AWS Billing and Pricing Models

One of the biggest advantages of cloud computing is the shift from large capital expenditure (CapEx) to operational expenditure (OpEx). Instead of buying hardware upfront, you pay only for what you use. However, AWS pricing can be complex — understanding the models helps you optimise costs and avoid bill shock.

---

Fundamental Pricing Principles

AWS pricing is built on three foundations:

Principle	Description
Pay as you go	No upfront commitments. Pay only for resources you consume, for as long as you use them
Pay less when you reserve	Commit to 1 or 3 years of usage and receive significant discounts
Pay less as you use more	Volume-based pricing tiers — the more you use, the lower the unit cost

---

How Services Are Priced

Different AWS services use different pricing dimensions:

Compute

Service	Pricing Dimensions
EC2	Per second or per hour, based on instance type, OS, and Region
Lambda	Per request + per GB-second of compute time
Fargate	Per vCPU-hour + per GB-hour of memory

Storage

Service	Pricing Dimensions
S3	Per GB stored per month + per request (GET, PUT, etc.) + data transfer out
EBS	Per GB provisioned per month + IOPS (for io1/io2 volumes)
Glacier	Per GB stored + per retrieval request (cost varies by retrieval speed)

Database

Service	Pricing Dimensions
RDS	Instance hours + storage + I/O + backup storage
DynamoDB	Read/write capacity units (provisioned) or per request (on-demand)
Aurora	Instance hours + I/O requests + storage

Networking

Type	Cost
Data in	Free (almost always)
Data out to internet	Charged per GB (first 100 GB/month free in many cases)
Data between AZs	Small charge per GB
Data between Regions	Charged per GB
Data within the same AZ	Free (using private IPs)

---

EC2 Pricing Options

EC2 offers several purchasing options, each suited to different use cases:

On-Demand Instances

• Pay by the second (Linux) or by the hour (Windows)
• No commitment, no upfront cost
• Full price — the "standard" rate

Best for: Unpredictable workloads, short-term development, testing

Reserved Instances (RIs)

• Commit to 1 or 3 years of usage
• Discounts of up to 72% compared to On-Demand
• Three payment options:

Payment Option	Discount Level	Cash Flow
All Upfront	Highest discount	Pay everything now
Partial Upfront	Medium discount	Pay some now, rest monthly
No Upfront	Lowest discount	Pay monthly

Best for: Steady-state, predictable workloads (production databases, always-on services)

Savings Plans

• Flexible alternative to Reserved Instances
• Commit to a dollar amount per hour of compute usage for 1 or 3 years
• Two types: