April 14, 2026Microsoft Fabric

Episode 322 – FabCon Atlanta 2026 Keynote recap

John was there with the 7,999 other attendees packed the State Farm Arena in Atlanta for FabCon and SQLCon 2026 — the third annual Fabric Community Conference and the first time SQLCon joined the party under the same roof. Jason was not there. The FabCon keynote fell on his birthday, and he was very happy watching it from his couch in San Antonio while the wildflowers bloomed in his backyard.

Episode 322 is their take on the keynote — not in linear order, but organized around what mattered most to two practitioners who have been in the Microsoft BI and data space for longer than they’d like to admit. This was a packed keynote. There’s a lot to cover. Jason put the SharePoint and Excel integration stuff first because that’s exactly where it belongs.

The Big One: SharePoint List Mirroring Is Here (Preview)

They spent more time on this one than any other topic in the episode — and rightfully so. SharePoint List Mirroring in Fabric is now in public preview, and both John and Jason are genuinely excited about what it means for reporting on SharePoint data.

The concept: changes to a SharePoint list get continuously replicated into OneLake as a mirrored database, exposed as Delta tables, with the full Direct Lake connection story on top. No more pulling SharePoint data into an import model and wrestling with refresh schedules. No more fragile connectors fighting SharePoint’s quirks. For organizations with real data living in SharePoint lists — and there are a lot of them — this is the answer they’ve been waiting for.

John noted that the Fabric team was a little surprised by the amount of interest this announcement generated. He and Jason were not surprised at all. Every Power BI practitioner has had the SharePoint list conversation. Everybody knows how much data lives there.

There’s a caveat worth calling out: as of the recording date (April 8, 2026), John and Jason had collectively tested this across five different tenants spanning Canada Central, US East, Central US, and Europe — and none of them had it working yet. 403 and 404 errors across the board. It’s rolling out, it’s a public preview, and they’re not complaining. But if you listened to this episode the week it dropped and ran to try it: check again. By the time you’re reading this, it may well be working. The second it is, these two are on it.

John also surfaced two implementation paths: one that queries raw backend SharePoint data, and one that leverages SharePoint views. Jason’s hope is that the view-based path sticks around, since views give you a lot more control over what you’re exposing. Worth watching as this feature matures.

Excel to Delta Table Transformation via OneLake Shortcuts (Preview)

The second Excel/SharePoint item is the new Excel to Delta Table transformation in OneLake Shortcuts, now in public preview. This one requires a bit of terminology precision because mirroring and shortcuts are different things — and John took a moment to make sure that was clear.

Mirroring creates a continuous read-only replica. A shortcut is a pointer back to the source — the data stays where it is. But shortcut transformations with conversion to Delta are a hybrid: they actually move the data into OneLake and keep it in sync. So when you point a shortcut at an Excel file in a SharePoint document library, Fabric can pull all the sheets out of that file and land them as Delta tables in OneLake — near-instantly, because it’s reaching back to the source the way a shortcut does. Drop a new Excel file in the folder and the tables appear.

For context: shortcut transformations with CSV conversion went GA at FabCon. The Excel piece is the new preview capability on top of that. The data needs to be reasonably clean — you can’t do Power Query-style transformations on the way in — but for well-structured files, this removes the need for a pipeline entirely.

CI/CD: Finally Getting Good

The CI/CD improvements got genuine enthusiasm on the show — particularly from practitioners who’ve tried to set up branching workflows in Fabric before. The old experience had a lot of moving parts: workspace setup, branch connection, Git back-and-forth. It worked, but it required patience.

What’s new and in public preview: selective branching and branch hierarchy. Instead of branching an entire workspace, you can now pick just the items you want to work on — a semantic model and a report, for example — branch those, iterate, and merge back in. Child branches surface under the parent in the workspace list. The workspace diff view is now available inside Fabric itself, so you can see what’s changed before you commit rather than jumping over to GitHub after the fact.

The Fabric CLI for deployment (dev → test → prod in a single command) is now GA. Azure DevOps pipeline extensions and the Fabric extensibility toolkit also hit GA. John added the ISV perspective: deployability is king for any software vendor building on Fabric, and every improvement here is a direct unlock for that ecosystem.

