Microsoft Fabric Updates & Release Notes

As announced at FabCon 2026 in Barcelona, we're releasing meaningful enhancements to the Capacity Metrics app in preview, designed to accelerate time to insight. This includes new and enhanced visuals and metrics across the health, compute, and storage pages—alongside improved navigation and a refreshed user interface design. A new item details page builds on the existing item history page, providing key insights for individual items running on the capacity.

The aim is to support key user journeys, including:

• Identifying issues as they happen.
• Reviewing issues that have occurred over the last two weeks.
• Identifying risks before they result in throttling.

Improved experiences

An admin can now land on the health page and quickly identify capacities most in need of review. New metrics introduce easily understandable concepts like percentage of time spent in throttling, which can be compared across capacities and time periods.

Figure: Health summary page displaying capacity health statistics for various services like Health, Compute, Storage, and Autoscale for Spark, with columns for different time periods, utilization, and various types of throttling – providing the capacity admin a view across capacities to help them keep operations smooth.

Once a capacity is identified for further review, the new compute page provides rich detail on the health state of the capacity and the biggest items impacting the capacity for the selected period. Key new visuals include:

• An updated utilization graph which expresses capacity unit (CU) consumption in CU seconds rather than a percentage, making it easy to see if your capacity has been scaled up or down, while also showing the current throttling state through conditional formatting.
• A new heat map visual that shows the capacities most utilized/overutilized state for each hour, making it easy to spot daily or hourly patterns.
• A new CU consumption table, which shows the item impacting the time window being reviewed (to a granularity of 1 hour). The visual shows smoothed CU consumption, meaning this truly reflects the highest impact items for the time being reviewed.

Figure: Compute dashboard with various metrics related to Fabric Capacity, including smoothed and unsmoothed usage percentages, time paused, and time no throttling, as well as different types of rejections and surge protection, all within a specified date range.

Once an item is identified as a possible cause of high utilization, or if the user just wants to see additional detail on the item, the new item details page provides a 14-day history. The page has been designed from the ground up to give detailed insights which were previously unavailable. One key visual is the new CU utilization chart which shows the impact the selected item had on the overall capacity.

Figure: Item details page including a column chart shows the impact the selected item is having on the capacity as a whole, aggregated CU usage across different fields such as users, and detailed information on consumption for the selected item.

Given the new experiences include changes to the existing metrics, it is being released as a preview alongside the existing app generally available experience, allowing time for you to adapt to the new app while still having access to the existing version. When you install the new version of the app, you will see the new preview as an additional option alongside the existing generally available versions of the reports.

Getting started

To support adoption of the new experience, explore the following resources for each new page:

• Existing Fabric Capacity Metrics app users will be notified when an update is available. Select Get it from the notification to update your current installation or install the latest version in a new workspace. Alternatively, you can install the app directly from the Microsoft Marketplace.
• Technical documentation with embedded video guidance:
  • Health page
  • Compute page
  • Storage page
  • Item details page
• User journey documentation, which explains how the app can be used in different situations.

We're excited to put these new capabilities into your hands and look forward to your feedback as we continue evolving the Capacity Metrics experience to help you better understand, manage, and optimize your Fabric capacities.

Original source

Custom SQL Pools in Microsoft Fabric Data Warehouse are now generally available, along with statement type classification.

Custom SQL Pools give administrators more control over how compute resources are allocated across different workloads in a Fabric workspace. You can isolate workloads into separate pools, set maximum resource allocation percentages, and help ensure that one workload doesn’t consume resources intended for another.

For example, you can create separate pools for data ingestion, reporting, and ad hoc analysis, then configure the maximum percentage of compute resources each pool can consume. Custom SQL Pools complement the autonomous workload-management capabilities built into Fabric Data Warehouse while giving organizations additional control for predictable and business-critical workloads.

Classify workloads by statement type

Statement type classification provides a simple way to separate read and write workloads. When configuring Custom SQL Pools, you can classify distributed SQL requests into two categories:

• SELECT for read queries.
• NONSELECT for statements such as data loading and modification operations.

Fabric creates separate pools for the two statement categories, with an initial maximum allocation of 50% for each pool. Administrators can adjust these percentages based on workload requirements, provided their combined maximum allocation doesn’t exceed 100%.

Statement type classification is useful when organizations want to protect interactive reporting capacity from intensive ingestion or data-transformation workloads without relying on application names or connection properties.

Custom SQL Pools also support application-based classification using an exact application name or a regular expression. Administrators can select the classification model that best reflects how workloads connect to and use their Fabric workspace.

Get started

To learn about Custom SQL Pools, supported classifiers, resource allocation, and workload behavior, review the Custom SQL Pools documentation.

For instructions to create and configure pools in the Fabric portal, review the Configure Custom SQL Pools in the Fabric Portal documentation.

Original source

Fabric App functions is a new capability that allows builders to add secure, server-side logic directly to their Fabric Apps without standing up separate infrastructure. Built on top of Fabric User Data Functions (UDFs), Fabric App functions give developers a simple way to execute business logic, integrate with external services, access protected resources, and keep secrets from the client.

What are Fabric App functions?

Fabric App functions are TypeScript-based serverless functions that run alongside your Fabric app. They provide a secure backend execution environment for code that should not run in the browser. Functions are packaged and deployed as part of the application experience and can be invoked directly from the application through a typed client.

Think of Functions as the custom logic layer for your Fabric App:

• Execute business rules and complex calculations
• Access protected data sources
• Call external APIs
• Perform orchestration and AI workflows
• Store and retrieve secrets securely
• Access downstream resources using the application’s identity via application authentication

Application authentication

Application authentication means the Fabric app function accesses a downstream resource using the application’s identity rather than impersonating the person currently using the app. The downstream service sees and authorizes the application, not the signed-in user.

How the identity flow works

The user signs in to the Fabric App, the app calls the function, and the function then calls the downstream service using the application identity.

• Grant the application only the permissions it needs.
• Authorize the application at the downstream service.
• Keep credentials and tokens out of client-side code.
• Add application-level validation and business logic before accessing data.

Secrets management

Functions allow builders to securely store and retrieve secrets without exposing them to browsers or application users.

Get started with functions

Fabric App functions are designed to feel like part of the application, not a separate platform. Key dependency is to install Azure Functions Core Tools version 4 before you get started.

npm install --global azure-functions-core-tools@4

Create an app with functions

Scaffold a new app, move to the app directory, and initialize its functions project:

npm create @microsoft/rayfin@latest -- "my-app"
cd my-app
npx rayfin functions init

The Fabric App functions init command adds the functions project, TypeScript configuration, and a starter function file to the existing app.

Build the app with GitHub Copilot

The scaffold includes the GitHub Copilot instructions and skills needed to work with the project. Open the project in Visual Studio Code, then describe the app you want to build in GitHub Copilot Chat.

For example, use this prompt to build an AI pull request review assistant:

Build an app that lets me view all pull requests for the repository <repoURL>, chat with my Azure AI Foundry agent <agentName> to review pull requests, and post review comments. Use functions with application authentication to connect to Azure AI Foundry and Azure DevOps.

GitHub Copilot can generate and update the required files and components. You can also ask the agent to run the app, debug functions locally, and deploy the app.

Debug functions

You can run functions locally to test your server-side logic before deploying to Fabric. Use the local Azure functions host to invoke functions, inspect errors, and validate your integration. Once validated, deploy the app and functions together. Run the following command to test your functions and deploy your changes.

npx rayfin dev
npx rayfin up

Conclusion

If you’ve been waiting for a secure way to connect Fabric Apps to external services, protect secrets, use application authentication, or build AI-powered experiences, Fabric App functions are the missing piece.

To learn more, refer to the Fabric Apps documentation.

Original source

If you haven't already, check out FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and Agents for a complete look at all of our FabCon Europe announcements across Fabric and Microsoft Databases.

Planning in Microsoft Fabric IQ is redefining what modern enterprise performance management looks like. Organizations can build budgets, forecasts, targets, and scenario models directly on top of governed Fabric data and shared semantics — on the Power BI semantic models they already trust. Planning is no longer a separate system that data is copied into, and results copied back out of.

That shift matters most in the era of AI. As organizations move from traditional applications to AI-powered, multi-agent systems, the advantage is shifting beyond the model you deploy. It now lies in the context an agent can reason over: how the business runs, the state it is in today, and where it is trying to get to. The first two are well served. The third, intent, usually lived somewhere else. Planning in Microsoft Fabric is different. The plan and the business intent behind it become a natural extension of how the business already measures performance and drives action. This ability to plan where the data lives has clearly resonated with customers — the growth and adoption of Fabric Planning since launch has been very humbling.

When Planning in Microsoft Fabric became generally available in July 2026, we said we would extend it toward an agentic future: because plans, forecasts, targets, and actuals live together on Fabric's shared semantic foundation, Fabric IQ's ontology and data agents can reason over the goals your teams are working toward, not only the results they have delivered. That is the difference between explaining variance after the quarter closes and course-correcting while it is still open.

Today we are taking the next step towards that future. We are announcing new capabilities that make planning more streamlined, more scalable, simpler to set up, easier to manage, and most importantly, more deeply integrated with Fabric IQ.

From semantic model to plan in a few clicks

We’ve made it simpler to get started with planning — instead of configuring shared cloud connections to the semantic model; planners can now connect to a semantic model directly by using their signed-in identity.

Create a Plan artifact, choose a semantic model, connect, and start planning.

Existing permissions and data access controls continue to apply, so nothing new has to be secured. For a finance team opening a new forecast cycle, that means less setup before they can start shaping the numbers — and for the organization, a faster path from trusted data to a working plan.

Native integration with Fabric IQ

Fabric IQ gives an organization a single, shared understanding of its business. OneLake and semantic models capture what happened. Real-Time Intelligence captures what is happening. Planning captures what should happen — the forward-looking layer of Fabric IQ.

We are introducing a new capability which streamlines the integration with Fabric IQ so that the plan can more easily become part of that layer. Copilot, data agents and operations agents then have the ability to reason over historical performance, current operations, and future intent.

The new writeback configuration experience makes it easier for planners to add budgets, forecasts, and scenarios back into the same semantic model the plan is already built on. Analytics and AI can then reason over more than just operational signals and historical KPIs — the foundation for intelligent action based on both actual performance and business intent.

Incorporating the plan into an ontology lets the organization represent planning concepts as business entities, properties and relationships, giving targets, forecasts and assumptions richer organizational context.

This enables a new class of data-aware and intent-aware agents —capable of analyzing variance, evaluating scenarios, and surfacing recommendations that align with business priorities. Together, they close the loop between what has happened, what the business expects to happen, and what it is trying to achieve.

"For us, the biggest benefit of Fabric Planning is speed: it lets us turn the semantic models and trusted business logic we have already invested in into planning applications, rather than rebuilding that foundation in a separate planning platform. And because plans sit in the same Fabric context as actuals, they can also give Copilot and agents the forward-looking business context to reason not only about what happened, but about what we expect and plan to happen next," said Patrick Sura, Director Business Intelligence EMEA, Dentsu.

