Top 10 Announcements from FabCon Europe 2026

FabCon Europe took place in Barcelona from September 28 to October 1, 2026. Microsoft announced a long list of updates across Fabric and Power BI. The September 2026 Fabric feature summary alone has well over 100 items.

These are the 10 announcements we’re most excited about. Most of them are in preview, so we’ve noted the release status for each one.

1. Fabric Policies (preview)

Policies were much needed. Until now, many Fabric capabilities were controlled by tenant switches that enable a feature for everyone or restrict it for everyone. Nothing was this fine-grained, which created blind spots and admin headaches. Fabric Policies let admins define who can perform an action, what the rule covers, and where it applies. If no rule matches, the action is denied.

The first three policy types are:

• Item creation (capacity scope): controls who can create specific item types in workspaces on a capacity.
• Edit workspace settings (tenant scope): controls who can change security-sensitive workspace settings, such as network security and customer-managed keys.
• External data sharing (tenant scope): controls who can share data externally, from which workspaces, for which sensitivity labels, and to which recipient domains.

Policies are stored in Policy Set items and managed in the Policies Center in the OneLake catalog. They support public APIs and Git integration, and policy changes are recorded in the Microsoft 365 audit log.

2. Deployment plans (preview)

A deployment plan is a new workspace item that controls the order in which related items deploy across environments. On a visual canvas, you group items into ordered sets and add actions that run before or after deployment. Actions can run notebooks, pipelines, copy jobs, Dataflow Gen2 items, and data functions.

For example, you can deploy a lakehouse, run a notebook that prepares its data, and only then deploy a warehouse that depends on the resulting tables. If an action or deployment fails, the plan stops before it reaches dependent items. Deployment plans work with both Git integration and deployment pipelines.

Deployment pipelines on their own have been limited. We covered 10 of their limitations in August. Deployment plans give you control over deployment order, including dependencies that Fabric lineage doesn’t detect.

3. Observability in Fabric (preview)

Observability is a new set of platform capabilities across workspace monitoring, the monitor hub, and operations agents. Workspace monitoring can now consolidate telemetry from multiple workspaces into a single eventhouse, with control over retention and caching.

The monitor hub adds Activator-powered alerts on job failures and successes, plus a capacity view for consumption and headroom. Microsoft says operations agents can suggest likely root causes and recommended actions.

The Capacity Metrics app has been limited and confusing. Building observability into the platform was the right move.

4. Project Osmos, the data engineering agent (preview)

Project Osmos is an agentic data engineering experience in Fabric, built on technology from Microsoft’s acquisition of Osmos in January 2026. You define the desired outcome, the context, and the guardrails, then delegate the work as a persistent task. Osmos inspects the workspace, plans the work, writes and runs Spark, creates or updates Fabric items, and validates the result.

It targets project-scale work such as platform migrations, ETL and lakehouse modernization, and performance optimization. Tasks inherit workspace permissions, so authorized team members can monitor and steer the same task. You can work with a task from Fabric or from GitHub Copilot, Visual Studio Code, Codex, and Claude Code.

Project Osmos is truly impressive. It can build medallion architectures and other data engineering patterns without requiring a lot of guidance. It’s still in preview, but it has the potential to change how data engineering in Fabric works.

5. Agentic app creation in Power BI Desktop (coming to preview)

Power BI Desktop is getting an experience for building data applications with natural language. You start from a semantic model, describe the app you need, preview and edit it, and publish it to Fabric. These apps can accept inputs, write back data, and preserve shared state.

These apps are built on the same technology as Fabric Apps, which gives each app a SQL database in Fabric, authentication, and security. Microsoft says apps in Power BI will be available in preview in the coming weeks to Power BI Pro, Premium Per User, and Fabric capacity customers. Pro and PPU customers get Fabric Apps and Fabric Database capabilities, up to 1 GB per app, at no additional cost.

Until now, people without Fabric capacity were left out of the app conversation. This brings Power BI-only customers into the modern era, where they can build apps to write back or take action on their data.

6. Fabric Notebook Toolkit (preview)

The Fabric Notebook Toolkit (FNTK) is a Python library that gives coding agents commands to edit, run, and debug Fabric notebooks. It supports GitHub Copilot CLI, OpenAI Codex, and Claude Code.

