SageMaker waste check
Previous Generation SageMaker Instance
previous_gen_sagemaker_instance
What it finds
Previous Generation SageMaker Instance: one of the SageMaker checks in the ml/ai optimization group. Checks in this group look for: machine learning endpoints and notebooks that are idle, stopped but still billing for storage, or larger than their workload needs.
Why it costs money
Hosted notebooks and endpoints bill for the instance hours and storage they hold, not for the models they serve.
How it is detected
Previous-gen instance type (ml.m4/c4/t2/r4/p2)
Detection confidence: HIGH. This is the criterion from our public docs. It runs in a local scan with your own read-only credentials.
AWS's own tools
None of Trusted Advisor, Cost Optimization Hub or Compute Optimizer flags this.
Checked against the AWS documentation.
Run it yourself
Nothing leaves your machine. The local scan has no telemetry.