AWS Benchmark of 71,000 Customers Reveals the Limits of Passive Discount Commitments

AWS published findings from an analysis evaluating optimization patterns across more than 71,000 anonymized customer accounts using its Cost Efficiency metric. The data reveals that while the median customer Cost Efficiency score sits at 83, the mean is 79, weighed down by a substantial long tail of under-optimized accounts. Critically, only 17.7% of eligible customers enable EC2 memory metrics—a single telemetry enhancement associated with an 8 to 30 percentage point boost in actionable savings per Compute Optimizer recommendation. Furthermore, although commitment coverage remains standard practice, only 47.1% of organizations with active Savings Plans combine them with workload rightsizing. This benchmark delivers a much-needed empirical reality check for engineering leaders and FinOps practitioners. Discount instruments like Savings Plans and Reserved Instances are easily executed at the procurement layer, leading organizations to mistake financial rate reduction for architectural efficiency. In reality, high commitment coverage (95% to 100%) masks between 65% and 80% of visible non-commitment optimization opportunities, including architectural rightsizing and Graviton migrations. Teams that blindly lock in commitments without cleaning waste end up paying discounted rates on zombie infrastructure. This dynamic mirrors the broader maturation curve in cloud financial management. For years, organizations prioritized speed-to-savings via centralized financial purchasing, leaving workload architectures untouched. As modern infrastructure budgets face intense pressure to fund capital-intensive AI initiatives, passive FinOps is proving insufficient. The trend is decisively pivoting toward proactive, engineering-integrated optimization: pairing detailed observability with automated rightsizing and customized recommendation thresholds. In practice, engineering teams should mandate the ingestion of memory utilization metrics across all compute fleets before executing subsequent Savings Plans cycles. Fine-tuning recommendation engines—such as adjusting headroom parameters in AWS Compute Optimizer—yields measurable improvements in efficiency scores without risking workload stability. Practitioners must enforce a strict 'shrink first, commit second' pipeline in their deployment governance to ensure baseline workloads run lean before locking in long-term financial commitments.