Plan at enterprise scale with the new Native Planning Engine (Preview)

The new Native Planning Engine, is built on Rust and Apache Arrow with tight OneLake integration and delivers up to 10x improvements in performance and scale. Teams can work across tens of millions of cells, with dependent calculations updating as inputs change.

That responsiveness is what changes the work. Planners can input and allocate across large, detailed business models, connect price and quantity through to revenue, cost, margin and profit, and keep complex plans connected across the enterprise — without splitting the model into smaller pieces to keep it workable.

"At Bosch Mobility, strategic planning starts with trusted data. Microsoft Fabric IQ Plan provides the scale and integration we need to connect complex market data across our business, creating the foundation for better decisions and future AI-driven insights," said Robert Finger, VP of Business Digitization, Bosch Mobility.

Plan driven automation – From planning decision to business action

A plan rarely stops when a value is submitted. Often a dependent downstream process needs to run, or a notification needs to be sent. Event triggers in Fabric Planning let planners configure what happens when a writeback event occurs.

Event triggers turn every planning update into action, automatically kicking off downstream processes when writeback starts, succeeds or fails.

• Orchestrate business processes with Fabric pipelines, from consolidation to allocation.
• Transform and distribute planning data with Dataflow Gen2.
• Refresh semantic models to bring updated plans into analytics.
• Run notebooks for forecasting, optimization and machine learning.
• Keep teams informed through automated Teams notifications and Outlook email.

This feature goes beyond simply starting a downstream process. Writeback details and execution metadata can be passed directly as parameters, giving downstream actions in the context of the planning event that triggered them. And with multiple triggers per planning sheet, teams can create different automation paths for different events, connecting planning changes to the right response across Fabric.

Planning becomes more than capturing a decision. It becomes the trigger that puts the wider Fabric ecosystem in motion.

Planning in organizational apps — One app for the business workflow

A plan is rarely reviewed on its own. The people approving a forecast also want the reports and analysis that explain it, and that usually means sending them to several different places.

Planning and Intelligence Sheets can now be embedded in an organizational app alongside Power BI reports and related analytics. A department manager opens one application link, reviews performance, reaches the plan, and continues the workflow — with navigation curated around the business activity rather than the workspace it happens to live in. Existing app audiences and access rules govern distribution, so leadership, managers and finance each see the content curated for them, wherever they open the app, including in Teams and SharePoint.

Advanced driver modelling — See the full impact of a decision

New multi-driver, multi-measure simulations help teams evaluate a decision as a connected set of trade-offs. A sales plan might assume higher volume and a lower selling price; operations can then test the material, labor and overtime implications and see how the combination changes profit. In Matrix or Tree View, changes flow through the configured driver relationships and relevant hierarchy detail, making the downstream impact visible. Teams can create scenarios such as Best Case and compare alternatives before committing a forecast. The result is a more informed decision: leaders see both the proposed action and its financial consequences in the same planning model.

More ways to model and deliver planning

The latest experience also expands how teams build and consume plans:

• Custom fiscal year planning supports calendars whose periods do not begin in January.
• Stepped, Outline and Tree layouts let teams shape the planning grid around the way their business works.
• Advanced lookups bring XLookup into the grid and enable PowerTable lookups into Fabric SQL.
• The goal is flexibility without fragmentation — different planning methods, one connected Fabric foundation.

"Fabric Planning lets us tailor forecasting to each area of the business — top down, bottom up, driver-based, or assumption-based — without forcing everyone into the same methodology. By building on validated semantic models and capturing assumptions, commentary, and the story with the numbers, we shorten the development cycle, bring multiple perspectives together quickly, and gain deeper insight without the inefficiency of emails, meetings, and reconciliation," said Andrew Gundrum, Executive Director, Finance Data and Analytics, Merck.

Billing and Governance Improvements

Recently, we announced a change to the billing for non-planning editors, which means for PowerTable-only data management, editors are billed as Stakeholders and consumers as Viewers, so you get full data-management value without paying for a Planner session.

Now also in preview are enhancements that provide administrators with new tenant-level settings to ensure you can put the right planning experience in the right hands. Administrators can control who can upgrade to Planner or Stakeholder sessions and can warn users before an action is likely to oversubscribe capacity. This ensures greater control over the amount of capacity consumption which can be incurred by planning sessions.

What comes next

The ambition is an AI-native planning experience — one where agents do not only read the plan, but take part in it.

The next step is an API surface for the planning lifecycle, so an agent can work inside a plan the way a planner does: draft a forecast from what the model already knows, run a simulation across drivers, compare scenarios before a commitment is made, write the result back, and start the downstream work it sets in motion.

Get started

With Planning in Microsoft Fabric IQ, organizations bring planning, analytics, reporting and AI together on a single governed platform. From one input to enterprise-wide action, Planning connects the decision, the data, and the work that follows.

Intent no longer has to live somewhere else. It sits in the model the business already trusts, beside the results it will be measured against — where the people running the business, and the agents working alongside them, can act on it together.

Learn more about Planning in Fabric IQ:

• Read the documentation
• Learn more about Fabric IQ

Original source

If you haven’t already, check out Arun Ulag’s hero blog “FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and Agents” for a complete look at all of our FabCon and SQLCon announcements across both Fabric and our database offerings.

Getting a data project working is only the beginning. As AI makes it easier for more people to build applications, explore information, and create agents, the harder questions arrive sooner: can the solution be deployed consistently, can a team explain what changed, and will someone notice when a dependency fails? Developers need room to experiment, but they also need a dependable route from an isolated feature branch to a production workload that others can operate. Manual setup, disconnected monitoring, and unclear data context make that route harder to repeat.

Administrators face the other side of that growth. More workspaces, users, and workloads mean more decisions about access, compute, spending, and accountability. A capacity spike can affect an unrelated team; a broad security setting may not fit every project. Admins need controls that reflect organizational requirements without turning each development step into a permission request. Developers, in turn, need those boundaries to be understandable and built into their workflows, rather than discovered just before release.

Today, we're announcing Microsoft Fabric platform updates to help you tackle both sides of that challenge. They connect operational signals with investigation, make delivery more repeatable, and bring governed discovery into the tools you use. We're also expanding your options for managing compute demand and applying controls across your estate.

Introducing observability in Fabric (Preview)

A failed job rarely tells the whole story. With the introduction of observability in Fabric, you can connect signals across jobs, workspaces, items, and capacities rather than investigate each in isolation. A Monitor item acts as a single management unit across capacities and your tenant, bringing early detection, contextual investigation, and central management into one operational experience.

Unified workspace monitoring (Preview)

We're introducing unified workspace monitoring so you can choose which Eventhouse receives your telemetry and consolidate multiple workspaces into that shared store. Instead of monitoring workspaces individually, you can bring related signals together and manage their storage centrally. You control retention and caching policies to suit your investigation and operational needs. When a failure crosses workspace boundaries, for example, you can correlate the telemetry in one place and follow the issue across those boundaries. Most Fabric items will be available at launch, with additional workloads becoming available over time.

The new experience also simplifies how you connect monitoring to your own tools. A custom endpoint on the monitoring Event Stream lets you copy data to external or custom destinations, while public APIs support integration with your external observability tools. Simplified, API-backed onboarding makes setup repeatable for automation and agents. You can establish a shared monitoring approach, then extend it to additional workspaces without treating every setup as a separate manual exercise.

Refreshed Monitor Hub (Preview)

The refreshed Monitor Hub helps you see where attention is needed before opening individual investigations. We're adding ready-made, at-scale views across workloads, jobs, apps, agents, alerts, and capacities, with a centralized health overview and data health monitoring. That broader view helps you spot emerging issues and consider their downstream impact rather than starting with a single failed activity.

Capacity consumption and available headroom sit alongside operational signals, so you can consider resource pressure as you assess an issue. Start with the health overview, narrow your focus to the affected area, and move into investigation with that context. The hub connects navigation and operational visibility; the capacity administration actions described later let you respond when the next step is a control change.

Agentic investigation and alerting with operations agents and Activator

We're also adding the ability to use operations agents to investigate end-to-end causes, develop evidence-backed hypotheses, assess impact, and recommend next steps. You can engage with the investigation, refine its direction, and choose your response. This brings the supporting evidence and relevant dependencies into the conversation instead of leaving you to assemble every clue yourself. Currently this capability can help you investigate pipeline failures, with further support planned.

New Activator-powered job alerts notify you about successes as well as failures. You can automate alerts and configure responses, connecting investigation with the operational workflows your team uses. Together, these enhancements help you move from noticing an event to understanding its likely impact and deciding what to do next.

Repeatable delivery and governed discovery for developers

Operational confidence starts before deployment. The latest enhancements, designed with developers in mind, help you package a solution, review changes, move definitions between environments, and find reusable data worth reusing with fewer manual handoffs between those steps.

Introducing deployment plans (Preview)

With the deployment plans now in preview as first-class workspace items, you can define the exact deployment order on a visual canvas, with YAML as the source of truth. You can include pre- and post-deployment steps, such as hydrating a lakehouse with a notebook, running a pipeline before dependent items, or validating data after deployment.

Your plan follows the solution through Git branch-out, initial sync, updates, deployment pipelines, and bulk import. Preparation and validation become part of the deployment sequence rather than a separate checklist.

Bulk Item Definition APIs, quotas, and API limits (Generally Available and Preview)

Bulk import and export item definition APIs are also now generally available. You can move multiple definitions in one call; imports automatically determine deployment order and update references for the destination workspace. This supports workspace-as-code, CI/CD, and migration.

We're also enhancing the Fabric API platform with a unified quota model and increased limits for core APIs. These updates provide more consistent throttling and quota management while supporting higher-scale automation, CI/CD, and enterprise workloads, making it easier to build and operate Fabric solutions at scale.

Git integration enhancements (Generally Available and Preview)

The new branch workspace administration delegation capability, now in preview, gives developers a governed, self-service way to create branch workspaces without permission to create workspaces or assign capacity. Administrators define the branch workspace admin profile, including the role, admins, capacity, and delegated settings applied to each new branch workspace, helping teams move faster while maintaining consistent guardrails.

We're also expanding the Git workflow with file-level commits for supported items (Preview) and contributor branch switching (Preview). You can commit only the files that are ready while leaving unfinished changes in the workspace, and contributors can switch the Git branch connected to a workspace when administrators enable the capability. Additionally, selective branching, change comparison and commits, and branched workspace relationships, now generally available, helping teams keep feature work focused, review changes before acting, and manage branching from a clearer source control experience.

Discover data in OneLake catalog and your coding workflow

The OneLake catalog gives you a richer, more granular way to explore and understand data across Fabric before you use it. With the latest updates, you can go beyond finding Fabric items to explore individual tables and columns, review descriptions, and preview the underlying data directly in the catalog. Quick actions, personalized views, and filters, including domains and tags, add business context so you can evaluate whether data is right for your needs before you start working with it (Generally Available).

