I recently had the privilege of attending Microsoft FabCon Europe in Vienna, where I joined more than 4,200 attendees from around the world. The event was packed with exciting announcements, insightful sessions, and opportunities to connect with industry peers — making it a truly valuable experience.

Microsoft is using Fabric to push forward its vision of data + AI being deeply integrated, contextual, and enterprise ready. By providing a complete data and analytics SaaS platform, Microsoft is breaking technical barriers, reducing data silos and increasing cost transparency. With Fabric, organisations can deliver value faster and spend less time putting together the components needed to run analytical and AI workloads.

According to Microsoft , the vision for Fabric is clear, to become the most comprehensive, enterprise-grade data platform on the planet. Here are some of the signs delivered so far:

From unified data to contextual & connected intelligence

Fabric’s goal is to centralise data through OneLake, providing organisations with a single foundation for their analytics. Among other benefits, this approach allows them to unlock greater value by enriching data with relationships (Graph), spatial context (Maps) and real-time signals, enabling AI systems to understand how entities connect rather than simply compute data.

Zero-ETL / low friction access to data wherever it lives

With mirroring, shortcuts, transformations, and cross-vendor support, Microsoft is pushing for architectures where data doesn’t have to be moved or transformed manually more than necessary. That’s more efficient, lowers latency, and reduces duplication of effort / errors.

Enterprise grade: security, governance, performance

Fabric is increasingly taking features you expect in mission-critical systems: capacity management, security controls, private networking, performance benchmarking, cost and usage predictability. Microsoft wants Fabric to not just be for experimentation, but be trusted for large scale, sensitive, uptime-critical workloads.

Developer productivity and lifecycle support

Automated tools, CI/CD, extensibility, better UI/UX, templates and integrations with developer environments reduces friction for teams building, deploying, iterating and collaborating.

Agents & AI built on top of a solid data foundation

Much of what’s announced is clearly influenced by the rise of “agents” or AI models that don’t work on static reports but need context, up-to-date data, relationships, location, etc. Graph + Maps + seamless data access + integrations with Azure AI Foundry points toward Microsoft positioning Fabric not just as analytics, but as a foundation for intelligent applications, agents and real-time decision making.

Interoperability & ecosystem

Microsoft is not trying to lock people into “just Fabric”; they are enabling connections with Snowflake, Oracle, BigQuery, Neo4j, etc. This is an admission that data estates are heterogeneous and organisations want flexibility. This also helps Microsoft argue that Fabric can be the glue across a wide variety of systems.

At FabCon, it became clear that Microsoft is listening more closely to the community — not only driving innovation and releasing new features, but also focusing on fixing issues and delivering the core functionalities that users rely on. You can check the latest announcements here, here and here.

What about what’s in the roadmap? You can check the published roadmap details here, but these are some of the features that were announced in the conference and that I’m looking forward to explore.

Administration, Governance and Security

• Capacity Metrics Chargeback
  This feature brings financial clarity, fairness, and accountability to Fabric usage. It transforms capacity from a shared “black box” into a transparent, governable resource that aligns IT costs with business value.
• Private Link support at a workspace level
  Already available for some items, this feature adds granular security, letting organisations lock down production workspaces while keeping dev/test accessible over the internet. It allows organisations to secure inbound traffic to specific workspaces instead of the entire tenant, allowing business users to still access their reports even if outside the private network.

Data Engineering

• Custom Live Pools
  When we use managed private endpoints or environments, we can no longer leverage the starter pools and have to wait 3 to 5 min for the custom pool to start. This can have a significant impact on critical workloads, however, with custom live pools, users will have the ability to create custom pools and keep them warm based on a schedule.
• High concurrency sessions in Spark
  High concurrency sessions are a great way to fully use the capacity and avoid spinning up multiple sessions per notebook, leading to livy issues. Up until now, we could only attach 5 notebooks to a single spark session, however, it will soon become possible to attach up to 50 notebooks.
• Performance improvements in environments
  Environments are a great feature that allows users to install libraries, configure new pools, set Spark properties, and upload scripts to a file system. When installing new libraries, users could expect a delay of up to 10min, however, we will soon expect a reduction of up to 70% when performing this action.
• Monitoring Enhancements
  Soon it will be possible to get a much more granular overview of the spark engine and get insights into pool start up times and why things may be taking longer than expected, better error messages and side-by-side performance analysis.

Data Factory

• Parameter support in schedules
  Probably you are asking yourself how is this not yet available?! You will be able to set up a schedule and pass a parameter, rather than hard code the values in a pipeline
• Connection property parameterisation
  A long waited feature and one that should have been available from day 1. As it stands, it is not possible to parameterise a connection in a data pipeline. As a result, when deploying to upper environments via the Fabric deployment pipelines, the connections need to be manually updated.

Data Warehouse

• Merge and Identity Columns
  If you work with SQL, you know what it means. No more workarounds trying to understand what was the last ID inserted in the table and long insert update statements to manage data changes.
• Data-driven alerts
  Easily configure alerts on business-defined queries and get notified via email or teams and act by automating key business workflows. Set alerts to detect missing updates, schema changes or anomalies during data ingestion and integration. Summarise failed data ingestion workloads, long running queries and get notified through preferred channels.

Data Science

• Unstructured data in Data Agents
  Leverage built-in notebook utilities to index PDFs or TXT files and eliminate the need to provision or manage search resources. This will allow you to not only leverage the structure data but also start exploring the potential of unstructured data by simply selecting a lakehouse folder and ask natural language questions.
• Data Agent: Support for SQL Views
  Leverage views and functions in SQL systems to generate more complex operations and scope data to a consumer’s use case.
• MLFlow3 in Fabric
  Designed to better support GenAI workflows in addition to the traditional ML lifecycle, it enhances experiment tracking, observability, performance evaluation, model versioning, prompt & agent workflows, and integrates human feedback more formally into the loop.

Real-Time Intelligence

• Operations Agent
  Train/tune AI agents using natural language prompts on what they should do to resolve issues, optimise process and drive outcomes in collaboration with humans.
• Anomaly Detector
  Continuously scan the events 24/7 as they arrive in Eventhouse for anomalies and discover the signals that hide patterns that can cause small to large improvements in your operations. With this feature, you can automatically train the anomaly detection models without data science expertise required.
• Graph
  Create graph models over your data and explore, discover and analyse multi-hop relationships to understand how entities connect.

Power BI

• End-to-end Power BI authoring in the browser
  This is a significant achievement for Mac users who now will be able to model in the Power BI service, given the core modelling parity between the web and Desktop experience.
• MCP servers for modelling and chat with your data
  Copilot can query semantic models through MCP servers, ensuring that prompts are grounded in governed data. Users can ask natural language questions, and the AI translates them into DAX/SQL against the model, while the MCP server ensures the context and schema are correctly applied. Developers can use MCP-assisted Copilot to generate or refine DAX measures, calculated columns, hierarchies, or even suggest star schema relationships.

Final Thoughts

At FabCon Vienna, Microsoft moved the narrative of Fabric from “data unification + analytics” toward “AI-readiness + action.” The announcements strengthen the foundation, add richer capabilities, improve developer experience, and push toward mission-critical workloads. If you are exploring Fabric and want to see how it can support everything from simple to mission-critical use cases, our team of experts is here to help.

Author

José Mendes
Head of Engineering at Telefónica Tech UK&I

Connect here