Databricks Releases Serverless Compute Access Controls (GA) and Rate Limits (Beta), Shifting Cost Governance from Clusters to Workloads

Databricks announced in a blog post on October 29, 2026, that its serverless cost controls have expanded to notebooks, jobs, and pipelines. Previously, this functionality was available for serverless SQL and Model Serving, but it is now unified across broader serverless compute scenarios.

The update includes three core pillars: cost attribution, access control, and spend limiting. Serverless Compute Access Controls is now Generally Available (GA), allowing administrators to govern who can run serverless workloads using 'Can Use' and 'Can Manage' permission objects. By default, all users have usage rights, but admins can remove the 'All Users' group to enforce stricter controls.

Rate Limits are currently in Beta, designed to prevent runaway spending. This feature sets maximum autoscaling ceilings for automated workloads using t-shirt sizes (S, M, L, XL), thereby bounding how much or how fast a job or pipeline can consume. Additionally, Serverless Compute Tags enable cost attribution, associating each DBU consumption with specific teams or projects, queryable via the system.billing.usage system tables.

Databricks stated that as serverless adoption grew more than threefold in the past year, platform teams increasingly demanded better visibility and budget control. These new tools aim to replace traditional cluster policies involving node counts and instance types, shifting governance focus from infrastructure to people and specific workloads.