Microsoft Fabric is a Microsoft product that brings all analytics workloads, from data integration to data warehousing and from data science to Power BI reporting, together on a single SaaS platform. At the Microsoft Fabric Community Conference (FabCon Europe), held in Barcelona in October 2026, a large number of new features were announced for the platform. In this article, we first summarise the core idea behind Fabric and then the key announcements.
The core idea of Fabric: one platform, one data lake
At the heart of Fabric is OneLake , a single data lake for the entire organisation. Data is stored in the open Delta Lake format, so the same table can be used without copying for data engineering with Spark, warehouse queries with SQL and Power BI reporting. Licensing is not per user but based on purchased capacity (F SKUs).
FabCon Barcelona 2026: now generally available
- Fabric IQ comes to Microsoft Copilot Chat: Governed insights from Power BI can be used inside Copilot Chat, while existing report definitions and row-level security rules are preserved.
- Power BI agent experiences: Agents can create and edit semantic models and reports from requirements or from a screenshot of a report. Some sub-capabilities are still in preview.
- Fabric IQ MCP endpoint: Trusted Power BI insights are opened up to agent development tools outside Fabric.
- Fabric Runtime 2.0: A new runtime with Apache Spark 4.1, Delta Lake 4.2 and Python 3.13. Existing workspaces need to be migrated and tested.
- Mirroring for Google BigQuery: Continuous replication of BigQuery data into OneLake.
- F4096 and F8192 capacities: Larger capacity options for very heavy workloads.
New in preview
- Agent-assisted data engineering: An agent that creates and runs Spark notebooks based on a described goal (e.g. bronze, silver and gold layers). Microsoft positions it for evaluation purposes for now.
- Fabric IQ ontologies: Defining business concepts and the relationships between them by reusing existing Power BI measures.
- Pay-as-you-go billing and the F0 capacity: Aims to let you start without reserving capacity in advance. However, F0 does not mean free compute, and it is still expected to roll out.
- Fabric policies: More granular, attribute-based governance rules that go beyond tenant-wide settings.
What does this mean for companies?
The common theme of the announcements is connecting AI agents directly to the data platform. But whether these capabilities deliver reliable results depends on having a consistent data model and clearly defined business metrics underneath. Our recommendation for companies considering a move to Fabric: first tidy up your existing Power BI models and measure definitions, then proceed with a small capacity and a single use case. Preview features must always be tested before being used in production.
To review your Fabric and Power BI roadmap together, take a look at our data analytics and Power BI service pages.