.

Table 1: Commitment-Based Pricing ModelsTable 2: Compute Resource RightsizingTable 3: Autoscaling and ElasticityTable 4: Storage Cost OptimizationTable 5: Network and Data Transfer OptimizationTable 6: Idle Resource Detection and EliminationTable 7: Cost Visibility and MonitoringTable 8: Cost Allocation and GovernanceTable 9: Shift-Left FinOps and IaC Cost GovernanceTable 10: Container and Kubernetes OptimizationTable 11: Serverless Cost OptimizationTable 12: AI/ML and GPU Cost OptimizationTable 13: Database Cost OptimizationTable 14: Advanced FinOps PracticesTable 15: Multi-Cloud Cost ManagementTable 16: Monitoring and Observability Cost OptimizationTable 17: Data Transfer and CDN OptimizationTable 18: Backup and Disaster Recovery Cost OptimizationTable 19: Cloud Cost Optimization KPIs and Metrics

Quick IndexSubscribe to unlock

A jump-to index of every table row in this cheat sheet.

Mind MapSubscribe to unlock

An interactive map of every table and concept in this topic.

Table 1: Commitment-Based Pricing Models

Commitment-based discounts are the single highest-leverage lever most organizations underuse — trading flexibility for deep, predictable savings on stable workloads. Choosing the right commitment type, term length, and coverage target requires understanding how each model balances lock-in risk against discount depth.

Strategy	Example	Description
Savings Plans	`Compute SP:` `$10/hour commit` `Any instance family/region`	• Flexible pricing model offering up to 72% savings based on dollar-per-hour commitment • automatically applies to EC2, Fargate, Lambda across instance families and regions.
Reserved Instances	`EC2 Standard RI:` `1-year, 72% discount` `Instance type locked`	Long-term commitment (1-3 years) for specific instance types offering deep discounts but limited flexibility across instance families or regions.
Spot Instances	`60-90% discount` `Interruptible workloads` `Use 10+ instance types`	• Spare capacity at massive discounts but can be preempted with 2-minute warning • ideal for fault-tolerant batch jobs, CI/CD, data processing, and AI/ML training.
Azure Reserved VM Instances	`3-year Azure RI:` `Up to 72% savings` `Size flexibility in family`	• Azure's commitment pricing with size flexibility within the same instance family and region • supports instance size exchanges.
Committed Use Discounts	`GCP 1-year CUD:` `37% compute discount` `Spend or resource-based`	• GCP commitment model offering up to 57% savings for 1 or 3-year terms • available as spend-based (flexible) or resource-based (specific machine types).
Preemptible VMs	`GCP: up to 91% off` `24-hour max runtime` `No SLA guarantees`	• Google Cloud's interruptible instances offering deepest discounts but limited to 24-hour sessions • requires checkpoint/restart logic.
Group Sharing	`Share across accounts` `RI/SP pooling` `Launched Nov 2025`	• AWS feature enabling Reserved Instances and Savings Plans to be shared across management groups rather than individual accounts • improves utilization across large organizations.
Flex Commitments	`Insurance for commitments` `Reduce lock-in risk` `Pay premium for flexibility`	• Third-party offerings that provide insurance against commitment underutilization • allows refunds or exchanges if commitment needs change unexpectedly.

Table 2: Compute Resource Rightsizing

Rightsizing is typically the highest-impact optimization action after commitments — most cloud workloads are over-provisioned by 30–50% based on actual utilization. The goal is matching instance type, family, and size to what the workload genuinely needs, using historical metrics rather than guesswork.

Technique	Example	Description
Rightsizing	`m5.xlarge → m5.large` `50% cost reduction` `CPU avg < 20%`	• Matching instance types and sizes to actual workload requirements based on historical utilization metrics • typically the highest-impact optimization lever after commitments.

