We need to completely disable Serverless Compute in our Azure Databricks workspace to prevent unexpected compute charges.

Question

We need to completely disable Serverless Compute in our Azure Databricks workspace to prevent unexpected compute charges.

zybertech1sql 0

Please help us with the following:

Identify any features or services (such as AI/Genie, SQL, Jobs, notebooks, Model Serving, or other workspace features) that are currently triggering Serverless Compute.
Advise how to disable Serverless Compute at the workspace level, if supported.
If workspace-level disabling is not possible, provide the recommended configuration changes to ensure Serverless Compute is never used.
Confirm whether there are any hidden or default settings that automatically start Serverless Compute and explain how to disable them.
Recommend any workspace policies or best practices to prevent future Serverless usage.

Peram Venkateswara Reddy 255 Reputation points Microsoft External Staff Moderator

2026-08-05T11:12:56.6666667+00:00
Hi @zybertech1sql , I would like to add a few points to @Sina Salam Answer,

Based on the current Azure Databricks documentation, Serverless Compute for notebooks, workflows, and Lakeflow pipelines is available by default in most workspaces that have Unity Catalog enabled and are deployed in a supported region. Microsoft Learn does not currently document a self-service workspace-level setting to disable Serverless Compute.

To prevent unexpected Serverless-related charges, I recommend first identifying which workloads or features are consuming serverless resources. In addition to Serverless notebooks, jobs, and Lakeflow pipelines, several Azure Databricks features use serverless infrastructure independently, including Serverless SQL Warehouses, Model Serving, Data Quality Monitoring, Predictive Optimization, and certain AI features. These services have separate configuration paths and can generate usage independently of Serverless Compute for notebooks and jobs.

You may want to review the following:

Serverless notebooks and workflows currently in use.

Serverless SQL Warehouses that may still be running.

Model Serving endpoints.

Lakeflow pipelines configured to use serverless resources.

Data Quality Monitoring and Predictive Optimization, which can generate serverless billing records even when no serverless notebooks or jobs have been intentionally executed.

To minimize future usage, consider:

Migrating Serverless notebooks and jobs to customer-managed compute.

Removing or replacing Serverless SQL Warehouses where appropriate.

Reviewing permissions for creating jobs, warehouses, pipelines, and serving endpoints.

Implementing budget and governance controls to improve cost visibility and accountability.

If your objective is a complete disablement of Serverless capabilities, the current Microsoft Learn documentation does not describe a supported self-service method to disable Serverless Compute at the workspace level. In such cases, you would need to work with Databricks administrative/support channels to determine whether any account-level or backend controls are available for your specific environment.

Ref: https://learn.microsoft.com/en-us/azure/databricks/compute/serverless/

Kind Regards,
Microsoft Support Team.

2 answers

Your answer

Answer 1

Hello zybertech1sql,

Welcome to the Microsoft Q&A and thank you for posting your questions here.

I understand that you would like to disable Serverless Compute in our Azure Databricks workspace to prevent unexpected compute charges.

Serverless Compute is available by default in many Unity Catalog-enabled Azure Databricks workspaces, and several Databricks services such as Notebooks, Jobs, SQL Warehouses, Lakeflow Pipelines, and Model Serving can independently consume Serverless resources. As a result, traditional cluster policies and compute governance controls may not fully prevent Serverless usage. - https://learn.microsoft.com/en-us/azure/databricks/compute/serverless/, and https://docs.azure.cn/en-us/databricks/compute/serverless/

What you need to do is to: - https://community.databricks.com/t5/administration-architecture/how-do-you-disable-serverless-interactive-compute-for-all-users/td-p/102880

Identify the exact Serverless workload generating charges using Databricks Usage and Billing data.
Remove or migrate all Serverless SQL Warehouses.
Convert Serverless Jobs to customer-managed clusters.
Review Notebook compute selections.
Remove unused Model Serving Endpoints.
Review Lakeflow Pipelines configured for Serverless execution.
Restrict user permissions for creating Warehouses, Jobs, Pipelines, and Endpoints.
Implement Budget Policies for ongoing visibility and accountability.
If no Serverless disable option exists in the workspace, engage Databricks Support via your immediate Portal to verify whether Serverless can be disabled at the backend for that workspace or account. - https://learn.microsoft.com/en-us/azure/databricks/compute/serverless/, and https://learn.microsoft.com/en-us/azure/databricks/compute/serverless/dependencies gives more insight.

After identifying the actual Serverless consumer and removing all Serverless-enabled resources, future Serverless usage can be minimized. However, if the workspace is provisioned with Serverless enabled by platform design, Databricks Support is required to confirm whether complete disablement is supported for that environment.

I hope this is helpful. Please! Do not hesitate to let me know if you have any other questions, steps or clarifications.

Please do not close the thread by upvoting and accepting the answer if any part of it is helpful.

Answer 2

You need a workspace-wide guarantee that no serverless workload incurs charges. Azure Databricks documents no universal kill switch: notebooks, jobs, and pipelines use serverless by default in Unity Catalog workspaces, while SQL, Model Serving, and AI features have separate controls. New notebooks and supported job tasks can default to serverless; the SQL UI defaults new warehouses to serverless. Genie and dashboards use their assigned SQL warehouse, and Model Serving is serverless. Query system.billing.usage using billing_origin_product, identity_metadata, and usage_metadata to identify workloads; inspect data-quality monitoring, predictive optimization, materialized views, Lakeflow Connect, and AI Search. Then move jobs and pipelines to classic compute, attach notebooks to classic compute, delete serverless warehouses, revoke Databricks SQL access and unrestricted compute creation, and remove serving endpoints. Usage policies and standard budgets only attribute costs or alert; they do not block general serverless usage. For an opt-out, open Support because no documented self-service control guarantees it.