Reserved Instances & Savings Plans: The Complete Enterprise Guide

What Are Reserved Instances?

Reserved Instances (RIs) are a billing mechanism — not a different type of virtual machine. When you purchase a Reserved Instance, you commit to paying for a specific amount of compute capacity over a 1-year or 3-year term. In exchange, the cloud provider reduces the per-hour rate you are charged for matching instances.

The physical compute experience is identical to on-demand instances. The difference exists only at the billing layer: when your account runs an instance that matches a reserved instance you own, the on-demand rate is replaced by the lower reserved rate.

AWS Reserved Instances in Detail

AWS offers two categories of Reserved Instances for EC2:

Standard Reserved Instances offer the highest discounts (up to 72%) but cannot be exchanged for a different instance type. They can be sold on the AWS Reserved Instance Marketplace if your needs change.

Convertible Reserved Instances offer slightly lower discounts (up to 54%) but can be exchanged for any RI of equal or greater value. This flexibility is valuable over 3-year terms as your workload evolves.

Prices are approximate US East (us-east-1) Linux rates. Verify current pricing at aws.amazon.com/ec2/pricing.

Azure Reserved VM Instances

Azure Reserved VM Instances work similarly to AWS, offering 1-year and 3-year commitments for specific VM families in specific regions. Azure uniquely offers an instance size flexibility benefit for most VM families: a reservation for a D4s_v3 (4 vCPU/16 GB) automatically covers equivalent compute capacity in D2s_v3 (2 vCPU/8 GB) or D8s_v3 (8 vCPU/32 GB) instances within the same family.

Azure also offers Azure Hybrid Benefit, which allows organizations with existing Windows Server or SQL Server licenses with Software Assurance to apply those licenses in Azure, saving 40–85% on Windows VM costs. Combined with Reserved Instances, this can reduce Windows VM costs by up to 80% compared to on-demand Windows pricing.

GCP Committed Use Discounts

Google Cloud Platform takes a different approach with Committed Use Discounts (CUDs). Rather than purchasing specific instance types, you commit to a minimum amount of vCPUs and memory for 1 or 3 years. This is more flexible than AWS Standard RIs because the commitment applies across all N2, N2D, C2, M1, M2, and custom machine types in a region.

CUD discounts for 1-year commitments are approximately 37% below on-demand for most general-purpose machine types. 3-year commitments provide approximately 55% discounts. GCP additionally provides Sustained Use Discounts (SUDs) automatically — no commitment needed — when a resource runs for more than 25% of a month. Full-month resources receive a 20–30% automatic discount on top of any committed use pricing.

Savings Plans vs Reserved Instances

AWS Savings Plans (introduced 2019) offer a more flexible alternative to Reserved Instances for many use cases:

RI Purchasing Strategy

The most effective enterprise RI strategy follows a tiered commitment approach:

1. Analyze 3 months of usage data to identify your consistent baseline capacity — the minimum vCPU and memory footprint that runs continuously regardless of traffic.
2. Cover the baseline with 3-year Reserved Instances or Committed Use Discounts. This floor represents your safest commitment because the capacity is always needed.
3. Cover predictable growth with 1-year Reserved Instances. Shorter commitments provide flexibility to right-size as your architecture evolves.
4. Use on-demand for traffic spikes above your reserved baseline. Auto-scaling should add on-demand capacity during peak periods.
5. Use spot instances for batch and background workloads to maximize discount depth on non-critical compute.

Breakeven Analysis Framework

Before purchasing Reserved Instances, calculate the breakeven utilization rate — the minimum percentage of time the instance must run for the RI to be cost-effective compared to on-demand:

For a 1-year No Upfront RI at 35% discount, the breakeven utilization is approximately 54%. If your instance runs less than 54% of the time (roughly 13 hours per day), on-demand is actually cheaper. For 3-year All Upfront at 57% discount, the breakeven drops to approximately 43%.

For production workloads running 24/7, Reserved Instances are almost always the right choice. For development environments running 8–10 hours per day on weekdays only (approximately 25–30% utilization), on-demand or Spot instances are typically more cost-effective.