Coding agents can already generate Spark code, but a developer still has to pick the right workspace and notebook, place the code in the right cell, run it, and diagnose failures. FNTK gives the agent commands for those steps, so it can inspect a notebook, make a targeted change, run it remotely, and attempt fixes. Operations that modify or remove remote content are validated and confirmed before they run.

A related update adds a local MCP server to the Fabric Data Engineering VS Code extension. It gives the Claude Code and Codex VS Code extensions the same notebook context GitHub Copilot already gets.

Being able to use your agent of choice to troubleshoot notebooks is a big win.

7. Fabric Apps updates (preview)

Fabric Apps launched in June 2026 as an application layer on Fabric, built with the open-source Rayfin SDK and CLI. At FabCon Europe, Microsoft added several capabilities:

• Data connectors: Apps can connect directly to Fabric SQL databases, warehouses, lakehouse SQL analytics endpoints, and semantic models. SQL databases and warehouses support read and write. Lakehouse SQL analytics endpoints and semantic models are read-only. With delegated authentication, the connector runs as the signed-in user, and the source enforces its own permissions, including row-level security.
• PostgreSQL support: Apps can store their own data in PostgreSQL as well as SQL Server, with the same data modeling experience and generated APIs.
• Functions and Secret Store: You can write server-side logic in TypeScript that deploys with the app. Secret Store encrypts credentials and provides them to functions only at runtime.
• App metrics: Every app includes built-in metrics for sign-ins, app loads, query volume, query errors, error rates, and average query duration.

It’s good to see Fabric Apps maturing.

8. T-SQL language enhancements in Fabric Data Warehouse (preview)

Anyone who has written T-SQL, changed the SELECT clause, and then received an error because the GROUP BY clause no longer matched knows the annoyance. Fabric Data Warehouse has fixed that with GROUP BY ALL.

Fabric Data Warehouse and the SQL analytics endpoint now support GROUP BY ALL, ORDER BY ALL, and the QUALIFY clause. GROUP BY ALL groups by every non-aggregated column in the SELECT list, so you don’t have to repeat them. QUALIFY filters on window function results without wrapping the query in a CTE or subquery.

Fabric also supports a FROM … SELECT query format, which lists data sources before the columns. Separately, new MEDIAN, QUANTILE, APPROX_MEDIAN, and APPROX_QUANTILE functions are in preview. PERCENTILE_CONT and PERCENTILE_DISC now support aggregate syntax.

9. Microsoft commits to Apache Ossie

Microsoft affirmed its commitment to Apache Ossie, a vendor-neutral specification for exchanging semantic metadata across analytics, AI, and BI platforms. Ossie was formerly known as Open Semantic Interchange. Snowflake helped found it, and it is currently incubating at the Apache Software Foundation.

Microsoft highlighted converting an Ossie document into both a Snowflake semantic view and a Power BI semantic model. Microsoft also plans to help advance Ossie support for ontologies.

This is early, but it matters for organizations that run both Snowflake and Fabric. Today, the same metrics often get defined twice. Ossie is meant to provide a shared definition both platforms can read.

10. Fabric IQ in Microsoft Copilot (generally available)

Fabric IQ is now generally available in Microsoft Copilot Chat and Copilot Cowork. Copilot users can ask business questions and get answers based on their organization’s Power BI semantic models and reports. Microsoft says this comes with no additional AI token costs.

People can get answers from Power BI data without having to go find a report. That convenience also creates governance work you should do now, before people start relying on the answers. Review the settings that make Fabric data available in Microsoft Copilot, in both the Microsoft 365 admin center and the Fabric admin portal. On your semantic models, set AI instructions and verified answers, and mark the models you trust as approved for Copilot. A new semantic model setting also lets model editors block read-only users from using the model through Copilot and Microsoft’s MCP tools.

What’s next

Most of these announcements are in preview, so expect details and timelines to change before general availability. Fabric IQ in Copilot is the only generally available item here, so start by getting your Copilot governance settings and semantic models ready.

If you’d like help evaluating these features or deciding where they fit in your Fabric environment, contact us at ProcureSQL.