Fabric Weekly

Issue 3

22^nd July 2026

👋 Welcome back to Fabric Weekly! Apologies for the radio silence - issue #3 was due out last week, but a thoroughly unseasonal bout of COVID (in July, of all times) had other plans for me. The upside is a proper bumper edition: it covers a longer stretch than usual, so there's extra to get through.

On the official side:

• The headline is GPU-accelerated query execution for Fabric Data Warehouse - an informal-but-rigorous TPC-H benchmark showing roughly 6x the throughput at 1000 concurrent users versus CPU execution, with p95 latency dropping from 97 seconds to 15.
• Also worth a look: role-based access control for API for GraphQL lands in preview, with field- and row-level security baked into the API itself rather than a custom app layer.
• OneLake picked up a genuinely money-saving GA feature: lifecycle management with storage tiers, which automatically moves cold data into cheaper cool/cold tiers based on access patterns without you lifting a finger.
• And Fabric Data Factory pipelines got a batch of new capabilities, headlined by an AI-powered Operations Agent (preview) that watches pipeline health and suggests fixes, plus activity-level monitoring so you can drill straight from a failed run into the specific step that broke.

Two bigger stories worth their own mention this issue: Reza Rad sat down with Yitzhak Kesselman (CVP for Real-Time Intelligence at Microsoft Fabric) for a expansive conversation on where Fabric IQ and Real-Time Intelligence are heading - ontologies, ops agents, and the "no AI without RTI" argument, with real customer numbers to back it up. And Rayfin, aka Fabric Apps - Microsoft's new open-source SDK for building full-stack web apps entirely inside a Fabric workspace - TypeScript frontend, Fabric SQL database or Power BI semantic model backend, all under one governed roof.

In the community:

• Erwin de Kreuk's FMD framework can now generate column-level lineage from Fabric Lakehouses straight into Microsoft Purview.
• And a deep, honest write-up on getting governed Databricks data into Fabric behind a fully private network compares four setups and lands on the same conclusion every time: the hard part isn't moving the data, it's carrying the Unity Catalog policies with it.
• And Chris Webb finally sets out the actual benefits of Direct Lake mode beyond "it's the new shiny thing" - cheaper refresh, no memory limits, and reuse of the same OneLake tables across multiple semantic models.

And, over on the endjin blog: a blog from yours truly which compares star schema, flat table, and header/detail designs in DAX Studio.

🤖 Copilot, AI & Agents

• The future of conversational analytics in Fabric
  Fabric is integrating conversational analytics across Fabric IQ, Power BI, Microsoft 365 Copilot, and custom data agents to enable natural language queries on business data for actionable insights.
• Microsoft Copilot: What Each Copilot Does and Where It Fits
  James Serra untangles the sprawling Microsoft Copilot family, making clear it's not one AI assistant but many tailored experiences. The bulk of the post catalogues every Copilot and AI surface inside Fabric specifically - conversational analytics/data agents in Power BI, report authoring, semantic model web modeling, Data Engineering/Data Science notebooks, Data Factory, Data Warehouse/SQL analytics, SQL database in Fabric, Real-Time Intelligence, and Copilot Studio's connected-agent scenarios - before briefly covering Copilots outside Fabric (Microsoft 365 Chat, Outlook, Teams, Word, PowerPoint, Excel, GitHub Copilot) for contrast.
• We're Now Using AI to Help AI Understand Your Data
  A clip from the Power Mates' April Fabric feature summary: Vesa and Heini flag the irony that Fabric's newest AI features use AI to auto-generate metadata and descriptions for fields and models, so that other AI (Copilot, agents) can understand your data - after years of humans managing without a single description in place.
• Fabric Real-Time Intelligence & Fabric IQ — A Conversation with Yitzhak Kesselman | Fabric Insider Ep. 6
  Yitzhak Kesselman explains the transition from Fabric to Real-Time Intelligence and Fabric IQ, emphasizing their role in enabling timely event-driven analysis and actions across business operations.
• Liar, liar! What to do when AI writes your pipelines confidently wrong?
  Argues AI-written pipeline code fails differently to human code: not with obvious bugs, but "confidently wrong" — syntactically clean, passes review, and still wrong because the agent didn't know a dimension table was Type-2 SCD, silently duplicating transactions. Proposes a "specify, bound, validate" workflow: precise specs, guardrails on what the agent can touch, and validating AI-generated pipelines by interrogating the output data (row-count reconciliation, uniqueness checks) rather than just reading the code.