We've also expanded the OneLake catalog search API to return relevance-ranked Fabric items, OneLake tables from lakehouses, mirrored sources, and semantic model tables in one result set (Preview). This brings the same granular discovery to programmatic search, helping applications find relevant data across Fabric regardless of where it lives.

OneLake catalog discovery in GitHub Copilot CLI is now generally available through Skills for Fabric. You can find Fabric items and tables from the CLI before writing code, bringing this richer, governed discovery experience directly into your coding workflow rather than requiring a separate search.

Bring catalog discovery into Excel and Microsoft Foundry

Finally, OneLake catalog in Excel is now in preview, enabling you to discover and analyze Fabric data in your workbooks without copying or moving the underlying source data. OneLake row- and column-level security and Microsoft Purview label-based protections apply, connecting familiar analysis with the protections associated with your source.

New OneLake catalog additions in Microsoft Foundry bring enterprise data discovery into your AI development workflow (Preview). With an embedded catalog and a unified view across Fabric and Foundry, you can find trusted data and add it to knowledge, agents, and workflows.

Effectively administer capacities and spending

As your solutions reach production, capacity administrators need to optimize spend for steady demand while seamlessly managing bursts. With today’s announcements, we are adding new billing choices and capacity controls, as well as tools that help you investigate usage, automate responses, and act across capacities.

Introducing on-demand compute and zero-provisioned capacity (Preview)

As data and AI workloads grow, organizations need flexible ways to align costs with actual usage. Building on Autoscale Billing for Spark, we're introducing on-demand billing across Microsoft Fabric in preview soon, enabling workloads to scale elastically when needed, minimizing throttling during demand spikes, and allowing you to pay only for the compute you consume.

We're also introducing Fabric zero-provisioned (F0), a new entry point to Fabric and OneLake that removes the need to provision compute before getting started. With on-demand billing enabled by default, organizations can immediately access Fabric services and pay only for the resources they use. Together, on-demand billing and Fabric zero-provisioned make it easier to start small, scale as demand grows, and combine consumption-based pricing with reserved or provisioned capacity to optimize both performance and cost. These capabilities will be available in the coming weeks.

Whether you're running workloads that tend to spike, experimenting with new projects, or supporting production applications, on-demand billing gives you greater flexibility without requiring upfront capacity planning.

Surge protection, overage, and larger capacities (Generally Available)

Workspace-level surge protection will be generally available soon. You can set a workspace consumption limit, automatically isolating runaway or non-critical workloads so they don't impact mission-critical workloads or cause capacity-wide throttling. You can manage these controls in OneLake catalog or Monitor Hub, which also allows you to mark mission-critical workspaces exempt from the limit.

Additionally, capacity overage is now generally available, letting you use additional paid compute with configured spending thresholds and notifications rather than sizing capacity solely for rare peaks. For sustained high demand, we've added the F4096 and F8192 capacity SKUs (also Generally Available), extending your options for larger workloads.

Monitor, act, and automate your capacities

Monitoring helps you understand who's consuming resources, identify demand spikes, and ensure critical workloads continue to perform as expected. New capacity metrics app enhancements help you investigate health-state duration, capacity units per second, daily and hourly heatmaps, and top item contributors (Generally Available). Clearer drilldowns support deeper analysis, while chargeback app integration helps you allocate usage.

Capacity overview events in Real-Time Hub (Generally Available), with new operation events (Preview) adding workspace- and item-level visibility. These signals enable proactive capacity monitoring by helping administrators identify emerging hotspots, usage anomalies, and resource-intensive operations before they impact users. Connect them to Real-Time Intelligence, Eventhouse, and Activator to automate alerts and actions, retain history for capacity planning, or integrate insights into your existing monitoring systems.

With the refreshed Monitor Hub, we're taking the first step toward bringing capacity insights and management actions together in a single experience. Capacity Insights & Actions (preview) lets administrators identify emerging issues and immediately respond by adjusting surge and overage settings, migrating workspaces, or resizing capacity. Use Monitor Hub for monitoring and operational decisions, and the capacity metrics app for deeper usage analysis.

Unify governance, security, and control in the OneLake catalog

As AI helps more people build with data, tenant admins are responsible for governing an estate that’s growing in both size and complexity. More workspaces, data items, and access requests mean more decisions about what people can share, which settings they can change, and how to protect sensitive data. Admins need to support new projects without turning every decision into a manual review, or losing visibility into how data is used.

That balance between enabling teams and maintaining the right guardrails is critical. As Jayakumar Pankajatchan, Distinguished Engineer at KPMG, shared, “Client trust starts with security. Every engagement needs clear data boundaries and strong guardrails, so information flows only where intended. With Microsoft Fabric, we strengthen governance across engagements while enabling teams to deliver insights with confidence at scale.”

We’re expanding governance and security in Fabric to help organizations put those guardrails into practice without slowing teams down. With more centralized management, AI-assisted governance guidance, granular policies, and broader network protection, you can give your teams room to innovate while maintaining controls that reflect your organization’s requirements.

Introducing policies in Fabric (Preview)

With policies in Fabric (Preview), we're introducing precise, attribute-based controls that help organizations enforce governance requirements based on user and data attributes, rather than relying solely on broad settings. Policies enable granular controls for key scenarios, with item creation and workspace security supported at launch and external data sharing coming soon.

A centralized policy management experience, monitoring, APIs, and CI/CD integration support policy lifecycle management at scale, supporting more consistent and automated governance across your organization.

"Fabric’s Item Creation Policies have strengthened our governance controls for Microsoft Fabric, allowing us to scale innovation across the organization by empowering users with the right capabilities while maintaining clear boundaries for platform management and compliance,” said Bernard Beeftink Lead Product Owner, BI, ABN AMRO.

Centralizing governance in the OneLake catalog

We're expanding the OneLake catalog Govern experience (Generally Available) with centralized governance capabilities that bring management of tenants, domains, capacities, workspaces, tags, and policies into a unified experience. Rich contextual signals, insights, and recommendations help administrators better understand their governance posture and take informed action across their data estate.

We're also introducing Govern Skills as part of skills-for-fabric. Govern Skills provide AI-powered guidance across the health, security, and trustworthiness of your data. Govern Skills help administrators identify governance risks, monitor the health of their data estate, and take recommended actions to maintain a secure, trusted, and well-governed foundation for analytics and AI.

Outbound access protection expansion (Generally Available and Preview)

We've expanded outbound access protection to OneLake shortcuts (Generally Available), along with new coverage for maps in Fabric and operations agents (Preview). You can extend your outbound controls across these experiences while building on existing foundations such as Microsoft Entra, Private Link, customer-managed keys, and workspace IP firewall rules.

FQDN allowlist for outbound protected workspaces (Preview)

FQDN support for Spark simplifies secure connectivity for organizations using OAP-enforced Fabric workspaces, where Private Endpoints are often required to access dependent services. Instead of creating and managing Private Endpoints for every resource a Spark workload needs to reach, administrators can use domain-based network controls to securely enable connectivity through approved FQDNs. This reduces networking complexity, lowers operational overhead, and accelerates onboarding while maintaining enterprise security and compliance requirements. This capability will be available in the coming weeks.

Fabric availability in Microsoft 365 Government Community Cloud High (GCC High)

We are also expanding Microsoft Fabric to Microsoft 365 Government Community Cloud High (GCC High) customers. Support for GCC High extends Fabric to organizations that require U.S. sovereign cloud environments and compliance with frameworks such as DoD Impact Level 4, CJIS, DFARS, and NIST 800-171.

Put the updates to work

Start with the friction you encounter most: investigating failures, reproducing deployments, finding data, or managing demand and controls. Our updates connect those responsibilities without requiring every team to use the same workflow. Match adoption to your readiness requirements, then extend your approach as your data estate grows.

In addition to the announcements above, we are also rolling out a broad set of innovation across Fabric. For a broad overview of all these announcements, read Arun Ulag’s hero blog “FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and Agents.” If you want a more detailed look, visit the Fabric September 2026 feature summary blog, the Power BI September 2026 feature summary blog, and the latest posts on the Fabric Updates channel and the Azure Data Tech Community blog channel.

Original source

If you haven't already, check out FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and Agents for a complete look at all of our FabCon Europe announcements across Fabric and Microsoft Databases.

Fabric Apps is the application layer built on Microsoft Fabric. At FabCon Europe, it gains backend functions, storage, a family of data connectors, PostgreSQL, built-in app metrics, and a clearer enterprise access model.

Our industry has gotten really good at helping people visualize data. With semantic models and reports, we can deliver trusted insights that create shared understanding. But how do those insights become something people can act on? How do they become part of the products and workflows where decisions happen?

That last mile needs an application: a place to turn an insight into a recommendation, make a decision, take action, and keep a durable record of it.

Traditionally, building that application has meant a lot of work: setting up databases and authentication, copying data, and going through lengthy security reviews. As a result, insights often get stuck in the report.

Agents have solved part of this problem. They’ve made front-end applications dramatically easier to create. But the backend (data, auth, security, and sharing) remains difficult.

We want to change that.

That’s why my team built Fabric Apps: an application layer on Microsoft Fabric that lets you build secure, data-rich applications directly on the data, semantic models, and services you already trust. Since launching in June 2026, more than 15,000 developers have built a Fabric App. Check out the gallery. The number is encouraging. But what I find even more exciting is who has been building.

We have seen a new group of “builders” creating Fabric Apps. Designers are shipping experiences they used to hand off. Product managers are testing ideas in a working application rather than a slide. Finance teams are building tools for the workflows they know better than anyone.

And today at FabCon Europe, we’re thrilled to announce that we are expanding Fabric Apps with backend functions, built-in storage, and direct connections to even more of your Fabric data, so these new builders can go even further.

Rayfin: The development framework behind Fabric Apps

Rayfin is the code-first framework that powers Fabric Apps.

You or your coding agent describe an application in TypeScript. Rayfin's SDK and CLI generate typed APIs and clients, run the whole application locally, and deploy it as a Fabric App, where database, storage, connectors, functions, and authentication are running and governed for you.

No infrastructure tickets, and no credentials scattered across environments. Your data stays where it is.

Build on the data you already have

Fabric data connectors

Fabric Apps now connects directly to Fabric SQL Databases, Warehouses, Lakehouse SQL analytics endpoints, and semantic models. Users don’t want to copy data into an application. Copies add cost, complicate lineage, and make every governance review harder. SQL Database and Warehouse support read and write; Lakehouse SQL analytics endpoints and semantic models are read-only. Schemas are discovered automatically and exposed through the same typed programming model you use for app-owned data.

