Microsoft Fabric is evolving rapidly with a wide range of preview capabilities that span storage, governance, data engineering, analytics, and Artificial Intelligence experiences. OneLake gains deeper interoperability through Apache Iceberg access, Azure Blob Storage shortcuts, centralized catalog governance, granular data access roles, fine-grained OneLake security, and Iceberg-compatible table APIs. Data access scenarios are expanded through Azure Monitor to Eventhouse routing, workspace outbound access protection, OneLake data in Excel, and cross-cloud integrations such as Microsoft Entra service principal authentication for Amazon S3 shortcuts, plus mirroring support for SAP, Google BigQuery, Cosmos DB, SQL Server 2025, and Azure Databricks catalogs, including workspaces behind private endpoints.
Across Data Factory and data engineering, new preview features focus on automation, performance, and lifecycle management. Copy job now supports change data capture, incremental copy across more sources, truncate and query-based copies, and multi-folder scenarios, with dbt jobs, Apache Airflow job file management APIs, and Spark Job Definition and Notebook activities that authenticate via service principal or workspace identity. Lakehouse lifecycle management gains git integration and deployment pipelines, while Fabric Runtime 2.0 with Spark 4.0 and Delta Lake 4.0, variable libraries in Dataflow Gen2 with CI/CD, partitioned compute, the modern evaluator, Livy REST API, Spark diagnostics, and JobInsight all work to improve developer productivity and operational visibility. Additional previews such as Fabric Connection inside Notebook, the Microsoft JDBC driver, Upsert actions in the Lakehouse connector, and a Spark connector for SQL databases extend integration options for Java, Spark, and SQL workloads.
Artificial Intelligence and Copilot experiences are being woven throughout Fabric, from Copilot for Data Warehouse chat, SQL analytics endpoints, Dataflow Gen2, notebooks, and Real-Time Intelligence, to a global “Copilot in Fabric” rollout that covers Power BI, Data Factory, Data Science, Data Engineering, and KQL authoring. Fabric AI Functions are now generally available, offering functions like ai.embed(), ai.analyze_sentiment, ai.extract, ai.generate_response, and ai.summarize with advanced configuration options and support for Azure OpenAI and Microsoft Foundry models such as Claude and LLaMA. New semantic and agent-oriented capabilities arrive through Fabric IQ and its ontology item, Fabric data agents integrated with Microsoft Copilot Studio and Azure AI Agent Service, prebuilt Azure AI services, OpenAI-powered plugins for Eventhouse, Artificial Intelligence-powered anomaly detection, and Notebook Copilot inline code completion. At the same time, foundational platform and governance enhancements such as workspace-level IP firewall rules, workspace-level surge protection, Fabric identities scaling from 1,000 to 10,000 identities, security insights in the OneLake catalog, SQL database auditing, customer-managed keys, and warehouse SQL audit logs strengthen security and compliance while reshaping how semantic models, warehouses, and streaming intelligence are governed and deployed, including the retirement of default Power BI semantic models by November 30, 2025.