Jason also called out variable libraries with connection properties as the feature that unlocked the most for him personally. The ability to point a semantic model at a different Lakehouse per workspace — having the connection swap automatically through deployment pipelines — is the kind of thing that makes proper environment isolation actually achievable. He described it as “magical.” Nobody else is doing this yet. They should be.

Fabric MCP Server: It’s More Than One Thing

The Fabric MCP server got some keynote attention, and John spent time clarifying something important: “Fabric MCP Server” is not one thing. It’s a family of things. There’s the Fabric MCP server for data engineering, the RTI MCP server, the Power BI MCP server — multiple specialized servers for different workloads. Jason’s framing was apt: saying “I use the Fabric MCP server” is like saying “I use Fabric” — the natural follow-up question is which part.

What got announced at FabCon: the local Fabric MCP server is now GA, and a remote hosted version is now in preview — meaning you no longer need to run an MCP server locally, manage tokens on your machine, or deal with local installation. Authentication flows through Fabric’s standard mechanism. That’s a meaningful quality-of-life improvement for day-to-day use.

Fabric Skills: John’s Favorite Announcement

The bigger deal for John personally — the thing he called his favorite FabCon announcement — is Fabric Skills. These are a set of pre-built agents and skills that know how to talk directly to the Fabric service APIs across all major workloads: Eventhouse, Data Warehouse, Data Engineering, Power BI, and more. They come with three agents and a broad set of skills and can be cloned into VS Code, where they work independently of needing an MCP server running locally.

What makes them different from standard MCP server usage is that they operate live against the service. You can ask questions about your data, make changes, run things, get results back — all in the context of your actual Fabric environment, not a local sync. John demoed this to himself the night before the episode: he found a broken notebook, described the problem, and had the agent diagnose and fix it without leaving VS Code. He also had it scan an entire workspace for empty tables, determine whether any of them were used anywhere, and clean them up. All through natural language. All live.

John’s endorsement was unambiguous. Jason is looking into it. This one is worth your time.

Tabbed UI Experience: GA

The tabbed UI experience in Fabric is now generally available. The general take: it’s a solid improvement and more fit-and-finish work is expected to continue as the experience matures. It’s the default navigation going forward.

Governance & Security: OneLake Security Coming to GA

The OneLake Catalog governed tab is now GA. More significantly: OneLake Security — table-level, column-level, and row-level security definitions that are honored across all Fabric engines — is coming to GA in the weeks following the keynote. This means security defined once in OneLake applies consistently whether you’re querying through T-SQL, Spark, KQL, or third-party engines via the new APIs.

The important caveat Jason highlighted: this is security you configure in OneLake. It does not read security from the source systems and propagate it automatically. You still have to define it in OneLake. That’s a logical next step that John acknowledged is coming — but it’s not here yet. The third-party API support means external querying engines can now integrate with OneLake security policies, which is a significant step toward true multi-engine governance.

Outbound Access Protection (OAP) has also been extended to 15 additional Fabric items, continuing the platform-wide rollout of that capability.

Mirroring Expands Significantly

Several new and updated mirroring sources were part of the FabCon announcements:

SharePoint Lists — public preview (covered above)
Dremio — public preview
Azure Monitor — coming soon (John’s personal most-anticipated; he’s actively building a monitoring solution that relies on this)
Oracle — generally available
SAP Datasphere — generally available
Change Data Feed (CDF) for mirrored sources — extended capabilities (paid option)
Mirroring Views on Snowflake — create views on top of mirrored data, starting with Snowflake (paid option)

John flagged that the extended capabilities (CDF and mirroring views) are paid add-ons on top of the core mirroring experience — worth understanding before building plans around them.

Delegated Shortcuts

Shortcuts have always used the consumer’s identity to authenticate back to the source. Delegated shortcuts, coming soon, allow you to specify an identity to use instead — a service principal, workspace identity, or similar. John called this out as important: it aligns shortcuts with how enterprise data access is supposed to work. Notably, this also supports shortcuts to OneLake data in another tenant — a wrinkle worth investigating further for multi-tenant organizations.

Azure Data Factory Migration to Fabric

The Azure Data Factory to Fabric migration assistant is now in public preview, and John described it as essentially feature-complete at this point. The big addition: SSIS transforms are included. Mapping Data Flows from ADF are coming to Fabric Data Factory by June 2026. If you’ve been waiting on SSIS or mapping data flow support before beginning your ADF migration, the window is getting short.