Declare the source in rayfin.yml:

connectors:
  - name: inventory
    type: fabric-warehouse
    config: { workspaceId: ${WS_ID}, itemId: ${ID} }
    auth: { type: delegated }
  - name: salesModel
    type: fabric-semanticmodel
    config: { workspaceId: ${WS_ID}, itemId: ${ID} }
    auth: { type: delegated }

Then use it from application code:

// Tables: query entities through a typed, fluent API.
const products = await client.connectors
  .inventory.Product
  .select(['name', 'sku'])
  .execute();

// Semantic models: query the model directly with DAX.
const revenue = await client.connectors
  .salesModel.executeQuery({ query: 'EVALUATE TOPN(10, Sales)', });

The second example is worth a closer look. Your revenue, target, and margin definitions already exist as measures that the business has debated, corrected, and agreed on. Rather than reimplementing that logic in JavaScript, an application can query the model and get the same answer everyone else gets.

Identity travels with the query. With auth: { type: delegated }, the connector runs as the signed-in user and the source enforces its own authorization, including row-level security. Application identity is available when an app legitimately needs its own permissions, but it does not inherit a user's permission boundary. Choose deliberately.

Give your application a place to write

PostgreSQL support, alongside SQL Server

A dashboard is read-only. An application needs somewhere to store the note someone captured, the decision made, and the plan that was approved.

Fabric Apps now supports PostgreSQL alongside SQL Server, using the same data-modeling experience, generated GraphQL APIs, typed clients, CLI, and local development workflow. Define entities in rayfin/data and register them in the schema; rayfin up provisions the service and applies the schema.

services:
  data:
    enabled: true
    dialect: mssql # or postgresql

const record = await client.data
  .CustomerFeedback.create(formValues);
const saved = await client.data
  .CustomerFeedback.findById(record.id);

The schema ships with the application as code, so there are no connection strings to manage and no migration scripts living on someone's desktop.

Keep documents next to the data

Storage

Records are structured, but the work around them usually arrives as attachments: a transcript, a signed PDF, an image someone needs to review before approving anything.

Storage is defined in code exactly like your database: folders, accepted file types, size limits, and permissions in TypeScript. The record goes in the database, the document goes in storage, and the record keeps a reference to it.

services:
  data:
    enabled: true
  storage:
    enabled: true

const { path } = await client.storage
  .documents.upload(file);
await client.data.CustomerFeedback.create({
  ...formValues,
  documentName: file.name,
  documentPath: path,
});

Write backend logic

Functions and Secret Store

Most applications eventually reach the edge of the standard building blocks, whether that's a business rule, a third-party integration, a multi-step workflow, or an AI call that has to run outside the browser.

Functions let you write server-side logic in TypeScript alongside your application code, while Fabric handles hosting, authentication, deployment, security, and scaling. Code lives in rayfin/functions and ships with the app on rayfin up.

services:
  functions:
    enabled: true

const result = await client.functions
  .generateProductIdeas.invoke({
    brief: 'Opportunities for the next quarter',
    context, // measures from a connector + records from app data
  });

Behind a single typed call, a function can query a semantic model through a connector, read application data, apply your business rules, and call a Microsoft Foundry model deployment through an AzureAI connection. Functions run under the app identity and can forward signed-in user context, so downstream calls can use either application or delegated authentication.

Secret Store keeps the API key out of your application entirely. It encrypts credentials and provides them to functions only at runtime, so they don't need to live in application code, source control, or a build pipeline.

Understand how your application performs

Built-in app metrics

Every Fabric App now includes usage and performance metrics out of the box: sign-ins, app loads, data-query volume, query errors, error rates, and average query duration.

Developers and application owners shouldn't have to stand up a separate telemetry service to understand adoption or to notice that a query slowed down after last week's change.

Start with a template

Universal App template and GitHub Copilot plugin

Install the Rayfin plugin and your coding agent gets guidance matched to the SDK version it is actually building against, instead of an API shape it half-remembers. With the new Universal App template, an agent can turn a prompt into a Fabric-authenticated, deployable project with very little setup.

Generated code is a starting point, not a production specification. Review it, check the permissions it requests, test the workflow, and put it through the same scrutiny as anything else you ship.

Your entire application, in code

Your screens, data schema, authentication configuration, functions, and data connectors are all defined in source code, in a single project.

That property is what makes a coding agent useful rather than risky. Typed entities make field names and call shapes explicit instead of inferred, so mismatches surface while you are still developing. The agent works from the application's actual model rather than a guess at it.

Type safety is not a substitute for runtime validation, authorization, or code review. All three still matter.

Built to be governed

Applications built with Fabric Apps are private by default. Microsoft Entra authentication happens before any application content is delivered, rather than as a redirect inside a page that has already loaded.

Builders then choose private or embedded-only access, or public access for approved anonymous scenarios. Tenant-level controls let administrators govern how applications can be exposed at all. We've also made sharing across supported Fabric items clearer.

One distinction matters more than any other:

Authentication determines who can open the application. Fabric permissions determine which data they can reach inside it. When a connector uses delegated identity, that second answer is the signed-in user's own permissions, enforced at the source. Fabric Apps keeps the two questions separate by design.

What's coming next

Our plans for later in 2026 include an admin and security portal, integration into Microsoft Copilot, app insights with every application, a Kusto data connector, app sharing with provenance, support for custom claims, OneLake-integrated application storage, embedding, and deployment pipelines.

We're early in this work, and there is a lot more to come.

Getting started

You can begin from wherever you work today:

• GitHub Copilot, install the Rayfin plugin and scaffold from the Universal App template.
• The Fabric portal, create a new App item in your workspace.
• Replit, our featured partner, directly in the browser.

I'd suggest starting with a single workflow that currently begins in a dashboard and ends in a spreadsheet and then building the ending.

In my experience that workflow is usually a decision someone makes every week, records informally, and then reconstructs from scratch the following month. It is also usually a small application, and usually best built by the person who runs the workflow rather than by whoever has room on an engineering roadmap.

Please tell us how it goes. What worked, and what got in your way? What do you need next? You can file issues in the Rayfin GitHub repository, and my team will read them.

If you're joining us in Barcelona, Ben Zulauf and I are presenting From Prompt to Production: Building Governed Apps on Fabric with Rayfin. Our team will also be at the Ask the Experts booths in the expo hall throughout the conference. Share your apps and feel free to bring a workflow you'd like to improve. We'd be glad to look at it with you.

Original source

Authors: Faisal Mohamood and Bogdan Crivat

If you haven't already, check out FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and Agents for a complete look at all of our FabCon Europe announcements across Fabric and Microsoft Databases.

Most enterprise analytics estates were assembled rather than designed. A tool for ingestion, another for transformation, a warehouse for data processing, a lake for everything else, and a layer of glue holding it together. That architecture did its job for a decade, but it was built for a world where data moved on a schedule and insight could wait.

AI changed the deadline. The organizations pulling ahead today are the ones consolidating on a single platform where data, analytics, and AI share one foundation instead of one integration backlog.

For teams running Azure Synapse Analytics and Azure Data Factory, Microsoft Fabric is the next step in that investment—not a departure from it. Fabric unifies data integration, engineering, and warehousing in one SaaS experience, reducing the operational burden of disparate tools and creating a clearer path to AI-driven workloads. The question customers now ask are not whether to modernize, but how to do so without disrupting the pipelines their businesses rely on every day.

Upgrade to Fabric Data Factory with confidence

Azure Data Factory has earned the trust of customers worldwide by powering mission-critical data integration workloads. Organizations that rely on Azure Data Factory and Azure Synapse pipelines have built critical data integration foundations that form the backbone of their business operations.

Fabric Data Factory builds on that trusted foundation while bringing customers into a unified data platform for AI transformation. With OneLake, Copilot-powered experiences, integrated analytics, and Fabric-native innovation, customers can modernize their data integration estate while preserving the investments that matter most.

At FabCon Barcelona, we are excited to announce a streamlined upgrade experience that helps Azure Data Factory customers begin their journey to Fabric Data Factory with confidence.

This new experience enables Azure Data Factory customers to bring their existing data factory into Microsoft Fabric, view their Azure Data Factory pipelines directly within Fabric, and continue using familiar management and monitoring experiences, all before making any migration decisions. Customers can experience Fabric immediately while preserving their existing Azure Data Factory investments. This is another important step toward our vision that every Azure Data Factory customer can become a Fabric Data Factory customer.

The experience is built around three principles we consistently heard from customers: transparency, control, and confidence. Customers want to understand what will change, what will remain the same, and when they should modernize. We provide a guided path that allows organizations to explore, assess, validate, and upgrade on their own terms.

The journey begins directly from Azure Data Factory with a simple view in Fabric experience. Customers can bring their Azure Data Factory environment into a Fabric workspace and immediately see their familiar data factory assets and pipelines within the Fabric experience. Existing Azure Data Factory assets remain the source of truth, while customers gain access to Fabric as a natural extension of their current experience. This allows teams to become familiar with Fabric, explore the broader Fabric platform, and understand the value of a unified data estate without disrupting production workloads.

Once inside Fabric, customers can evaluate their readiness for Fabric Data Factory using built-in migration assessment capabilities. The assessment provides visibility into pipeline and activity compatibility, helping customers understand which workloads are ready to upgrade, which may require review, and which capabilities are planned for future support. By surfacing compatibility information upfront, customers can make informed modernization decisions and prioritize migration efforts based on business needs. Connection mapping and migration guidance further simplify planning and reduce manual effort. This assessment-first approach is a direct response to customer feedback. Trust is built through visibility, and customers consistently told us they want a clear understanding of readiness before committing to change.

Figure: Seamless upgrade experience from Azure Data Factory to Fabric Data Factory.

After reviewing assessment results, organizations can choose when and how they upgrade to Fabric Data Factory. Supported pipelines can be upgraded incrementally, allowing teams to validate outcomes, gain operational confidence, and modernize at the pace that makes sense for their business. Existing Azure Data Factory workloads can continue operating while organizations evaluate and transition workloads over time. This staged approach helps reduce risk while enabling customers to realize the benefits of Fabric Data Factory and the broader Microsoft Fabric platform.

The upgrade journey from ADF to FDF allows customers to plan, explore Fabric, assess readiness, validate outcomes, and upgrade when ready. The journey from Azure Data Factory to Fabric Data Factory is designed to be transparent, low-risk, and customer-driven, helping every Azure Data Factory customer become a Fabric Data Factory customer on their own terms.

Guided migration with Microsoft-approved partners

Microsoft-approved partners can guide customers’ upgrade from start to finish, or customers can explore the upgrade independently with no effect on their existing Azure Data Factory pipelines. Customers can now request partner-guided upgrade support online and get connected with Microsoft partner expertise to help plan and execute their Fabric Data Factory upgrade.

Partners can help assess the customer’s environment, prioritize workloads, address compatibility requirements, develop the upgrade plan, execute and validate upgrades, and accelerate deployment to production.

Figure: Guided migration by Microsoft-approved partners to move your Azure Data Factory to Fabric Data Factory.

Modernizing Spark workloads with Fabric Data Engineering

Fabric Data Engineering provides a unified foundation for building and scaling Apache Spark workloads alongside the rest of the analytics estate. With capabilities such as the Native Execution Engine, high concurrency, and starter pools, teams can improve performance, reduce operational overhead, and work more easily with shared data in OneLake.