📊 Reporting & Insights

• What Is Fabric App or Rayfin and Why You Should Care
  Reza Rad explains Rayfin (aka Fabric Apps), Microsoft's open-source SDK/CLI for building full-stack web applications entirely inside a Fabric workspace: a TypeScript frontend backed by either a Fabric SQL database or a Power BI semantic model, hosted and authenticated (via Entra ID) inside Fabric itself. Covers the four starter templates (blank, Data App, To-Do full/simplified), the four-command workflow to scaffold, run and deploy a project, and why this turns Fabric from a data platform into an application platform - notably the Data App template, which lets you deliver a custom branded web experience over a trusted semantic model without a Power BI Pro license per viewer.
• Power BI: Let Users Choose Between Full and Abbreviated Numbers ⭐
  Learn how to empower Power BI users with a slicer, dynamic format strings, and DAX UDFs for toggling between full numbers and abbreviations, enhancing report flexibility.
• Using Fabric Operations Agents And Workspace Monitoring With Power BI
  The Operations Agent monitors Power BI for DAX query errors, providing detailed alerts with user info, timestamps, error messages, and model/report/visual IDs when configured within Workspace Monitoring.
• New YouTube Video: Azure Databricks & Microsoft Fabric – Mirroring, Semantic Models and Power BI
  The video demonstrates how Azure Databricks Mirroring integrates with Microsoft Fabric to automatically replicate data for use in Power BI reports, streamlining analytics workflows.
• Building an Apache Arrow Parser for Power BI’s ExecuteDaxQueries Endpoint
  This article describes developing an Apache Arrow IPC parser for Power BI's ExecuteDaxQueries endpoint to natively parse DAX query results in Power Query while maintaining data types and INFO functions.
• Dynamic formatting by hierarchy level with ISINSCOPE and ISATLEVEL
  This article demonstrates applying distinct formatting rules at each hierarchy level in Power BI using ISINSCOPE within measures and ISATLEVEL in visual calculations, showcasing year-, quarter-, and month-specific color logic for a matrix visual.
• The Benefits Of Using Direct Lake Mode In Power BI
  Chris Webb lays out the concrete reasons to use Direct Lake mode beyond novelty: near-instant, low-CU "refresh" (framing) versus costly Import refreshes; avoidance of Import mode's memory limits and refresh timeouts; reuse of the same OneLake tables across multiple semantic models with consistent OneLake security; and shared access to the same data from other Fabric engines like SQL endpoints and notebooks.
• Optimising DAX: Model Design Comparisons and Testing Tips
  Compares star schema, flat table, and header/detail model designs loaded into the same DAX Studio workshop: star schema wins on size and compression, flat tables compress reasonably well via repeated values, and header/detail is usually worst because the high-cardinality join key is stored twice. Also covers when calculated columns help (lower-cardinality bucketed values) and practical DAX Studio testing tips.
• Secret Behind Professional Dashboards
  A round-up of Power BI visual-design tips: conditional-formatting data bars inside a table, standard and custom tooltip pages (including a top-5-products tooltip), HTML card visuals for cleaner KPIs, and dynamic DAX format strings — aimed at making table/matrix visuals and dashboards look more polished.
• DP-700 training: Monitor semantic model refresh
  A DP-700 exam-prep walkthrough of Import-mode semantic model refresh in the OneLake catalog: scheduling refreshes (up to 8/day on Pro, 48/day or 30-minute intervals on PPU/Fabric capacity), timezone pitfalls with scheduled times, checking refresh history and failure details, and setting up failure notifications — noting Direct Query/Direct Lake don't need scheduled refresh at all.