The rest of this section is part of the full cheat sheet

Table 3: Autoscaling and Elasticity

Autoscaling transforms cloud infrastructure from static provisioning to dynamic capacity that tracks real demand — paying only for what's used, when it's used. Getting scaling policies right (especially combining predictive, reactive, and schedule-based strategies) is key to avoiding both waste and performance failures.

Method	Example	Description
Target Tracking Scaling	`Keep CPU at 70%` `Auto-adjust capacity` `Simple configuration`	• Autoscaling policy that maintains a specific metric target (e.g., CPU utilization) • automatically scales in/out to maintain the setpoint; easiest to configure correctly.

The rest of this section is part of the full cheat sheet

Table 4: Storage Cost Optimization

Storage costs grow silently — data accumulates, snapshots multiply, and lifecycle policies are never configured. The core principle is moving data to the cheapest tier that meets its access-time requirements, automating transitions, and aggressively deleting what is no longer needed.

Technique	Example	Description
Storage Tiering	`Hot → Cool → Archive` `S3 Standard → IA → Glacier` `90%+ cost reduction`	• Moving data between storage classes based on access frequency and retention requirements • dramatic cost reduction for infrequently accessed data.

The rest of this section is part of the full cheat sheet

Table 1: Commitment-Based Pricing Models

Strategy	Example	Description
Savings Plans	`Compute SP:` `$10/hour commit` `Any instance family/region`	• Flexible pricing model offering up to 72% savings based on dollar-per-hour commitment • automatically applies to EC2, Fargate, Lambda across instance families and regions.
Reserved Instances	`EC2 Standard RI:` `1-year, 72% discount` `Instance type locked`	Long-term commitment (1-3 years) for specific instance types offering deep discounts but limited flexibility across instance families or regions.
Spot Instances	`60-90% discount` `Interruptible workloads` `Use 10+ instance types`	• Spare capacity at massive discounts but can be preempted with 2-minute warning • ideal for fault-tolerant batch jobs, CI/CD, data processing, and AI/ML training.
Azure Reserved VM Instances	`3-year Azure RI:` `Up to 72% savings` `Size flexibility in family`	• Azure's commitment pricing with size flexibility within the same instance family and region • supports instance size exchanges.
Committed Use Discounts	`GCP 1-year CUD:` `37% compute discount` `Spend or resource-based`	• GCP commitment model offering up to 57% savings for 1 or 3-year terms • available as spend-based (flexible) or resource-based (specific machine types).
Preemptible VMs	`GCP: up to 91% off` `24-hour max runtime` `No SLA guarantees`	• Google Cloud's interruptible instances offering deepest discounts but limited to 24-hour sessions • requires checkpoint/restart logic.
Group Sharing	`Share across accounts` `RI/SP pooling` `Launched Nov 2025`	• AWS feature enabling Reserved Instances and Savings Plans to be shared across management groups rather than individual accounts • improves utilization across large organizations.
Flex Commitments	`Insurance for commitments` `Reduce lock-in risk` `Pay premium for flexibility`	• Third-party offerings that provide insurance against commitment underutilization • allows refunds or exchanges if commitment needs change unexpectedly.

Table 2: Compute Resource Rightsizing

Technique	Example	Description
Rightsizing	`m5.xlarge → m5.large` `50% cost reduction` `CPU avg < 20%`	• Matching instance types and sizes to actual workload requirements based on historical utilization metrics • typically the highest-impact optimization lever after commitments.

Table 3: Autoscaling and Elasticity

Method	Example	Description
Target Tracking Scaling	`Keep CPU at 70%` `Auto-adjust capacity` `Simple configuration`	• Autoscaling policy that maintains a specific metric target (e.g., CPU utilization) • automatically scales in/out to maintain the setpoint; easiest to configure correctly.

Table 4: Storage Cost Optimization

Technique	Example	Description
Storage Tiering	`Hot → Cool → Archive` `S3 Standard → IA → Glacier` `90%+ cost reduction`	• Moving data between storage classes based on access frequency and retention requirements • dramatic cost reduction for infrequently accessed data.