The Spark in Azure Synapse to Fabric Data Engineering migration assistant accelerates modernization by migrating Spark pools, notebooks, and Spark job definitions, while mapping lake databases through OneLake catalog shortcuts. Artifacts are copied while data remains in place, allowing existing pipelines to continue running without downtime. A guided experience tracks progress and provides a detailed report of what migrated successfully and what requires attention.

OBOS, one of the Nordic region’s leading residential developers, used a phased approach to migrate more than 600 notebooks and pipelines and consolidate over 4,500 data objects into 87 lakehouses. By validating technical results and live business KPIs across both environments before cutover, OBOS reduced operational costs by 20% and accelerated data processing by 30%.

“We deliberately ran the old and new environments in parallel, comparing row and column counts and validating business KPIs through live reports before we switched anything over. That rigorous validation mattered, because for a data platform like ours, trust is the product.” - Dag Erlend Berger, Product Lead for the Data Platform, OBOS

Complementing the migration assistant, the Fabric data engineering agent (Project Osmos), now in preview, helps accelerate the engineering work that goes beyond moving artifacts. Teams can delegate complex migration and modernization tasks, such as adapting Spark notebooks, refactoring pipelines, and validating migrated workloads. Osmos inspects the relevant data and code, plans and executes the work in Fabric, and iteratively validates and refines the results against the task’s requirements. Authorized teammates can contribute context and steer the same persistent task, with Fabric permissions governing access. This helps reduce manual engineering effort while keeping quality, maintainability, and team oversight central to the transition.

Accelerating query performance with Fabric Data Warehouse

For organizations managing data volumes in terabytes or petabytes, Fabric Data Warehouse provides superior performance and cost efficiency compared to Synapse dedicated SQL pools.

The migration assistant for data warehouse simplifies and automates this transition for existing Synapse customers. The assistant, built directly into the Fabric experience, provides a guided path to convert artifacts from Synapse dedicated SQL pools into Fabric Data Warehouse, helping with schema conversion and data movement alike. It delivers a clear, actionable snapshot of migration results alongside a detailed summary of items that need attention, and Copilot is available to help resolve issues on the spot.

To further reduce migration effort, the migration assistant gives customers the flexibility to migrate only what they need. Customers can select specific schemas or individual objects instead of migrating an entire environment, with visibility into object dependencies and guidance on related artifacts that should be migrated together. This allows organizations to break large modernization projects into manageable phases while maintaining confidence that dependent objects are accounted for.

Large-scale modernization programs are inherently iterative. Source systems continue to evolve while migration projects are underway, creating a challenge for organizations that need to keep target environments synchronized without repeatedly restarting migration efforts. To address this, we introduced incremental migration support in Migration Assistant. Customers can return to an existing migrated warehouse at any time and apply new changes from their source system, whether they are using a live connection or uploaded SQL files. The experience intelligently identifies differences, helps resolve conflicts, and updates only what has changed, transforming migration from a disruptive, start-over exercise into a repeatable, low-friction process. This significantly reduces the operational burden on migration teams and makes it easier for organizations to move to Fabric with greater confidence.

Figure: End-to-end Migration experience using migration assistant for Fabric data warehouse

Dentsu, a global advertising and public relations company, was dealing with data growing faster than what its teams could keep up. In partnership with OneLake GmbH, Dentsu accelerated its migration to Fabric, applying deep technical expertise to design a scalable, enterprise-grade analytics platform.

“Fabric has evolved beyond self-service analytics to a full enterprise-class platform. By combining data warehousing, lake storage, and Power BI, we’ve unlocked a foundation that sparks curiosity and innovation. People naturally ask, ‘What can I build on top of this?” —Patrick Sura, Global Reporting Architect, Dentsu

Nebraska Furniture Mart, a Berkshire Hathaway company managing a large BI estate under significant performance and resource pressure, migrated to Fabric Data Warehouse using the migration assistant.

“The tool facilitated every step of the process, automatically addressed column type conflicts, and effectively handled data migration right after the metadata transfer.” — Siri Kakumanu, Senior Manager, Data Analytics and Business Intelligence, Nebraska Furniture Mart

The future of your data estate starts now

Whether you are modernizing data pipelines, scaling analytics workloads, or simplifying Spark operations, these tools are designed to help you move forward with confidence and at your own pace. If you want to learn about the migration paths available to Fabric, start with Azure to Fabric Migration site. If you’ve scoped your migration, use cases and are ready to get started, check out our migration documentation for guidance and actionable steps.

If you’re looking for hands-on support, connect with your Microsoft account team or visit the Microsoft Fabric Community to learn from customers who have already made the move.

Original source

If you haven't already, check out FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and Agents for a complete look at all of our FabCon Europe announcements across Fabric and Microsoft Databases.

Welcome to the September 2026 Fabric Feature Summary!

As the Microsoft Fabric community gathers in Barcelona for the European Microsoft Fabric + SQL Community Conference, this month brings one of our biggest updates yet. From governance, CI/CD, and data engineering enhancements to AI-powered experiences, analytics, data integration, real-time intelligence, and Fabric IQ innovations, we're sharing the latest capabilities designed to help organizations build, manage, and scale their data estate with confidence. Whether you're joining us in person or following along from afar, this roundup is your guide to everything new across Microsoft Fabric this month.

Table of Contents

• Fabric Platform

  • Policies in Fabric (Preview)
  • Fabric CI/CD - Deployment Plan (Preview)
  • Restrict access to sensitive data with data loss prevention policies (Generally Available)
  • Browse and explore objects in OneLake Catalog
  • Discover tables and more with the improved OneLake Catalog Search API (Preview)
  • Find the right data faster with the new Global Search experience
  • Fabric Core MCP Server (Generally Available)

• Git Integration

  • Branch workspace admin profile (Preview)
  • Compare and commit changes (Generally Available)
  • File level commit (Preview)
  • Branch out with selective branching (Generally Available)
  • Branch workspace (Generally Available)
  • Bulk export / import API (Generally Available)

• OneLake

  • OneLake Outbound Access Protection with external shortcuts (Generally Available)
  • OneLake Table Read API (Preview)
  • New OneLake security UX with Members and Data Tabs (Generally Available)

• Data Engineering

  • Custom Live Pools for Fabric Data Engineering (Generally Available)
  • Native JSON acceleration in the Native Execution Engine (Preview)
  • Delegated management of Custom Live Pools for workspace Members (Preview)
  • Notebook toolkit for agentic development support (Preview)
  • Integration with Codex/Claude Code in Fabric Data Engineering VS Code extension
  • Spark connector for SQL databases (Generally Available)
  • High Concurrency Support for the Fabric Livy API (Generally Available)
  • High-concurrency support in the Microsoft ADO.NET driver (Generally Available)
  • High-concurrency mode for the Microsoft JDBC driver (Generally Available)
  • High-concurrency support in the Microsoft ODBC driver (Generally Available)
  • Resize columns in your Lakehouse Explorer table preview (Generally Available)
  • Project Osmos: Data engineering in Microsoft Fabric (Preview)

• Data Science

  • Spark 4.1 Support for ML Endpoints (Preview)
  • Tensor-Based Models on Endpoints (Preview)

• Data Warehouse

  • Modernize Teradata Workloads Faster with Migration Assistant (Preview)
  • Fabric warehouse Custom SQL pools (Generally Available)
  • SQL pools Statement Type classifier (Generally Available)
  • Distributed Bitmap Filters (Generally Available)
  • New analytical functions (Preview)
  • T-SQL language enhancements (Preview)
  • Integration with Fabric User Data Functions (Preview)
  • Automate Every Deployment with Confidence: Introducing Pre/Post Deployment Support (Preview)
  • Automate Fabric Warehouse Lifecycle Management with Item definition support in CRUD APIs
  • Diagnose warehouse workloads with the SQL DW operations skill (Generally Available)
  • More ways to use scalar UDFs (Preview)

• Real-Time Intelligence

  • Business Events (Generally Available)
  • Create richer visuals and denser layouts for smarter live monitoring (Generally Available)

• Real-Time Hub

  • Manage all your Activator rules from Real-time hub (Preview)
  • Learn to monitor streaming data in Real-time hub in minutes (Generally Available)
  • Create capacity alerts quickly with ready-made templates (Generally Available)
  • Capacity Overview Events in Real-time hub (Generally Available)
  • Capacity Operation Events in Real-time hub (Preview)
  • New Item Soft-Delete and Recovery Events in Real-time hub (Preview)

• Eventstreams

  • Fabric Event Schemaset and Schema Registry (Generally Available)
  • Schema-aware Fabric Eventstreams (Preview)
  • Eventstream Custom Stream Connector (Preview)
  • Expanding Connectivity with New Connector Releases in Eventstream
  • Workspace Private link for Eventstream (Preview)
  • Workspace Monitoring Azure Stream Analytics Log Support for Eventstream

• Real-Time Dashboard

  • Copilot creation of Real-Time Dashboard (Preview)
  • AI-native custom visual builder in Real-Time Dashboard (Preview)
  • Real-Time Dashboards with Ontology as a source (Preview)

• Activator

  • KQL/SQL query visibility in Activator UI (Generally Available)
  • Create and manage rules from OneLake catalog & Workspace (Preview)
  • Trigger Copy job as an action (Generally Available)
  • Trigger Dataflow as an action (Generally Available)
  • Trigger UDF as an action (Generally Available)
  • Stateful conditions in Activator for Event based triggers (Generally Available)
  • Dynamic thresholds in Activator (Generally Available)

• Eventhouse

  • Large Messages support in Eventstream and Eventhouse (Preview)
  • Granular Sharing for Shortcut Databases in Eventhouse (Generally Available)
  • Update Policy on Accelerated Shortcuts in Eventhouse (Preview)
  • Entity Diagram in Eventhouse (Generally Available)
  • New Workspace Monitoring logs for Eventhouse: throttling events, sub-optimal size, and scale-out events (Preview)
  • Support for workspace outbound access protection (Generally Available)
  • Workspace-level Private Link (Generally Available)
  • Cluster-level Customer-Managed Key Support (Generally Available)

• Maps

  • Connect to External Geospatial Services with Feature Service Support (Preview)
  • Bring Fabric Maps into Real-Time Dashboards (Preview)
  • Set alerts directly from table visuals in Fabric Real-Time Dashboards

• Fabric IQ

• Conversational Analytics

  • Fabric IQ MCP Server (Generally Available)
  • Expansion of Fabric IQ to include answers from data agents and ontologies (Preview)
  • Visualizations in the Fabric Data Agent (Generally Available)
  • Scale Business Context with Topics in Fabric Data Agents (Preview)
  • Model Upgrades on the Data Agent Runtime (Preview)
  • Ontology as Context Source (Preview)
  • Fabric data agents now return up to 1000 rows (Preview)
  • Monitor and improve data agents with user feedback in Microsoft 365 Copilot (Preview)
  • Microsoft Fabric data agent now supports Model Context Protocol tasks (Generally Available)

