AWS now supports spend limits

Spend limits let project owners cap a project's pre-tax monthly charges so workloads run within a predictable budget. When a project's usage reaches its limit AWS pauses the project and stops resources; data is preserved but some resources may need manual restart. Limits are project-scoped, visible to anyone with project access, require a Paid Plan, exclude credits, and can be applied to up to 10 projects. The minimum limit is the greater of $20 or a conservative estimate based on month-to-date activity, prior-month activity, and currently running resources; this minimum exists to avoid trivial fluctuations causing disruptive pauses. Notifications are sent at 50%, 75%, 90% and when on track to exceed the limit in 10 days. Spend limits are created and managed in AWS Settings under Billing.

Optional early cost controls intervene before the hard limit: about 7 days prior AWS stops new resource launches via service control policies; about 5 days prior idle resources are paused using Compute Optimizer recommendations (examples: pause EC2 if CPU <5% and network <5 MB/day over 14 days, special G/P GPU criteria; pause RDS MySQL/Postgres with no connections and low activity; pause SageMaker endpoints with zero invocations). About 4 days prior, top cost drivers from EC2, RDS, Lambda, Bedrock, and SageMaker can be paused: EC2 top drivers are terminated after creating EBS snapshots and releasing Elastic IPs; RDS is stopped (auto-restarts after 7 days); Lambda event sources are disabled and provisioned concurrency removed; Bedrock deletes provisioned throughput and may delete fine-tuned model weights. If a project remains paused for 90 days without action, AWS permanently deletes project data.