⚙️ Data Engineering

• Why Does Fabric Have Two Kinds of Lakehouses?
  In Microsoft Fabric, two lakehouses exist due to the introduction of a new catalog for managing tables and schemas instead of restricting the older Hive-based service.
• Streamlining orchestration with smarter pipelines and deeper monitoring
  Fabric Data Factory adds four pipeline features: auto-populated base parameters for Notebook activities; L2 activity-level workspace monitoring so you can drill from a pipeline run into the specific failed activity; an AI-powered Operations Agent (preview) that watches pipeline health and suggests fixes; and a Refresh Materialized Lake View activity (preview) for triggering MLV refreshes directly from a pipeline.
• Intelligent streaming: Real-time game-day operations with Fabric Eventstreams, Spark and AI functions
  Fabric's Eventstreams, Spark, and AI Functions streamline real-time game-day operations by capturing guest requests, classifying them via AI, and routing actionable items to relevant staff for swift responses.
• Microsoft Fabric Updates: What's New in June 2026
  A monthly Fabric feature recap from The Power Mates (dual Microsoft MVPs Gastón Cruz and Alex Rostan), rounding up June 2026 updates for enterprise teams and highlighting their own Tenant Scan Fabric governance-audit tool. This is the full-length video behind the "Fabric Updates JUNE 2026" YouTube Shorts series also pulled in this week; the shorts are excluded here as duplicates.
• No Code Dataflow Gen 2 in Microsoft Fabric Using GitHub Copilot and Fabric Skills
  No Code Dataflow Gen 2 in Microsoft Fabric enables users to create comprehensive dataflows, including merges and fiscal date tables, by simply describing requirements in plain English, with AI tools automating the generation of M code.
• Stream Oracle database changes into Microsoft Fabric Eventstream (Preview)
  Introduces the new Oracle Database CDC source connector for Fabric Eventstream (Preview), built on Oracle LogMiner, which streams every insert/update/delete from Oracle 12c+ (including Autonomous Database) as structured before/after change events — enabling real-time routing to Eventhouse, Lakehouse, and Activator without a separate CDC stack.
• What’s New in Fabric Eventstreams — 2026 Q2 Edition
  Recaps Q2 2026 Fabric Eventstreams releases: GA Kafka and Service Bus connectors plus a new Oracle CDC connector, GA SQL operator with multi-destination fan-out and event-time windowing, a Preview Business Events publisher, Eventstream AI Skills for natural-language topology authoring, and in-canvas Activator rule configuration — illustrated through a retailer's real-time order, warehouse-monitoring, and clickstream scenarios.
• From Mirrored Database to Real-Time Action: A Fabric Eventstreams scenario
  Walks through the "row-to-action" pattern for Fabric: mirroring an operational database, enabling Delta Change Data Feed, and streaming those changes into Eventstream, then using SQL operators, Eventhouse, and Activator to detect and act on events like manufacturing defect clusters within roughly 90 seconds of the source row changing.
• Using Python Notebooks to Merge Data into a Microsoft Fabric Lakehouse
  A pattern for using Fabric's low-CU pure-Python notebook runtime (not PySpark) to incrementally upsert/merge API data into a Lakehouse Delta table: a composite key identifies rows for the merge, a watermark column tracks the last-loaded date for incremental pulls, and parameters (workspace/Lakehouse GUIDs) are kept out of hard-coded attachments so the notebook moves cleanly between workspaces.
• We Put a GPU Behind Fabric DW
  An independent (if informal) benchmark of Fabric Data Warehouse's new GPU-accelerated query execution (Preview) against 100GB TPC-H on an F64 trial capacity: single-user queries all complete under 500ms, but the real story is concurrency — GPU sustains ~100+ queries/sec at 1000 concurrent users versus CPU's ~17 q/s, with p95 latency of 15s versus 97s. Roughly 6x the throughput from the same capacity, one toggle to enable; the win is concentrated at high concurrency, not single-user queries.