Analytics Engines: Runtime 2.0, Custom Live Pools, Materialized Views

Fabric Runtime 2.0 — built on Apache Spark 4.x, Delta Lake 4.x, Scala 2.13, and Azure Linux Mariner 3.0 — entered public preview at FabCon. This is the foundation for large-scale computation going forward. Materialized lake views went GA, simplifying medallion architecture in Spark SQL and PySpark with always-up-to-date pipelines and no manual orchestration.

Custom live pools are also new: the near-instant startup experience previously limited to standard Spark pools is coming to some custom pool configurations via pre-warming. Jason’s open question — and it’s the right one — is what the cost model looks like for keeping custom pools warm given Spark’s capacity unit cost. Worth watching.

Multimodal AI functions in notebooks (PDFs and images in Spark, in preview) round out the data engineering announcements.

Data Warehousing: Performance Milestone and New Capabilities

The warehouse team reported a 60% performance improvement over the last six months. Additional capabilities announced: custom SQL pools for workload isolation to eliminate noisy neighbor disruption, AI functions directly in T-SQL, and real-time alerts via Activator integration. These are meaningful improvements for teams running serious workloads in Fabric Data Warehouse.

Database Hub: A Single Pane of Glass for Your Entire Database Estate

Database Hub is new, in early access, and neither John nor Jason has had hands on it yet — but Jason described being “desperate” to try it. The pitch: a single unified view for your entire database estate, whether on-premises, IaaS, SaaS, or multi-vendor. Azure SQL, Cosmos DB, PostgreSQL, SQL Server (via Azure Arc), MySQL, Fabric SQL — all visible and manageable from one place.

John’s framing was clarifying: think of it as the engine-level equivalent of what OneLake shortcuts do for data. OneLake gives you a single logical view of data wherever it lives. Database Hub does the same for the database engines themselves. For organizations evaluating Fabric SQL alongside Azure SQL — which Jason is actively doing with a client — this is a capability worth tracking closely.

Also in the databases section: database agents, a Fabric SQL migration assistant for assessing SQL Server workload readiness, new enterprise security controls (auditing, customer-managed keys, dynamic data masking, availability zones), GitHub Copilot in SSMS 22 GA, SQL Hyperscale at 192 and 160 vCore options, and an Azure Cosmos DB agent kit (open-source, with AI coding assistance baked in). John also noted a keynote statement that Hyperscale infrastructure is coming to Fabric SQL — which would eventually close the gap between Fabric SQL’s current 32 vCore ceiling and what Azure SQL Hyperscale can do today.

Databricks and Snowflake Bidirectional Integration

The interoperability story at FabCon was strong. Snowflake bidirectional integration went GA. Databricks bidirectional integration — specifically the Unity Catalog component allowing native reads of OneLake data from Databricks — moved from public beta to public preview, with the reverse direction (writing from OneLake back to Databricks) in progress. The net result: the major data platforms are converging around OneLake as a shared data layer rather than everyone keeping their own copies.

Planning in Fabric: Watch This Space

A new workload called Planning was announced — think budgeting, forecasting, and scenario modeling natively in Fabric. Old-timers will recognize the itch it’s scratching: it’s the Performance Point planning story, brought forward with modern technology and a partner co-development model.

The reaction on the show was genuinely curious and cautiously optimistic. A FabCon attendee on John’s team had a striking reaction: they’d just purchased an ERP partly for this capability, and in hindsight might not have needed to. That’s a strong signal — and a bold claim worth watching. Enterprise planning carries real complexity: the number of variables, scenario modeling requirements, and finance team expectations are not trivial problems to solve. If Microsoft can deliver on the vision and make it cost-effective, this has the potential to be a meaningful market driver. The honest take on the show: the promise is real, and the proof will be in the execution. Both John and Jason are watching closely.

One naming note worth flagging: there’s a rumor circulating that Planning may be renamed “Planner” in Fabric — and Planner already exists in Microsoft 365. The show had some fun with Microsoft’s long history of naming things (Goals, anyone?), so consider this a gentle heads-up to watch the official naming before baking it into your documentation.

Fabric IQ, Data Agents GA, and Ontologies

Data Agents Are Generally Available

Data Agents went GA at FabCon — with enterprise upgrades including graph support, object-level access permissions, Microsoft Purview integration for audit and governance, and the ability to monitor Copilot and data agents through Purview. Jason’s been using data agents and is glad the GA milestone is here, since it means broader adoption can start. Integration into Microsoft 365 Copilot is now natively easier.