• Ontology

  • Bring Power BI DAX measures into the Ontology
  • Add richer business context with metadata
  • Reuse common concepts with type hierarchies and inheritance
  • Capture business logic with natural-language rules
  • Reuse existing standards with RDF/OWL import and export
  • Build and query ontologies conversationally
  • Organize enterprise concepts with namespaces
  • Bind faster and query at source scale
  • Configurable Ontology graph experiences
  • Apply enterprise security and network controls
  • Manage change with built-in versioning
  • Simplified user experience and migration to new experience
  • Fabric Graph Updates for Flexible, Relationship-Heavy Analysis

• Databases

  • Database Hub (Preview)
  • Database agent (Preview)
  • Azure Cosmos DB Mirroring in Microsoft Fabric Now Supports Virtual Network Data Gateways
  • Change Event Streaming for SQL database in Microsoft Fabric (Preview)

• Data Factory

• Pipelines

  • Lakehouse maintenance activity (Generally Available)
  • Support for activity retry back-off logic (Generally Available)
  • Support for pipeline-level dependencies with run conditions (Preview)
  • Richer pipeline monitoring in workspace monitoring (Preview)
  • Fabric Actions activity (Preview)
  • Business Actions activity (Preview)

• Mirroring

  • Mirroring for Google Big Query (Generally Available)
  • Mirroring for SharePoint List (Generally Available)
  • Extended capabilities for Mirroring (Generally Available)
  • Mirroring Security Roles for Snowflake (Preview)

• Copy Job

  • Change data capture in Copy job (Generally Available)
  • SCD Type 2 in Copy job (Generally Available)
  • Audit columns in Copy job (Generally Available)
  • Copy job integration with Eventstream as a source and destination (Preview)
  • Copy job supports SQL CDC custom capture instance names (Generally Available)
  • Copy job integration with workspace monitoring (Generally Available)
  • Custom staging database and schema for CDC and SCD Type 2 in Copy job (Generally Available)
  • CDC from read-only Snowflake databases in Copy job (Generally Available)
  • Oracle CDC initial-load improvements (Generally Available)
  • Azure Database for PostgreSQL connector for Copy job (Generally Available)
  • Multi-folder copy into separate destination tables (Generally Available)
  • Audit columns for Snowflake destinations (Generally Available)
  • SharePoint folders as source and destination in Copy job (Generally Available)

• Dataflow Gen2

  • Optimized copy to Lakehouse (Generally Available)
  • Offset transformation: offset a date, datetime or time column backwards or forward in time (Preview)
  • Custom sort dialog (Preview)
  • Now and today ribbon command
  • Data visuals in Dataflows (Preview)
  • My Queries: your personal library of queries (Generally Available)
  • Shared queries: discover, share and consume queries with colleagues (Preview)
  • New data destinations (Generally Available)
  • Dynamic expression editor for data destinations (Preview)
  • Send email from Dataflow Gen2 (Preview)
  • Dataflow diagnostics download (Generally Available)
  • Introducing Preview only step in Navigator
  • Expanded support for the modern evaluator in Dataflow Gen2
  • Refresh details in the Fabric Monitoring hub (Preview)
  • Dynamic schema support for the Fabric Data Warehouse destination (Preview)
  • Backup query results to a Lakehouse (Preview)
  • Modern Get Data: browse Azure resources (Preview)
  • New enhancements for upgrading Power BI Dataflows Gen1 to Dataflows Gen2 (CI/CD) (Preview)
  • New enhancements for Mapping Data Flow Transforms in Dataflow Gen2 (Preview)
  • Upgrade to Fabric Data Factory with confidence

• Connection Governance

  • Data connection policies: Authentication allowlist (Generally Available)
  • Data connection policies: Tenant allowlist (Generally Available)
  • Admin APIs for cloud connections (Generally Available)

• App Development

  • Functions, Connectors and more with Fabric Apps (Preview)

• Coming Soon

  • On-demand billing and Zero-provisioned (F0) Capacity (Preview)
  • On-demand billing for Fabric Data Warehouse
  • Estimated graphical query plan
  • Data Warehouse in Workspace Monitoring
  • AI Insights in DW Monitor

That’s a wrap!

The European Microsoft Fabric + SQL Community Conference is just getting started, and there's never been a more exciting time to be part of the Fabric community. We hope these updates help you discover new capabilities, streamline operations, strengthen governance, and unlock more value from your data. Be sure to explore the links throughout this post for deeper technical guidance, and follow along with the latest announcements, demos, and sessions from Barcelona. We're excited to see what you build with Microsoft Fabric next.

Original source

If you haven’t already, check out Arun Ulag’s hero blog “FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and Agents” for a complete look at all our FabCon and SQLCon announcements across both Fabric and our database offerings.

Organizations are moving quickly to turn AI into meaningful business outcomes. But the value of AI still depends on the data behind it: whether teams can find the right data, access it securely, and use it across the tools and platforms where agents work. When data remains fragmented across clouds, business units, and technology stacks, every new AI initiative starts with the same costly work of moving, copying, and reconciling data.

Microsoft OneLake provides a unified data foundation for Microsoft Fabric, helping organizations connect and govern data across their estate without creating more silos. Every Fabric user and customer is also a OneLake user and customer and at FabCon and SQLCon Barcelona 2026, I’m excited to share that we're expanding OneLake further to continue on our principal of providing customers an open, secure, and scalable foundation for connecting and sharing data and putting governed data to work across analytics and AI,

Connect and share data across your distributed estate

Data is increasingly distributed across clouds, applications, partners, and platforms. OneLake helps organizations work across that distributed estate without requiring every team to move and maintain another copy of the data.

Share governed data and context with IQ sharing

An organization’s unique IQ is its data, knowledge, and workflows. When copilots, agents, and applications can draw on this shared understanding of the business, they can deliver more consistent and relevant experiences. But that context needs to be governed and reusable so the same terminology, relationships, instructions, and domain knowledge can be applied across teams and applications.

Today, we’re introducing IQ sharing in preview soon, a new capability that enables organizations to securely share governed data and business context across teams, business units, customers, partners, and ecosystems. This makes trusted organizational context portable and reusable, extending it beyond a single team or organization.

At launch, IQ sharing supports data, metadata and context represented in files, including Markdown (.md) agent instructions and RDF (.rdf) ontologies. We plan to expand these capabilities with native support for Fabric IQ ontologies and Power BI semantic models, making it easier to securely share and reuse business context across AI experiences.

EY, TomTom, S&P Global, Financial Fabric, and xMentium are already building integrations with IQ sharing, with more partners to come. TomTom, for example, is bringing its geospatial and transportation intelligence into Fabric through IQ sharing, with available data products discoverable through the OneLake catalog.

"Trusted location data is a foundational ingredient for our customers' applications, AI solutions, and intelligent agents," said Leo Sei, Chief Product Officer, TomTom. "We're excited to be one of the first partners to onboard to IQ sharing and extend access to TomTom location data products across our customers' analytics and AI ecosystems. IQ share makes our data easy to discover, securely distribute, and centrally govern while preserving the control, context, and trust that are critical for delivering AI-ready data products."

John Whittaker, VP of Cloud Marketplace at EY, sees similar potential for IQ sharing to make business context more portable and reusable:

“In collaboration with Microsoft, IQ sharing is an important step toward making context reusable, governed, and portable, with the potential to bring trusted knowledge, methods, and domain insight to clients in new AI-enabled ways, helping people move faster, support stronger decision-making, and scale value with confidence.”

Together, OneLake and IQ sharing provide an open foundation for distributing governed intelligence, helping organizations reuse trusted data and business context to power analytics, applications, copilots, and agents.

Connect and transform data across your entire estate

We’re continuing to expand OneLake’s open data network across platforms, clouds, and business applications, giving organizations more ways to work with distributed data through a shared data foundation without having to move, duplicate, or maintain additional copies.

Connect Salesforce Data 360 to OneLake with bidirectional integration

OneLake and Salesforce Data 360 are coming together to give customers a simpler way to use their Salesforce data across Microsoft Fabric without moving or duplicating it. With OneLake shortcuts, Salesforce Data 360 data becomes part of the unified data estate, where teams can discover it in the OneLake catalog and use it across Fabric workloads, including SQL, notebooks, and Power BI. Customers can also use Salesforce data to build ontologies that provide rich business context to Fabric data agents, helping them reason over it alongside the rest of their enterprise data. Get started now.

Because OneLake shortcuts access data directly from Salesforce storage, customers can work from a single copy of always-current data without building pipelines, paying for duplicate storage, or keeping copies in sync. Queries use Fabric compute without placing additional load on Salesforce instances, while Salesforce administrators remain in control of what data is shared. Salesforce controls and OneLake security provide governed access across Fabric, giving customers the flexibility to use their Salesforce data for analytics and AI while keeping it securely managed.

Access versioned lakeFS data through OneLake shortcuts

OneLake shortcuts now provide a complementary way to access versioned data managed in lakeFS from Microsoft Fabric. Customers can connect to the latest data from a lakeFS branch or reference a specific immutable commit, helping them create reproducible reports, experiments, and audits without building separate pipelines to copy and maintain snapshots of their data.

Mirror Dynamics 365 Business Central data into OneLake

We’re also extending this connected data foundation to Dynamics 365 Business Central. A new integration enables organizations to bring selected Business Central tables and company data directly into OneLake through mirroring. Built-in monitoring provides both overview and detailed logs, helping teams track integration health and troubleshoot issues.

Read and write OneLake data from Azure Databricks

For organizations working across data platforms, we’re strengthening interoperability with Azure Databricks through native storage in OneLake, now supported in production. This gives joint customers a more direct way to use OneLake data from Azure Databricks without copying it into another storage layer, helping teams preserve a consistent data foundation across platforms.

Connect more external data to OneLake through open standards

We’re continuing to extend the range of external data that organizations can make available through OneLake. Mirroring for Google BigQuery is now available, making it easier to work with Google Cloud data in Fabric for analytics and AI while reducing duplication and bespoke integrations. New Apache Iceberg-based catalog federation support for AWS Glue Data Catalog, Google Lakehouse runtime catalog, and CONNECT from AVEVA further expands the data organizations can discover and use through OneLake, bringing more of the distributed data estate within reach of Fabric.

Operate and scale OneLake with greater flexibility

An enterprise data foundation must be flexible to operate and powerful to use. Organizations need simpler ways to align costs with demand, monitor their data estate, and scale as analytics and AI workloads grow.

On-demand billing brings more flexibility to OneLake compute

In the coming weeks, we’re introducing on-demand billing for OneLake compute in preview, allowing customers to pay only for the OneLake compute they use. Previously, access to data in OneLake depended on a running Fabric capacity, which could be affected by pauses or throttling. With on-demand billing, OneLake compute can scale elastically when needed, helping minimize throttling during demand spikes.

We’re also introducing Fabric zero-provisioned (F0), a new entry point to Fabric and OneLake that removes the need to provision compute before getting started. With on-demand billing enabled by default, organizations can immediately access Fabric services and pay only for the resources they use. Together, on-demand billing and Fabric zero-provisioned allow organizations to read and write data without maintaining dedicated capacity.

