We are excited to announce Cosmos DB in Microsoft Fabric is now in preview. Cosmos DB in Fabric makes it easy to build agentic AI apps, offering an AI-optimized database that is automatically configured to meet your application's needs. It's built on the high availability, dynamic scaling, and AI-ready capabilities of Azure Cosmos DB.
Building on the momentum from the launch of SQL database in Fabric, we are expanding databases workload in Fabric with this new addition. You can now store semi-structured NoSQL data in Cosmos DB in Fabric, alongside your relational data in SQL databases, enabling a unified data platform for your applications. This further positions Fabric as a complete data platform to handle all your organizational needs, from operational to analytics and BI.
Your existing or new applications can instantly benefit from deep integration with Fabric OneLake, bringing databases, analytics, data science, real-time intelligence, and Copilot-powered BI in one place, rather than assembling them individually.
We built Cosmos DB on three key themes, to make it the database of choice for AI apps, without overhead: simplified, autonomous, and optimized for AI.
Simplified and Autonomous
We are removing the database management overhead with the right defaults, so you can focus on building your applications.Simple to get started
You can get started with Cosmos DB database with just a name. Complex settings such as networking, backup, or scale are not required. The database starts within seconds, with optimal settings for your apps, by default.You can also easily upload data to your databases with just a couple of clicks in familiar data explorer. Exploring with sample data and connecting your apps is just a click away. If you want to import custom data, you can easily do so through JSON files or bulk inserts through SDK. We are working on adding data ingestion through pipelines in future.
You can also query the transactional data directly from the database using the familiar SQL queries.
Automatic scaling with unified CU usage reporting
Cosmos DB in Fabric is automatically configured with the right scale settings for most of your app needs. Currently Cosmos DB usage is not billed in Fabric. In future, scale usage will be reported into Fabric as Consumption Units (CUs) for unified billing. More details on billing will be shared later.Secure by default
Cosmos DB in Fabric is configured with the right security settings from the start. Primary authentication is based on Microsoft Entra IDs. Workspace permissions are automatically inherited and enforced on data plane requests to Cosmos DB.Cosmos DB supports tenant-level Private Links for network security. Workspace-level private links will be supported in the future.
Simple application connectivity
Apps or clients, in Azure or Fabric, can connect to the Fabric database with the database endpoint, using existing SDKs. Supported authentication methods are Microsoft Entra IDs of the users logged into Fabric, with sufficient permissions to the workspace where the database is created. Apps can also connect to the database with Service Principal.Databases created in Fabric can be managed programmatically through REST APIs, and we are working on adding Application lifecycle management.
Account keys are not supported for now and any clients that need key-based connections will not be supported. Connectivity from Azure Functions, Azure Databases extension for VS code, and other Microsoft services will work if the services allow Entra IDs or managed identity or Service Principal for authentication.
Cosmos DB in Fabric supports NoSQL API for now.
Automatically available in OneLake
Cosmos DB data is automatically made available in OneLake in Delta Parquet format. This allows you to seamlessly join Cosmos DB data in Fabric with all other data in OneLake, such as other databases like SQL databases, Warehouses, Lakehouses, mirrored or shortcuts data from various sources. You can also perform cross-container, cross-database queries within or across Cosmos DB databases.Your data is also instantly made available to analytics and BI in Fabric. You can write back the data analyzed in Spark notebooks into Cosmos DB in Fabric, for low-latency millisecond serving through your applications; also known as, reverse-ETL.