Operational Agents: New and in Preview

Operational Agents are new and in public preview. John has a practical use case already in flight: a comprehensive monitoring solution for Fabric, using agents to watch for specific conditions in Eventhouse — long ingestion times, failed queries, capacity pressure events — and take action automatically. He’s building it now. Jason raised an interesting question about whether the architecture (which appears to stitch together Power Automate and possibly Activator) will hold up compared to emerging agentic harnesses. Worth watching how the implementation matures.

Ontologies: Metadata for the Metadata

Fabric IQ Ontologies are read-only for now and were first introduced at Build, but got significant FabCon attention. The honest reaction on the show: the vision is starting to come into focus, even if it’s still early days for understanding the full practical value.

The concept: an ontology sits above your semantic models and data, creates an authoritative abstraction layer, establishes relationships between things, and makes it possible to do graph-style queries about how data elements relate to each other. The AI layer on top of all that is where it gets genuinely interesting — a well-built ontology makes it dramatically easier for AI to reason about your data estate accurately.

The practical question the show raised: ontologies require someone to define the metadata layer on top of your existing metadata. That’s a real adoption hurdle — most teams don’t reliably add metadata to their documentation today. The path to wide adoption probably runs through automation: an AI agent that scans your data estate, generates the ontology structure, and surfaces it for human validation and refinement. Once that’s the workflow, this gets genuinely interesting. Both John and Jason are aligned on that point completely.

Real-Time Intelligence: Eventstreams, Business Events, and Graph

Eventstreams with Spark Structured Streaming: you can now embed a full Spark notebook inside an Eventstream for processing streaming data. John has use cases he wants to explore immediately. The transformational capabilities inside Eventstreams have historically been limited — the SQL operator helped, and now having a full notebook in the pipeline is a significant step up.

SQL Operator fan-out: the SQL operator in Eventstreams previously could only output to a single destination. It can now fan out to multiple streams. Jason hadn’t caught this one and called it out as interesting.

Business Events and Event Schemas: a new concept allowing you to define custom event schemas at the business level that watch for changes to your data and stream those changes through an Eventstream. John hasn’t dug into this deeply yet, but Jason immediately connected it to real operational use cases — watching customer signals in real-time, predicting churn before it shows up in reporting. The shift from rear-view mirror to lead indicator is exactly where real-time intelligence is most valuable.

Capacity Events in the Real-Time Hub: capacity changes and pressure events now surface in the Real-Time Hub, letting you take action before throttling kicks in. Related: capacity overage protection is a new platform-level option that lets you authorize paying for overage usage rather than getting throttled — a meaningful operational lever for teams running sustained heavy workloads.

Graph in Eventhouse: the existing graph query capability has been significantly upgraded — billion-node graphs with multi-hop connections, designed for relationship-heavy data like customer networks, supply chains, and partner ecosystems. Coming to GA in the weeks following the keynote.

Maps in Fabric: GA

Maps in Fabric went GA. The show connected this directly to the ontology and Fabric IQ story — geospatial awareness as part of the intelligence layer, not just the visualization layer. Azure AI Foundry can now connect directly to OneLake as a data source, tying the broader AI ecosystem into Fabric’s data layer as the final item on the keynote list.

What’s Coming Next

There’s more to come from FabCon. John and Jason explicitly acknowledged that the March 2026 Feature Summary blog post — released during FabCon week — deserves its own deeper coverage, and they’re planning to pull out the items from that post that didn’t get covered in the keynote recap.

Links

Microsoft Blog Posts Referenced in This Episode

Previous Episodes

Events

Subscribe: SoundCloud | iTunes | Spotify | TuneIn | Amazon Music

Tags: Azure Data Factory, CI/CD, data agents, Database Hub, Databricks, Eventstreams, Excel Delta Table, FabCon 2026, FabCon Atlanta, Fabric CLI, Fabric IQ, Fabric MCP server, Fabric Skills, Feature Summary, Graph, keynote recap, Microsoft Fabric, Mirroring, OneLake, OneLake Security, Ontologies, Operational Agents, Planning Fabric, Power BI, Real-Time Intelligence, Runtime 2.0, selective branching, SharePoint List Mirroring, Snowflake, SQLCon 2026, tabbed UI, Variable Libraries