Databricks on AWS
1201 releases that have landed on AWS, newest first.
- AUTO CDC APIs replace APPLY CHANGES
Supports change data feeds in Spark Declarative Pipelines on Lakeflow. Replaces APPLY CHANGES APIs.
- Convert to Unity Catalog managed table from external table is in Public Preview
Uses ALTER TABLE SET MANAGED command. Requires participating customer status.
- Databricks Jobs is now Lakeflow Jobs
No migration is required.
- Databricks SQL Serverless engine upgrades
Upgrades bring up to 25% lower latency and add Predictive Query Execution and Photon vectorized shuffle.
- DLT is now Spark Declarative Pipelines on Lakeflow
No migration is required.
- Foreign Apache Iceberg tables are in Public Preview
Supports reading Iceberg tables from HMS, Glue, and Snowflake Horizon Catalog. Requires Lakehouse Federation.
- Run the change, do not just read it. Hands-on labs in your own Databricks workspace, graded when you submit. Browse labs
- Lakebase, a managed PostgreSQL OLTP database, is in Public Preview
Provides PostgreSQL-compatible storage and compute for transactional workloads.
- Managed Apache Iceberg tables are in Public Preview
Supports read and write from Databricks or external Iceberg engines. Integrates with predictive optimization.
- Microsoft SharePoint connector (Beta)
ingests SharePoint data into Databricks via API-based pipeline creation.
- Monitor and revoke personal access tokens in your account (GA)
Account admins can monitor and revoke PATs in the account console, replacing manual tracking methods, and OAuth tokens are recommended instead.
- Deployment jobs (Public Preview)
Automates tasks for new model versions, integrating with Unity Catalog models and Lakeflow Jobs.
- MLflow 3.0 is generally available
Delivers state-of-the-art experiment tracking and performance evaluation for machine learning models.
- Model Context Protocol (MCP) for AI agents is in Beta
Lets AI agents access tools and data via a consistent interface. Supports Databricks-hosted or custom servers.
- Serverless performance targets is now GA
Enables faster startup and execution time when performance optimized setting is used.
- Configure PrivateLink connections from serverless compute to S3 Storage buckets (Public Preview)
Enables private connectivity to S3 from serverless compute, requires AWS setup.
- Cross-platform view sharing is now GA
Replaces manual data sharing methods, updates billing and access controls. Adds system table to track shared data history.
- Manage network policies for serverless egress control (Generally Available)
Configures outbound access for serverless workloads, including SQL warehouses and model serving endpoints.
- Tag policies is in Beta
Enforces consistent tagging across data assets, defining allowed keys and values. Managed at account level.
- Get this table by email. One Monday mail covering the week, filtered to the clouds and products you run. Weekly digest
- Databricks Apps support for compliance security profile standards
Supports HIPAA, PCI-DSS, and FedRAMP Moderate standards.
- PrivateLink connections from serverless compute to resources in your VPC through an NLB is in Public Preview
Uses an NLB, requires AWS PrivateLink setup.
- Serverless GPU compute is in Beta
Supports custom single and multi-node deep learning workloads with AI Runtime.
- Consumer access entitlement is generally available
Grants limited workspace UI access and SQL warehouse querying.
- Dashboard scheduling and filtering improvements
Fixes issues with scheduling and cross-filtering on large datasets.
- Genie entity matching and SQL editor improvements
Genie now supports up to 60 entity matching columns. SQL editor no longer disappears when deleting generated queries.
- New consumer entitlement is generally available
Grants limited UI access and SQL querying. Replaces broader entitlements for business users.
- New region: AWS Asia Pacific (Jakarta) (ap-southeast-3)
Supports creating workspaces and products in ap-southeast-3.
- User interface updates
Emits listHistoryQueries and getHistoryQuery events on Query History page and query profile opens.
- Salesforce Data Cloud File Sharing connector (Public Preview)
Offers zero-copy querying via file federation, reading data directly from cloud object storage without JDBC.
- Corrected job_name values in system.billing.usage
Replaces task keys with user-provided names, except for one-time job runs.
- History sharing now enabled by default to improve table read performance for Databricks-to-Databricks Delta Sharing (GA)
Requires Databricks Runtime 16.2 and above.
- Mosaic AI Vector Search storage-optimized endpoints are Public Preview
Supports over 1 billion vectors at dimension 768, with 10-20x faster indexing.
- Unity Catalog HTTP connections support OAuth User-to-Machine Per User credentials (Public Preview)
Replaces shared OAuth credentials with per-user credentials.
- The sourceIpAddress field in audit logs will no longer include a port number Coming soon
Removes port number, which was always 0, for consistency with other logs.
- Dashboard catalog picker and filter fixes
Select catalog and schema from SQL editor, use table name only in queries. Fixes global filters on full-screen widgets.
- Genie knowledge store and metadata improvements
Adds popular query suggestions, table metadata improvements, and knowledge store features for building reliable Genie Agents.
- Metric views are in Public Preview
Requires a SQL warehouse on the 2025.16 Preview channel. Abstracts complex business logic into a centralized definition.
- User interface updates
New queries are created in Drafts folder by default, moving out when saved or renamed.
- Dashboard filter enhancements and visualization improvements
Multi-select filters now include exclude option, global filters apply across all pages.
- Genie usage guidance and query performance
Adds usage guidance and speeds up query result display.
- New alert version in Beta
Simplifies creating and managing alerts with a single interface, can be used alongside legacy alerts.