The announcement builds on Microsoft’s broader vision of OneLake as a shared data foundation across platforms such as Fabric, Azure Databricks, and Snowflake. Recent interoperability enhancements allow organizations to store and access a single copy of data across these environments, reducing duplication and simplifying data architectures. By decoupling OneLake access from Fabric capacity, on-demand billing makes it easier for customers to adopt OneLake as their centralized enterprise data lake while continuing to use the compute engines that best fit their needs. This applies to AI workloads as well. Microsoft Foundry can seamlessly be used with OneLake on-demand billing as well.

Put OneLake data to work across analytics, AI, and applications

Connecting data is only the first step. Organizations also need to make governed data easy to discover and use across the tools, applications, and AI experiences where people and agents work.

Discover and use governed data in Microsoft Foundry and Excel

Data becomes more valuable when people can discover and use it from the tools where they already work. A new OneLake catalog integration in Excel, available through the modern Get Data experience, helps business users discover and analyze organizational data directly from Excel. By reducing the steps between discovery and analysis, the integration helps more users work with governed Fabric data through a familiar experience.

We are also announcing deeper integration with Microsoft Foundry, our end-to-end platform for customizing, hosting, running, and managing AI solutions. With this integration, you can access the OneLake catalog in Foundry, making it easy to discover and build knowledge directly from OneLake data. Foundry users can explore their data estate including rich metadata, endorsements, sensitivity labels, and descriptions, and connect their OneLake items with a single selection to add data to Knowledge.

Give agents, applications, and data engines secure access to OneLake tables

The OneLake Table Read API (Preview), gives AI agents and applications secure, serverless access to Delta Lake and Apache Iceberg tables without requiring SQL. OneLake security automatically enforces table-, row-, and column-level access controls on every query, returning only authorized data. Built on open-source Apache Arrow, the API helps developers efficiently integrate governed enterprise data into AI agents and applications.

We’re also expanding the ways external data engines can work with OneLake through open APIs. ClickHouse support for reading Apache Iceberg tables from OneLake using OneLake Table APIs is now generally available, enabling customers to query OneLake data directly without creating another copy. Support for writing Iceberg tables to OneLake using Table APIs and credential vending (Preview), allowing results produced in ClickHouse to become part of the shared OneLake data foundation and available across Fabric.

In addition to working with OneLake data directly through Table APIs, customers can use the new ClickHouse workload in Microsoft Fabric when they want dedicated analytical compute. Now in public preview, the experience provisions a dedicated ClickHouse Cloud service on Azure and includes an embedded SQL console within Fabric. Teams can synchronize OneLake tables as accelerated, point-in-time copies for analysis in ClickHouse Cloud, while OneLake remains the source of truth. This gives developers a dedicated environment for responsive analytics, interactive applications, and AI agent experiences while keeping those workloads connected to their Fabric data.

Teradata Autonomous Knowledge Platform is integrated OneLake Table APIs, enabling enterprise customers to run Teradata's high-performance enterprise AI directly on data stored in OneLake, without extract, transform, and load (ETL) pipelines, data duplication, or migration. Built on open Apache Iceberg standards, the integration allows Teradata users to query OneLake tables in place using standard Iceberg APIs, with cross-platform authentication and access controls handled natively. Enterprises with data in both environments can now join and analyze it without first consolidating it into a single system or rebuilding governance from scratch.

LakeSail Cloud further expands the engines that can work with data in OneLake by bringing Rust-native Spark compatibility to Delta Lake and Apache Iceberg tables. With LakeSail, teams can use Spark SQL, DataFrame, and Python workloads to read and write OneLake tables without deploying separate data connectors or changing existing Spark code. This gives organizations another way to use the processing engines that fit their workloads while keeping data in the shared OneLake foundation.

Build operational applications on OneLake

Organizations increasingly need to turn governed data into applications and workflows that help people make decisions and take action. OneLake provides a shared foundation for these operational experiences, keeping applications and their data connected across Fabric.

Last May, we introduced Fabric Apps to help developers build and deploy data and AI-powered applications on trusted, governed data in Microsoft Fabric. We’re now expanding Fabric Apps with support for OneLake sources alongside existing SQL databases. Developers can connect applications to Fabric warehouses, lakehouses, and Power BI semantic models while preserving existing permissions and trusted business logic. This brings applications closer to the data they rely on and reduces the need to move or duplicate data across systems.

We’re also deepening the integration between Planning in Fabric (Preview) and OneLake. Now, planning data and writeback can be stored directly in OneLake, enabling teams to work across larger models and with more contributors while maintaining a responsive experience. These improvements increase planning performance and scale by up to 10x, while keeping planning outputs readily available for use across the broader Fabric ecosystem. Check out all the innovation coming to Planning in Fabric in my other blog, Planning in Microsoft Fabric: From insight to action.

Discover, govern, and protect your data estate

As data becomes available to more people, applications, and agents, governance and security become even more important. Organizations need a unified way to discover their data, manage access, and apply consistent protections wherever that data is used.

Discover more of your data with deeper lineage in the OneLake catalog

The OneLake catalog is becoming a richer, more granular place to discover and understand data across Fabric. Users can now go beyond finding Fabric items to explore individual tables and columns, view their details, and preview the underlying data directly in the catalog, making it easier to evaluate data before putting it to work.

More granular search makes it easier to find that data across Fabric. The OneLake catalog search API now returns OneLake tables and semantic model tables alongside Fabric items in a single relevance-ranked result set. New Spark granular runtime lineage, coming soon, adds another layer of context, helping users understand how data is used and transformed across their data estate.

Manage and govern your data estate in the OneLake catalog

Discovery is only part of the equation. The OneLake catalog Govern experience brings centralized governance capabilities into the OneLake catalog, helping administrators manage their data across tenant, domain, capacity, and workspace scopes from a unified experience. Automatically generated signals and proactive recommendations help administrators understand their governance posture, identify areas that need attention, and take informed action.

Governance also extends into developer workflows. Govern Skills for Fabric enable AI coding assistants such as GitHub Copilot to discover existing Fabric items and tables through the OneLake catalog. This supports a find-before-you-build approach, helping developers find and reuse governed data before creating new assets or writing code.

Connect OneLake data with trusted context from Atlan

Understanding data requires more than knowing where it lives. Teams also need the business context, lineage, and governance information that explain what data means and how it is used. We’re expanding our integration with Atlan, an industry-leading provider of data and AI governance that helps organizations discover, understand, and govern data across their enterprise.

Atlan’s native Microsoft Fabric connector (Generally Available) and adds deeper lineage across semantic models and measures, Spark workloads through the new Fabric Platform Lineage APIs, and Dataflow Gen2. The integration also extends in the other direction with OneLake shortcuts, enabling teams to access Atlan context and Atlan lakehouses from within Fabric and query it using Fabric Spark and SQL capabilities. Together, these integrations bring OneLake data and the context surrounding it closer together, helping teams better understand and govern the data they use for analytics and AI.

Apply more precise policies across Fabric

Policies in Fabric (Preview) introduce precise, attribute-based controls that help organizations enforce governance requirements based on user and data attributes, rather than relying only on broad administrator settings.

Policies support granular controls for key scenarios, with item creation and workspace security available at launch and external data sharing coming soon. Policies also come with a centralized policy management experience, along with monitoring and APIs that make it easier for administrators to manage and track policies across Fabric. CI/CD integration helps teams automate the policy lifecycle as their environments grow.

For example, external data sharing policies will give administrators more flexibility over cross-tenant collaboration. Rather than simply turning external sharing on or off, organizations can apply more granular controls based on attributes such as users, workspaces, and recipients, helping teams collaborate externally while maintaining the controls needed for security, governance, and compliance.

Strengthen data protection with OneLake security

As access to data expands across users, applications, AI agents, and platforms, organizations need security controls that remain consistent wherever that data is used. New OneLake security capabilities make it easier to understand and manage access within OneLake, preserve existing protections as data moves across platforms, and control how OneLake connects to external sources.

We’ve redesigned the View users in role experience for OneLake security, now generally available, to simplify permissions auditing and management. New views help administrators quickly understand who can access a specific table and what data an individual user can see. From the same experience, they can assign users to multiple roles or revoke access, making it easier to review and manage permissions as data usage grows.

We’re also making it easier to maintain consistent protections for data used across platforms. Trust3.AI is introducing a capability that synchronizes row- and column-level policies between Databricks, Snowflake, and OneLake security. By mapping policies from these platforms to OneLake security rules, the integration helps organizations maintain consistent access controls for users, applications, and AI agents without manually recreating policies in each environment.

Mirroring for Snowflake brings Snowflake data into OneLake and keeps it synchronized for use across Fabric. Until now, customers have had to recreate the security policies governing that data after it was mirrored. With Snowflake Security Roles Replication, now in public preview, organizations can replicate source-defined roles and permissions alongside their data and translate them into OneLake security. This carries important governance context with the data and reduces the need to manually rebuild access controls as organizations bring their Snowflake and Fabric data estates together.

Finally, outbound access protection for OneLake shortcuts (Generally Available), giving Fabric administrators control over which external data sources and endpoints shortcuts in a workspace can connect to. These controls help reduce the risk of unintended or malicious data exfiltration while still allowing approved connections that support business needs.

Expand what you can do through the Fabric partner ecosystem

OneLake provides a shared data foundation that partners can build on to bring specialized capabilities directly into Microsoft Fabric. At FabCon and SQLCon Barcelona, we’re expanding the Fabric partner ecosystem with new workloads that give organizations more ways to analyze, manage, and act on the data in OneLake without leaving Fabric.

Esri: bring location context to business analysis (Generally Available)

Location can reveal patterns and relationships that are difficult to see in rows and columns alone. Esri’s ArcGIS Maps for Microsoft Fabric brings mapping and spatial analysis into Fabric, helping teams explore their business data geographically. Stream Layers add support for real-time information, while expanded Parquet and OneLake mapping makes more organizational data available for spatial analysis. Enhanced authoring, symbology, and sketching capabilities give analysts more flexibility to tailor maps to the questions they want to investigate.

Telmai: Build confidence in the data behind every decision

Reliable analytics and AI depend on knowing whether the underlying data is complete, current, and ready to use. Telmai Data Reliability for Microsoft Fabric, now generally available, continuously monitors critical OneLake tables for changes in data volume, schema, freshness, and completeness.

Telmai uses context from the OneLake catalog to help teams focus their monitoring on critical data assets. AI-assisted incident correlation connects related issues across the data estate, while Ask Telmai provides a conversational way to investigate potential causes instead of reviewing isolated alerts one by one. Telmai can also make machine-readable trust signals available to Fabric data agents and other compatible AI clients, helping both people and AI-driven experiences account for data reliability when using OneLake data.

Celonis: understand how work flows and where it can improve (Generally Available)

Business processes rarely follow a simple path. Celonis Process Intelligence brings process visualizations and business context into Fabric, helping teams understand how work moves across their organization using the data they already have. Users can explore process flows, investigate inefficiencies and their root causes, and identify opportunities for improvement alongside the analytics they use to understand the rest of their business.

