Cost efficiency as an NFR applies FinOps principles to ensure infrastructure spend is proportional to business value delivered. The key framework is Total Cost of Ownership (TCO), which includes compute, storage, network egress, licensing, personnel, and opportunity cost. Cloud cost optimization follows a hierarchy: eliminate waste first (idle resources), right-size second (over-provisioned instances), then commit (reserved instances or savings plans), and finally architect for cost (spot instances, serverless, tiered storage). Organizations that treat cost as a first-class NFR alongside performance and availability build unit economics awareness into every architectural decision.

Key Points

  • Reserved Instances / Committed Use Discounts save 30-60% over on-demand for stable baseline workloads — commit to 1-year for most services, 3-year only for commodity infrastructure.
  • Spot/Preemptible instances offer 60-90% discount but can be reclaimed in 2 minutes — suitable for stateless, fault-tolerant, and batch workloads with checkpointing.
  • Network egress is frequently the largest surprise cost — data leaving a cloud region costs $0.09/GB on AWS; design for data locality, use CDNs, and minimize cross-region calls.
  • Serverless (Lambda, Cloud Run) eliminates idle compute cost and scales to zero; cost breaks even vs. containers at ~50-70% CPU utilization of the equivalent container size.
  • Storage tiering: S3 Intelligent-Tiering automatically moves infrequently accessed data to cheaper tiers; move logs >30 days to Glacier, >1 year to Glacier Deep Archive (saving ~95%).
  • FinOps maturity model: Inform (tagging, cost visibility) → Optimize (right-sizing, commitments) → Operate (showback/chargeback to teams, unit economics KPIs).
  • Cost anomaly detection alerts (AWS Cost Anomaly Detection, GCP Budget Alerts) catch runaway spend — set at 20% above expected weekly spend as a minimum.
  • Carbon efficiency increasingly overlaps with cost efficiency — Spot instances and serverless architectures have lower carbon footprints per unit of work, aligning sustainability and FinOps goals.

Real-World Example

Dropbox's "Magic Pocket" project migrated ~500 PB of data from AWS S3 to their own physical infrastructure over 2016-2017, saving approximately $75M per year by the time the migration was complete — a landmark example that at sufficient scale, cloud costs can justify significant capital investment in owned infrastructure.