Management and Governance
AWS Cost Anomaly Detection
AWS Cost Anomaly Detection uses machine learning to identify unusual AWS spend and issue alerts. It can surface unexpected model, accelerator, data-transfer, or logging cost changes after a GenAI release.
Key points
- Monitors can scope analysis by service, account, tag, or cost category
- Alert subscriptions define thresholds and recipients
- Detection follows billing data and is not a real-time quota control
When to use it
- Detect a sudden rise in Bedrock invocation spend
- Flag an idle GPU fleet or runaway logging volume
Exam tips
- Use anomaly detection for unexpected spend and budgets or application quotas for planned limits
- Tag shared GenAI resources so an anomaly can be attributed to a team or feature