# SageMaker Studio Applications Without Idle Shutdown Canonical: https://www.pointfive.co/efficiency-hub/inefficiencies/sagemaker-studio-applications-without-idle-shutdown In SageMaker Studio, each running JupyterLab or Code Editor application runs on an instance that is billed for as long as the application is running,... By: PointFive Updated: 2026-09-28 [Cloud Efficiency Hub](https://www.pointfive.co/efficiency-hub) The short version In SageMaker Studio, each running JupyterLab or Code Editor application runs on an instance that is billed for as long as the application is running, whether or not the user is doing anything. PointFive Research Cloud cost research at PointFive AWS service [AWS SageMaker](https://www.pointfive.co/efficiency-hub/cloud-services/aws-sagemaker) Category [AI](https://www.pointfive.co/efficiency-hub/service-category/ai) Reference CER-0383 Type Inefficient Configuration ## Explanation Why the waste happens and who it affects. Data scientists routinely leave spaces running overnight, over weekends and after a project ends, often on large CPU or GPU instance types chosen for a single experiment. Closing the browser tab does not stop the application. Studio has a built-in idle shutdown feature, but it is not on by default and must be enabled separately for JupyterLab and for Code Editor, at the domain or user-profile level. It only works on SageMaker Distribution images v2.0 or newer (or custom images built on them), and when enabled from the console without a value it defaults to 10,080 minutes, a full week. Domains created before the feature, domains on older images with no auto-shutdown lifecycle configuration, and domains with week-long timeouts all leave idle instances billing. The AWS Well-Architected Machine Learning Lens lists implementing Studio idle shutdown under MLCOST04-BP08, Stop resources when not in use. ## Billing model The pricing dimensions that drive this cost. Studio itself is free to access, and charges come from the resources applications run on. Application instance JupyterLab and Code Editor applications are charged for the chosen instance type based on the duration of use, for as long as the application is running Space storage Each space's storage volume is billed per GB-month while the space exists, independent of whether an application is running Idle shutdown timer Starts only when the application becomes idle, and the application keeps billing until the configured timeout elapses ## How to detect 5 checks to find it in your estate. - Run describe-domain and describe-user-profile and check JupyterLabAppSettings and CodeEditorAppSettings for AppLifecycleManagement.IdleSettings; a missing block or LifecycleManagement not set to ENABLED means idle shutdown is off for that application type - Flag IdleTimeoutInMinutes values that are far longer than a working day, including the 10,080-minute console default, and user profiles that override a stricter domain setting - Check the SageMaker Distribution image version used by each space; applications on images older than v2.0 cannot use idle shutdown and need a lifecycle configuration instead - Use list-apps to find JupyterLab and CodeEditor apps in InService status, and correlate their instance types and uptime with user activity; apps running continuously across nights and weekends are the main target - Review SageMaker spend in Cost Explorer by usage type to see how much comes from Studio application instances, especially GPU instance types ## How to fix 5 ways to remove the waste. - Enable idle shutdown at the domain level for both JupyterLab and Code Editor with a timeout that matches how the team works, and restart running applications, since settings only take effect after a restart - To enforce a single timeout, set IdleTimeoutInMinutes, MinIdleTimeoutInMinutes and MaxIdleTimeoutInMinutes to the same value; to allow some user choice, set a bounded maximum instead of leaving the week-long default - Move spaces to SageMaker Distribution v2.0 or newer, or build custom images on it; where that is not possible, attach an auto-shutdown lifecycle configuration - Be aware of the idle definition: JupyterLab counts as idle only with no active kernel or terminal sessions, so abandoned kernels keep the instance running; encourage users to shut down kernels and commit work - Limit the instance types users can select for Studio applications and delete spaces that are no longer used, which also removes their storage charges ## Documentation Vendor references for pricing and configuration. - [Idle shutdown docs.aws.amazon.com](https://docs.aws.amazon.com/sagemaker/latest/dg/studio-updated-idle-shutdown.html) - [Set up idle shutdown docs.aws.amazon.com](https://docs.aws.amazon.com/sagemaker/latest/dg/studio-updated-idle-shutdown-setup.html) - [MLCOST04-BP08 Stop resources when not in use docs.aws.amazon.com](https://docs.aws.amazon.com/wellarchitected/latest/machine-learning-lens/mlcost04-bp08.html) - [Amazon SageMaker AI Pricing aws.amazon.com](https://aws.amazon.com/sagemaker/ai/pricing/) ## Related inefficiencies [Browse the library](https://www.pointfive.co/efficiency-hub) - AWS SageMaker CER-0333 ### [Idle SageMaker Notebook Instances Left Running Continuously](https://www.pointfive.co/efficiency-hub/inefficiencies/idle-sagemaker-notebook-instances-left-running-continuously) SageMaker notebook instances are billed continuously while in an active state - and critically, they do not automatically shut down when idle. Closing a browser tab, shutting down a Jupyter kernel, or simply walking away does not stop the... AI - AWS SageMaker CER-0384 ### [On-Demand SageMaker Training Jobs Without Managed Spot Training](https://www.pointfive.co/efficiency-hub/inefficiencies/on-demand-sagemaker-training-jobs-without-managed-spot-training) SageMaker training and hyperparameter tuning jobs run on on-demand instances unless managed spot training is explicitly enabled on the job. Many training workloads tolerate interruption: experiments, scheduled retraining, tuning sweeps and... AI - AWS SageMaker CER-0559 ### [Idle SageMaker Real-Time Inference Endpoint](https://www.pointfive.co/efficiency-hub/inefficiencies/idle-sagemaker-real-time-inference-endpoint) A SageMaker real-time inference endpoint keeps its instances running and billed for as long as the endpoint is in service, whether or not any application calls it. Endpoints are created for model evaluations, demos, A/B tests, proofs of... AI --- Source: the public page above. Product screenshots and illustrative interfaces are examples, not live customer data.