Microsoft Fabric October 2025 Feature Update Brings Native Graph, Adaptive File Sizing, and Secure Outbound Controls

Microsoft Fabric’s October 2025 update delivers a robust set of new features and upgrades, further enhancing its reputation as a leader in unified analytics, AI integration, and enterprise security. This month’s highlights include advanced security with Outbound Access Protection, streamlined data engineering with Adaptive Target File Size, and a new native graph service that redefines how organizations view and manipulate connected data.​

---

Events & Community Announcements

 

This October, Microsoft is focusing on engagement and professional growth:

• Fabric Data Days kick off November 4th, offering 50+ days of community sessions, learning, contests, and exam discounts—ideal for aspiring data professionals and experts alike.​ Register and Sign up for live sessions here.

• FabCon Returns March 16-20, 2026 in Atlanta, GA, featuring training, networking, and keynotes from the industry’s top leaders. Register with code FABCOMM to save $200. See FabCon Europe 2025.

---

Microsoft Fabric Platform: Developer Experience Upgrade

• Keyboard Shortcut Support: Developers get workflow acceleration through keyboard shortcuts for horizontal tabs and the object explorer, echoing modern IDE experiences.​

• 

  Focus Mode: A new mode allows users to maximize screen space by collapsing the navigator and object explorer, reducing distractions during code or query edits.​

---

Security: Outbound Access Protection & Private Links

• Outbound Access Protection for Spark (GA): Now organizations can govern outbound connections from Spark jobs to external and internal destinations, reducing risk of data exfiltration across all Fabric workloads.​ 

• Workspace-Level Private Link (GA): Fabric workspaces now support secure, private network connectivity. Critical workloads can run within corporate virtual networks without touching the public internet—ensuring regulatory compliance and isolation.​ 

---

OneLake: Better Access Controls

• 

  ReadWrite Access: OneLake security now allows precise control at the ReadWrite level, permitting users to write data in lakehouses (even as viewers) without over-scoping permissions. Teams can securely manage uploads and workflows—maintaining least-privilege principles.​

---

Data Engineering

• Adaptive Target File Size: This pivotal feature solves the “file size dilemma” by automatically adjusting Delta table file sizes using real-time telemetry. Smaller tables get appropriately sized files for query efficiency, while larger tables consolidate for performance and cost savings. Early results: 30% faster ELT cycles and up to 1.6x faster compaction phases.​ 

• Fast Optimize & Auto Compaction: File maintenance now happens without “write amplification.” New logic evaluates whether compaction jobs will meaningfully improve performance and triggers only when needed. Customers have seen up to 40% reduced Spark costs and 5x faster data pipelines after enabling auto-compaction.​ 

• Spark Connector for SQL Databases (Preview): Spark developers can now read/write securely to Azure SQL, SQL Server, and Fabric SQL—without complex installs. PySpark and Scala are fully supported, with authentication options for every scenario.​ 

• Spark Executor Rolling Logs: For complex jobs, log files now split into manageable hourly segments, making it easier to troubleshoot and download logs for specific windows.​ 

---

Lakehouse and Data Agent Usability

• Table Deep Links: Users can now generate direct URLs for any table in Lakehouse Explorer, fostering faster collaboration and instant data previews.​ 

• Data Agent Integration: Lakehouse users can seamlessly connect their data to Data Agents for instant AI-driven insights and multitasking navigation—streamlining the journey from raw data to actionable intelligence.​ 

---

Data Science: Enhanced Authoring Experience

• New UI for Data Agent Creators: A dual-tabbed interface (Data and Setup) allows for efficient configuration, iterative schema testing, and seamless change tracking with built-in Git/CI integration.​ 

• Markdown Editor: Authors can write instructions for agents using Markdown, making for clear, structured, and context-rich communications that large language models process more accurately.​ 

---

Data Warehouse: Smarter Data Operations

• JSONL Ingestion: OPENROWSET in Fabric Data Warehouse now supports easy import and querying of rich JSONL data, from flat files to deep nested structures, using T-SQL syntax.​ 

• Compaction Preemption: New logic prevents background storage optimization from clashing with user transactions, reducing query failures and ensuring smoother high-throughput workloads.​

• Relative File Paths: Data source integration gets easier with simplified pathing. Users create a named data source for the root directory and reference all files by relative paths—cleaning up queries and scripts.​

---

Real-Time Intelligence: Native Graph and Streaming

• Native Graph in Fabric (Preview): Fabric now includes a horizontally scalable, native graph analytics and visualization engine, inspired by LinkedIn’s architecture. This enables organizations to model and explore complex connections—perfect for powering next-gen AI agents and sophisticated data relationships.​ 

• Eventstream Enhancements:

  • MongoDB CDC Source: Real-time event ingestion from MongoDB deployments, including Atlas and on-premises, now streams directly into Eventstream.​

  • EventHub Schema Validation: Payloads from Azure EventHub are now validated against schemas using Fabric’s registry, eliminating the need for downstream code-based data checks. 

  • Derived Streams Pause/Resume: Operational flexibility allows for cost savings during downtime and fast resumptions when needed—crucial for business resiliency.​

---

Data Factory: Broader Format and Export Support

• Expanded Copy Job Formats: More file types, including ORC, Excel, Avro, and XML, can now move effortlessly within Fabric, improving compatibility.​ 

• CSV Handling: Enhanced CSV copying now supports nuanced controls like custom quote and escape characters, plus encoding options, for global data integration.​ 

• Variable Library Integration: Deeper Dataflow Gen2 and Copy Job support allows dynamic pipeline behaviors and better CI/CD management.

• Export Query Results (Preview): Power BI users can now export cleaned or transformed queries directly to OneLake, Dataflows, or other cloud stores—linking desktop analytics with enterprise-scale workflows in just a few clicks.
  

  
  

---

Building the Foundation for the Next Data Era

The October 2025 Microsoft Fabric update further positions Fabric as the go-to solution for secure, scalable, and AI-ready data workloads. By introducing foundational features like native graph analytics, adaptive automation, granular security, and seamless integration across tools, Microsoft is helping organizations unlock new insights, optimize workflows, and prepare for the future of intelligent data.​

For technical details and demos, see more information in the official feature summary on the Microsoft Fabric blog.