Partners

Cloud

Reserved Instances vs Savings Plans: Picking the Right Cloud Commit

Apr 26, 2026 5 min read

If you bought blindly into a three-year RI, you are probably overspending. A walkthrough of when each commitment model actually saves money in 2026.

What You Are Actually Committing To

Reserved Instances (RIs) on AWS commit you to a specific instance type, size, region, and optionally platform and tenancy — in exchange for discounts of 40–72% off on-demand pricing over 1 or 3 years. Savings Plans are more flexible: Compute Savings Plans apply to any instance family, size, region, or OS, and Fargate usage. EC2 Instance Savings Plans are more restrictive but offer higher discounts. Azure has Reserved Virtual Machine Instances and Savings Plans with similar mechanics. The flexibility-discount tradeoff is the same across all three clouds.

When Reserved Instances Still Win

RIs make sense when your workload is predictable, long-lived, and runs on a specific instance family that you do not expect to change. A production database cluster running on db.r6g.2xlarge with consistent throughput for the next three years is a clear RI candidate. The discount is higher than a Savings Plan, and there is no ambiguity in what is covered. RIs also let you share coverage across accounts in an AWS Organisation, which matters at scale.

When Savings Plans Are the Smarter Bet

If your team is migrating from one instance family to another, scaling unpredictably, or running a mix of container, serverless, and EC2 workloads, Compute Savings Plans give you the coverage flexibility to handle that without stranding committed capacity. The downside is a slightly lower discount ceiling. For most mid-market companies that are still in architectural flux, Savings Plans are the safer starting point — you can always layer RIs for rock-stable workloads later.

The 30-Day Optimisation Sequence

Pull 30 days of CloudWatch/Cost Explorer data and identify your top 20 steady-state compute resources by spend. For each, determine: has this instance type changed in the past 12 months? Is it likely to change in the next 24? If no to both, RI it at 3 years all-upfront. Everything else, cover with a Compute Savings Plan at 1 year. Leave 15–20% of compute spend on-demand as a buffer. Run this exercise quarterly — the cloud bill changes faster than most finance teams realise.

Key Takeaways

Ready to Put This Into Practice?

Talk to VSERV about cloud cost optimisation and building a commitment strategy that actually reduces your AWS or Azure bill.