🏛️ Storage & Platform

• Bring your Snowflake Iceberg tables to OneLake with Fabric mirroring
  Fabric mirroring enables seamless integration of Snowflake Iceberg tables into OneLake as Delta Lake for immediate querying without data duplication.
• Microsoft Fabric + SPARROW: Real-World Data Engineering
  A Power Mates walkthrough of a real-world Fabric case study: SPARROW (Solar-Powered Acoustic Remote Recording), a solar-powered wildlife acoustic sensor platform, feeding conservation data into OneLake via shortcuts, through Fabric pipelines, and out to real-time Power BI biodiversity dashboards. This is the full-length version; five shorter clips covering the same SPARROW story were also pulled in this week and are excluded here as duplicates.
• Controlling storage costs with OneLake lifecycle management (Generally Available)
  OneLake's new features automatically tier data across hot, cool, and cold storage based on access patterns, lowering long-term costs while preserving performance.
• Understand your data with the OneLake storage report (Generally Available)
  The OneLake storage report provides workspace admins with instant visibility into item-level storage usage for better data management in Microsoft Fabric.
• The Lakehouse Migration API Is Whitelist-Only (Here's How You Get In)
  Migrating a non-schema Fabric lakehouse to a schema-enabled one requires the Lakehouse Migration API, which is currently whitelist-only: Microsoft enables it per tenant after you submit the aka.ms/lakehouseswitch form, per a clipped conversation with Microsoft's Ted.
• What’s new for Dataverse Fabric Link: More control, stronger security, and fresher data
  Dataverse Fabric Link now lets admins choose which Dataverse/finance-and-operations tables sync into Fabric instead of linking everything, adds Workspace Identity and Service Principal authentication for production use (instead of relying on individual user credentials), and cuts sync latency from up to two hours down to under 15 minutes via low-latency sync in supported regions.
• OneLake architectural guidance: A practical blueprint for the AI Era
  The OneLake Architectural Guidance introduces five patterns to help enterprises modernize their fragmented data estates for AI and analytics, supported by real customer examples.
• I Roasted Fabric Error Messages to Microsoft's Face
  A pointed critique of confusing or unhelpful Fabric error messages, taken directly to Microsoft — an opinionated look at where the platform's error handling still falls short of a good developer experience.
• There's a Cosmos DB Hiding Inside Your Fabric Lakehouse
  Looks at how mirrored Cosmos DB data shows up inside a Fabric Lakehouse as Delta tables in OneLake, and what that means for querying Cosmos DB data alongside other lakehouse sources.
• I Asked Microsoft Why Schema-Enabled Lakehouses Took So Long
  Puts the question of why schema support for Fabric Lakehouses took so long to ship directly to Microsoft, digging into the architectural tradeoffs behind adding schemas to the lakehouse storage layer.
• Getting governed Databricks data into Fabric behind a private network
  A hands-on comparison of four ways to bring governed Databricks/Unity Catalog data into Fabric behind a fully private network: mirroring Unity Catalog (GA, zero-copy, read-only), storing directly in OneLake (Beta), storing plus publishing metadata (Beta/Preview), and shortcuts or copy/ingest (GA). The deciding factor is governance — Unity Catalog row filters and column masks don't travel automatically, and Policy Weaver is needed to translate them into OneLake data access roles.