Spectral Core: simplify database migration to Fabric (Generally Available)

Spectral Core Fabric Migration brings data movement, SQL code translation, schema and object conversion, and validation into a guided Fabric experience. Teams can work through the steps involved in migrating databases with a clearer view of what is being moved and where additional attention is needed.

Financial Fabric: bring capital markets data closer to analysis (Generally Available)

Financial Fabric’s Capital Markets DataHub brings data from more than 200 data providers and 10,000 datasets into Fabric, helping investment, risk, and operations teams reduce the work required to source and access financial data. Ask DataHub provides conversational exploration, while Power BI Direct Lake analytics connects that data to familiar analysis experiences.

IntuigenceAI: turn industrial information into usable knowledge (Generally Available)

IntuigenceAI helps energy and process manufacturing teams turn information spread across drawings, schematics, datasheets, and logs into a queryable knowledge graph. Its AI synthetic engineers provide a multilingual conversational experience for retrieving information and supporting engineering work, with data residing in the customer’s OneLake.

Metropolis: connect communications data with business outcomes (Preview)

Metropolis VoiceLake normalizes supported communications data into a Fabric Lakehouse and provides ready-made Power BI reports and a semantic model for exploration. Organizations can combine communications records with other business data for analytics and AI, providing a broader view of service activity and operational performance.

Figure: Exploring and validating data with natural-language analysis in Daivio.

And there are many more. Together, these partner workloads extend the ways organizations can analyze, manage, and act on their data in Fabric. Customers can explore these experiences in the Fabric Workload Hub and choose the specialized capabilities that fit their business needs.

Build on a more open, secure, and manageable data foundation

These announcements advance a common goal: helping organizations make more of their data available for analytics and AI without creating more copies or giving up control. IQ sharing, shortcuts, mirroring, and deeper platform interoperability connect data across organizational and technology boundaries. The OneLake catalog, security, policies, and expanded partner integrations help teams discover, understand, govern, and put that data to work. And with more flexible billing and stronger monitoring, organizations can operate and scale their data foundation as their needs grow.

Explore more Microsoft Fabric innovation

Alongside these OneLake announcements, we are introducing several innovations across Fabric. For a broad overview of all these announcements, read Arun Ulag’s hero blog “FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and Agents.” If you want more details, check out the Fabric September 2026 Feature summary blog and the latest posts on the Fabric Updates channel, including my other blog: Planning in Microsoft Fabric: From insight to action

Original source

My workspace governance features, which give administrators governance controls over their users' personal My workspaces, are now generally available.

As My workspaces continue to be used for business-critical content, organizations need governance capabilities that scale with their environment. These enhancements give administrators greater flexibility to access, manage, and recover content when needed, while enabling automation that reduces operational overhead and improves consistency across the tenant. These enhanced tenant administrator controls for My workspace are also available in government clouds.

What's new

Two new admin REST APIs make temporary admin access to a personal workspace fully programmable:

• Grant Admin Temporary Access: grants a Fabric administrator 24-hour access to a user's My workspace.
• Remove Admin Temporary Access: revokes that access before the 24-hour window ends.

Both can be called by a Fabric administrator or by a service principal. The service principal identity must be enabled for Admin API access in your tenant settings. SPN support and these new APIs allow tenant administrators to automate management of tenant users’ My workspaces contents instead of doing it by hand in the Admin Portal.

Reminder: Restoring orphaned My workspaces

When a user is hard deleted in Microsoft Entra, their My workspace becomes orphaned and moves to a Deleted state, with contents retained for 30 days. During that window, you can restore it with the Restore Workspace API.

Important: new workspace name is required when restoring a former My workspace, and you assign a new admin to own the recovered content.

Try the Admin controls and new APIs and let us know what you think in the comment section.

Original source

Important business moments rarely wait for the next scheduled refresh. An order is approved, a shipment enters a risk state, a capacity approaches a limit, or a workspace item is accidentally deleted. When teams discover those changes late, the delay affects operations, analytics, and customer outcomes at the same time.

Microsoft Fabric helps teams react when those moments happen. Real-Time Hub provides a common place to discover event sources, connect consumers, and move from a signal to an action. With Business Events now generally available and new Fabric Events capabilities arriving in the September 2026 release, organizations can connect business meaning and platform operations through one event-driven model.

One model, three outcomes

Publish a governed signal once, let operational teams react independently, and retain the same event for real-time and historical analysis.

Turn business meaning into a governed event (Generally Available)

A Business Event represents a meaningful occurrence or state change that matters to the organization. It is not raw telemetry or a diagnostic record. It captures business intent: a payment failed, a cold-chain shipment is at risk, inventory fell below a threshold, or an approval is required.

Every Business Event follows a versioned schema in an Event Schema Set. Publishers and consumers agree on the event name, fields, data types, and meaning before they integrate. This contract reduces point-to-point dependencies because the publisher does not need to know which teams will react, and consumer teams can add or change their responses without modifying publisher code.

Use the Fabric experience closest to the business logic

Business Events meet teams where their detection logic already runs. A notebook can publish the result of analytics or machine learning. A User Data Function can emit an event from reusable domain logic after validating an application request. Activator can convert a condition detected in Power BI, a Real-Time Dashboard, KQL, or a warehouse query into a governed signal.

This flexibility keeps ownership clear. The team that understands the business condition owns when the event is published. The Event Schema Set owns the shared contract. Downstream teams own how they respond.

The following animation follows a cold-chain shipment risk event through Real-Time Hub. It shows how a user opens the governed event, reviews its schema and connected consumers, publishes representative data, and verifies the event flow before relying on production logic.

Figure: The Business Events experience in Real-Time Hub brings the event contract, publishers, consumers, and validation tools into one workflow.

By bringing the event contract, publishers, consumers, and validation tools into one experience, Real-Time Hub gives teams a shared view of how an event moves through the system. Teams can use that visibility to confirm that the event structure and connections work as expected before the production publisher is ready.

Validate integrations before production code is ready

A faster inner loop is one of the most practical advantages of the new experience. The sample publisher generates a JSON payload from the selected schema, so teams can replace the values or upload a representative file and publish an event in seconds. Data preview then shows recent payloads for each publisher and consumer together with event metadata and the active schema version.

Teams can publish a sample Business Event, inspect Business Events data, and verify the contract, permissions, subscription, and routing before connecting the live publisher. This removes temporary test code and gives both sides of the integration direct evidence that the expected payload arrived.

For application-oriented scenarios, the guided User Data Function publisher experience prepares the Fabric connection and sample publisher code from Real-Time Hub. Developers can focus on the business condition and payload instead of integration plumbing.

Process, act, and analyze without changing the publisher

After an event is published, independent consumers can choose the response that matches their responsibility. Activator can send notifications, start workflows, or invoke downstream logic. Eventhouse can retain every event in a dedicated Kusto Query Language (KQL) table for operational monitoring, historical analysis, Real-Time Dashboards, anomaly detection, and AI or machine learning scenarios.

• React immediately to a high-severity condition with an alert or automated action.
• Route and transform the signal for downstream destinations through Eventstream.
• Retain durable evidence for audit, trends, service-level analysis, and continuous improvement.

Eventstream as a Business Events consumer is in Preview. It can filter, enrich, transform, route, and fan out an existing governed signal while the original publisher remains independent. Teams can add Business Events from an Eventstream or start from the event in Real-Time Hub through a prepopulated consumer flow. Review the Eventstream consumer guidance for configuration details.

For the generally available paths, learn how to publish Business Events from Activator and analyze Business Events in Eventhouse.

Monitor capacity health as events occur (Generally Available)

Capacity Overview Events include two event types. Capacity Summary events provide a recurring high-level view of utilization and health, while Capacity State events identify changes such as pause, resume, and throttling-related conditions. Capacity administrators no longer need to rely only on periodic checks to discover that action may be required.

These signals can drive no-code alerts in Activator, automated workflows, or Eventstream processing. They can also land in Eventhouse for KQL analysis, historical retention, and dashboards that help teams recognize patterns before users are affected.

Figure: Capacity Overview Events expose capacity state and summary signals that administrators can connect to alerts, processing, and analysis.

Follow the capacity health monitoring tutorial and review the Capacity Overview Events schema to get started.

Find the operations driving capacity pressure (Preview)

Capacity Operation Events extend capacity visibility from overall health to operation-level root cause. Fabric emits an event for each operation that consumes capacity, with details such as the workspace and item, capacity unit consumption, duration, throttling delay, status, and smoothing-window attribution.

Administrators can identify which refreshes consumed the most capacity, which operations were throttled, and where repeated failures originate. Activator can respond when an expensive operation crosses a threshold, while Eventstream and Eventhouse support routing, chargeback analysis, governance, and longer-term optimization.

Figure: Capacity Operation Events add operation-level context for troubleshooting utilization, throttling, and failures.

Review the Capacity Operation Events documentation for the complete schema.

React to item deletion and recovery (Generally Available)

Real-Time Hub now includes workspace item lifecycle events for soft-delete and recovery operations. Teams can subscribe when an item is moved into the recovery state, when recovery succeeds, or when either operation fails. These events make item lifecycle activity available to the same alerting and stream-processing tools already used for other Fabric Events.

A workspace administrator can receive an alert when a critical item is accidentally soft deleted, trigger a governance workflow, or retain the event for audit and historical analysis. Because the new events use the existing workspace item event model, teams can extend current consumers instead of building a separate monitoring path.

Figure: Workspace item events now include soft-delete and recovery outcomes for real-time notification, automation, and governance.

Explore the Fabric workspace item events documentation for event names and schema details.

See how Business Events accelerate discount approvals

In the Protect the Margin with Business Events in Fabric demonstration, a sales application sends proposal details to a User Data Function. The function applies discount and margin policies, then publishes the governed Sales.Deals.DiscountApprovalRequired event only when manager or executive approval is required.

The team first creates the event contract, connects Eventstream, publishes representative samples, and confirms delivery with Data preview. It then connects the production function and submits a live proposal through the same validated path. Eventstream remains independent from the sales application and can send the event to a Spark notebook for scoring, enrichment, or approval analytics.

Watch the end-to-end demonstration

Watch Protect the Margin with Business Events in Fabric to see how a business signal connects sales policy, real-time validation, and independent downstream processing.

Get started

Start with one event that is easy to explain and costly to miss. Define the contract, publish a representative sample, verify delivery, and add one consumer that produces a measurable operational or analytical outcome. Once the pattern is proven, additional teams can subscribe without changing the original publisher.

Choose the event path that matches your scenario: explore the Business Events overview to publish governed business signals, or the Fabric Events and Azure Events overview to react to changes across Fabric or bring Azure resource events into Fabric. Visit the Fabric Events developer site for architecture guidance, end-to-end scenarios, and samples, and use the Real-Time Hub overview to connect your first event to an action, a stream-processing path, or durable analytical history.

Original source