AWS State of Cost Efficiency report: what the score means
Published and reviewed: 1 October 2026
Scope my cloud cost reviewAWS published its State of Cost Efficiency report on 9 June 2026.

On this page
What AWS published on 9 June 2026
The AWS Cloud Financial Management blog published « The AWS State of Cost Efficiency Report » on 9 June 2026. AWS analyzed optimization patterns across more than 71,000 anonymized, opted-in customers over the most recent quarter. The metric is the Cost Efficiency score in AWS Cost Optimization Hub: a daily score from 0 to 100% that measures the share of your optimizable spend that is already well optimized. It combines workload optimization (rightsizing, idle cleanup) and rate optimization (Savings Plans, Reserved Instances) in one number.
As of May 2026, AWS reports a median score of 83 and a mean of 79 across customers. The sample is customers who opted in, described by AWS itself, so it is a view of AWS's own population, not of the market. AWS also notes that the score can change when new recommendation types are released.
A high score can coexist with oversized resources
Because Savings Plans count as optimized spend, a higher coverage raises the score. AWS says so directly and warns that this can mislead: the underlying resources may still be oversized or idle. In its data, customers with 95 to 100% Savings Plans coverage see the remaining non-Savings-Plans optimization opportunity drop by 65 to 80% compared with customers at 0 to 25% coverage. The opportunity is masked rather than gone, and AWS says it becomes actionable again when commitments come up for renewal or workloads change.
The order AWS recommends: shrink first, then commit
The report's action list is ordered by risk and effort. Idle resources first: AWS gives as an example of an idle EC2 instance one with peak CPU below 5% and network I/O under 5 MB per day over 14 days, and advises checking that the resource is no longer needed, taking a snapshot before deleting an EBS volume, and considering stopping rather than deleting an RDS instance. Then rightsizing, starting in non-production. Then commitments, layered in smaller increments against the leaner footprint, which avoids a large cliff of expiring commitments.
The reasoning is concrete: if commitments are bought on over-provisioned capacity, a later rightsizing frees up commitment rather than lowering the bill. The Savings Plans versus Reserved Instances guide walks through the purchase decision, and the rightsizing guide covers the review that should precede it.
Memory metrics and recommendation settings
AWS says that enabling EC2 memory metrics is among the highest impact steps for EC2 optimization, since for many instances memory data decides whether a rightsizing recommendation appears at all. It reports that 17.7% of eligible customers have it enabled. It also says the score may dip after you enable them, because new opportunities surface. A fall in the score after an observability change is therefore not a deterioration of the bill, and a report that presents it as one has misread the metric.
Questions to settle and documents to keep
- How does the proposal treat existing commitments and overlapping recommendations when it lists savings?
- Does the benefit ledger keep gross and net benefit apart, with engineering effort and fees in the net figure?
- Are rate actions and usage actions recomputed together, so a rightsizing does not leave you over-committed?
Keep the metric definition, the date of your score and the boundary of optimizable spend, compared with your total invoice. The FinOps KPIs guide describes how to verify realized benefit after implementation on comparable demand.
Primary sources
The AWS Cloud Financial Management blog post, read in full for this page:
Continue your review
Request a quote for my project
Describe your project in a few lines and receive a quote from a specialist.
Request my quote