🔐 Governance & Security

• Governing the Flow: A Beginner’s Guide to Microsoft Fabric
  Governing the Flow details how Microsoft Fabric's governance pillars impact speed versus stability in data workflows.
• Securing Outbound AI Connectivity in SQL Server 2025 (Generally Available)
  SQL Server 2025 enhances secure outbound AI connectivity with governed REST invocations and external model objects, preserving data trust and ensuring security, governance, and auditability.
• Fabric Metadata Driven Framework July 2026 Release Notes
  The July 2026 release of the Fabric Metadata Driven Framework introduces enhanced lineage integration, automated SQL connection creation using Service Principals, and improved Key Vault support for secure credential management.
• The Auditor’s Toolkit: Reporting on Your Tenant
  A PowerShell/Fabric CLI script using the Power BI module to audit workspace access across an entire tenant in one pass — pulling every workspace, user, and access-right combination into a single CSV for compliance and security review, instead of manually clicking through the Admin Portal.
• Announcing role-based access control for API for GraphQL in Microsoft Fabric (Preview)
  Adds role-based access control to Fabric's API for GraphQL (Preview): API owners can define roles with separate query (read) and mutation (write) permissions, plus field-level and row-level security to restrict which parts of the schema and which records a role can see — configured directly on the GraphQL item rather than in a custom application layer.
• Microsoft Fabric Workspace Topology Patterns
  James Broome walks through six escalating workspace-topology patterns for Fabric — single workspace, Dev/QA/Prod separation, splitting data engineering from reporting, use-case/domain bounding via Fabric Domains, separating storage from compute, and full medallion-layer splits — with the tipping point that pushes a team from one pattern to the next, and a recommended starting combination (environment separation + DE/reporting split + domain bounding) for most production deployments.
• Bringing Column-Level Lineage from Microsoft Fabric to Microsoft Purview with FMD
  The open-source Fabric Metadata-Driven Framework (FMD) now automatically extracts column-level lineage from Fabric Lakehouse transformations and publishes it into Microsoft Purview, closing a gap where Purview previously only supported item-level lineage for Fabric. The core lineage-extraction logic is also available as a standalone notebook for teams not using FMD.

💰 Management & Cost

• Remove Hardcoded Workspace and Item IDs with Fabric Variable Libraries to use Across Dev/Test/Prod
  Gilbert Quevauvilliers walks through using Fabric Variable Libraries to strip hardcoded workspace and item GUIDs out of a pipeline and notebook, using a worked example (a pipeline + notebook + Lakehouse pulling public holiday data) that deploys cleanly from dev to prod just by pointing a same-named Variable Library at the new workspace's IDs, with no code changes required.
• Use User Data Functions to securely call Fabric REST APIs
  User Data Functions in Azure Fabric offer a secure method to invoke REST APIs by centralizing authentication and authorization in reusable functions, minimizing direct credential embedding while ensuring explicit access control via service principals and Azure Key Vault.
• VNET Data Gateway Autoscaling (Preview)
  VNET Data Gateway Autoscaling (Preview) dynamically adjusts capacity based on demand, reducing costs and improving reliability for Microsoft Fabric workloads while maintaining security and isolation.
• Reassigning Microsoft Fabric Capacities in Bulk: Don’t Let an Expiring Trial Catch You by Surprise
  Automate bulk reassignment of Microsoft Fabric workspaces with the Fabric CLI to prevent item loss when trial capacities expire, facilitating seamless transitions for development, testing, or cost optimization.
• The Fabric Blueprint: Architecting Workspaces for Enterprise Success
  An internal-standards style guide for Fabric governance: environment-based (Dev/Test/Prod) workspace separation, domain-based workspace grouping, shared vs. project-specific workspaces, Git-integrated CI/CD with deployment rules, Entra ID security groups over individual permissions, naming conventions, and capacity segregation to avoid noisy-neighbour issues.
• How To Deploy Fabric SQL and Azure SQL Databases with Azure DevOps
  Fabric's built-in CI/CD refuses schema changes that could lose data (column length changes, dropped columns/tables). This walks through an alternative: an Azure DevOps build pipeline that extracts a DACPAC via SqlPackage, and a release pipeline that publishes it through QA/Production with explicit switches (DropObjectsNotInSource, BlockOnPossibleDataLoss, etc.) — working identically for Fabric SQL Database and Azure SQL Database since they share the same engine.
• Terraform with Microsoft Fabric: Where to Draw the Line
  [Video, no transcript available] Weighs which parts of a Fabric estate are sensible to manage via Terraform versus which should stay in Fabric's native deployment tooling, based on the title and the presenter's Fabric-focused channel.
• Creating Microsoft Fabric Sandbox Capacities Chatbot
  [Provisional — source content unreachable despite two fetch attempts] Builds a chatbot using Azure Logic Apps to interact with Microsoft Fabric sandbox/trial capacities. This summary is based on the title only; please flag if the full article becomes accessible so it can be rewritten properly.