In-depth guides on AWS cost management, FinOps practices, rightsizing, Reserved Instances, and Savings Plans. Our engineers share real billing teardowns, unit economics frameworks, and cost anomaly playbooks from production environments.
EC2 On-Demand is not one price — it is a matrix of family, size, OS, tenancy, and region. In us-east-1 (June 2026), a m7g.large Linux instance runs $0.0816/hr while a GPU g5.xlarge is $1.006/hr. Most bill surprises come from wrong family choice, not wrong size.
ElastiCache bills node-hours for Redis/Valkey clusters and ECPUs for Serverless. A cache.r7g.large with two replicas is 3× the primary node cost before data transfer. Serverless wins below ~$800/mo steady cache spend; provisioned nodes win above it.
Fargate bills per vCPU-second and GB-second — not per task hour. A 0.25 vCPU / 0.5 GB task running 24/7 costs ~$9/mo in us-east-1; a 4 vCPU / 16 GB task costs ~$145/mo. ECS on EC2 adds cluster overhead but wins at sustained utilization above ~55%.
A 22-account AWS Organization spent $1.1M/yr on Compute Savings Plans but applied only 61% to production — dev sandboxes burned the commit while prod stayed On-Demand. Group Sharing (April 2026) fixed attribution; the waste was $312k/yr before they restructured purchases.
CUR 2.0 (generally available 2024, default for new exports in 2025) adds line-item resource IDs and Savings Plan attribution columns Athena needs for serious chargeback. A standard FinOps query pack answers top-10 services, untagged spend, and SP coverage in under 60 seconds.
Compute Optimizer added 32-day lookback for cyclic workloads in 2025 — critical for monthly batch jobs that looked oversized on 14-day windows. In a June 2026 audit of a $48k/mo account, rightsizing recommendations alone modeled $11k/mo savings at low performance risk.
Aurora bills instance hours plus storage plus I/O — or storage at a 125% premium with I/O bundled (I/O-Optimized). Serverless v2 at $0.12 per ACU-hour wins on variable workloads, loses on steady traffic above 4 ACU. Global Database doubles the price per secondary region plus replicated-write fees. Backtrack adds per-change-record cost most teams never factor in.
Bedrock AgentCore is metered across twelve distinct components — Runtime, Browser, Code Interpreter, Gateway, Identity, Memory (two tiers), Observability, Evaluations, Payments, Search, and the underlying model spend. Two of them drive 80% of the bill.
CloudFront bills $0.085/GB egress in North America tiered down to $0.020/GB at extreme volume, plus $0.0075–$0.0100 per 10K requests, plus origin egress. Regional price classes drop the bill 30–60% by skipping expensive geographies. Real-time logs at $0.01 per million entries surprise high-traffic sites. Lambda@Edge is dramatically more expensive than CloudFront Functions.
CloudWatch bills across ten distinct dimensions — Logs ingestion at $0.50/GB, Logs storage, custom metrics with cardinality multipliers, alarms tiered by resolution and type, dashboards at $3 each, Synthetics canary runs, RUM events, X-Ray traces, Container Insights, and cross-region replication. Logs ingestion is the largest line on most accounts.
DynamoDB on-demand bills $1.25 per million writes and $0.25 per million reads — pay-per-request convenience with no capacity planning. Provisioned at $0.00065/WCU-hour wins on steady traffic above ~50% utilization. Every GSI doubles the write cost, IA tier cuts storage 60% but adds 25% to request fees, and Global Tables compound the bill per region.
EBS is billed across six dimensions — gp3 storage at $0.08/GB-month, separate IOPS and throughput line items, io2 Block Express, snapshots at $0.05/GB-month, fast snapshot restore at $0.75/DSU-hour. Deleting an EC2 instance never stops the volume bill, and the gp2→gp3 migration is one of the cleanest 20% wins in cloud cost.
