FabCon Atlanta 2026: Every Announcement
Complete coverage of 30+ announcements from the combined FabCon + SQLCon event: Fabric IQ, OneLake Security, Runtime 2.0, Database Hub, and more.
Fabric is no longer an analytics platform. It's the control plane for the entire enterprise data estate. This reference covers every announcement from the combined FabCon + SQLCon event (March 16–20, Atlanta), with previous state context and official citations.
The strategic reframing
Microsoft is positioning Fabric as the operating system for data, context, and action, unifying databases, analytics, governance, and AI agents. The co-location of SQLCon signals that the boundary between "database people" and "analytics people" is deliberately being collapsed.
Master Announcement Table
| Announcement | Status | Previous State | Category |
|---|---|---|---|
| Fabric IQ | Preview | Announced at Ignite 2025 (preview). Expanded at FabCon. | AI |
| Ontologies Workload | Preview | DNE. Entity/relationship modeling was ad-hoc. | AI |
| Graph in Fabric | GA Soon | Preview (Ignite 2025). | AI |
| Fabric Data Agents | GA | Preview, F64+ → F2+ (Apr 2025). | AI |
| Fabric Remote MCP | Preview | DNE. | AI |
| Fabric Local MCP | GA | DNE. | AI/DevOps |
| Maps in Fabric | GA | FabCon Europe '25: Preview. Ignite '25: additions. | Analytics |
| Fabric CLI v1.5 + deploy command | GA | v1.x: basic workspace mgmt. | DevOps |
| Variable Library: Item Reference | GA | String-based GUIDs. | DevOps |
| Runtime 2.0 | EPP | Runtime 1.3 (Spark 3.5, Delta 3.2, Python 3.11, Java 11). | Eng |
| Materialized Lake Views | GA | Preview (Build 2025). | Eng |
| Notebook Public APIs | GA | Manual portal management. | Eng |
| Mapping Data Flows → Fabric | June 2026 | ADF-only. Not available in Fabric. | Factory |
| OneLake Catalog: Auto-Summary, Parent-Child | GA | Basic catalog. No AI summaries. | Governance |
| Databricks Unity Catalog → OneLake | Preview | One-way (OneLake→DBX) only. | Interop |
| Snowflake Interop | GA | Preview. | Interop |
| Self-Service Workload Publishing | GA | Required Microsoft involvement. | ISV |
| Fabric Migration Assistants (ADF, Synapse, SQL) | GA | FabCon LV 2025: Synapse only. | Migration |
| Mirroring: Oracle, SAP | GA | Oracle PV (Build '25). SAP PV (Ignite '25). | OneLake |
| Mirroring: SharePoint, Dremio | Preview | DNE. | OneLake |
| Mirroring: CDF + Views (Snowflake) | Preview | Basic mirroring only, no change feed. | OneLake |
| Shortcut Transformations | GA | Preview (FabCon LV 2025). | OneLake |
| OneLake in Windows File Explorer | GA | Portal-only access. | OneLake |
| Direct Lake on OneLake | GA | Preview (Ignite 2024). Desktop authoring (Mar 2025). | PBI |
| Translytical Task Flows | GA | Preview. | PBI |
| Eventstreams Deltaflow | Preview | IoT/streaming sources only. | RTI |
| OneLake Security | GA Soon | Private preview (FabCon LV 2025). | Security |
| DLP for Structured Data (OneLake) | Preview | DLP for unstructured only. | Security |
| Purview DSPM for AI | Preview | DNE. | Security |
| SPN & Workspace Identity (all activities) | GA | Limited activity support. | Security |
| SQL Server 2025 EXPOSE TO FABRIC | Preview | DNE. ETL-based integration only. | SQL |
| Database Hub in Fabric | Early Access | DNE. Databases managed in separate Azure Portal blades. | SQL/DB |
Use the tabs above to explore announcements by area - OneLake, AI & Fabric IQ, Engineering, and more; with citations.
OneLake: The Unified Data Lake
OneLake remains the gravitational center: the single logical data lake unifying multi-cloud data without ETL. FabCon Atlanta dramatically expanded reach, governance, and discoverability.
Mirroring Expansion: 7+ New/Updated Sources
Zero-ETL with auto CDC, schema evolution. New: Change Data Feed and views on mirrored data (Snowflake).
Shortcut Transforms
Auto-shape data as it connects to OneLake. No separate pipeline needed.
OneLake Security
Define roles once → enforced in Spark, T-SQL, Power BI, data agents. Industry-first for lakehouses.
Interoperability: Multi-Platform Open Access
Bi-directional. Multi-platform estates use OneLake as the shared layer without ETL.
Windows File Explorer
Business users access enterprise data lakes through the familiar file system.
OneLake in Microsoft Foundry
AI builders using Foundry can discover and consume OneLake data natively.
Catalog: Auto-Summary
Quick understanding of unfamiliar items. Regenerated on each visit for freshness.
Catalog: Parent-Child Structure
Understand how data items relate without digging into each one.
Variable Library: Item Reference
Stronger security (authorized items only), meaningful UI display (name, type, location), resilient cross-environment deployment.
Core thesis: AI agents are only as reliable as the data you feed them
Every AI announcement traces to grounding agents in trusted, semantically rich enterprise data. Fabric IQ, Ontologies, semantic models, Graph, and MCP are all agent reliability infrastructure.
Semantic Intelligence Layer for AI
An umbrella semantic intelligence layer unifying five capabilities: (1) Semantic Models, now extended beyond BI into AI grounding; (2) Graph model for multi-hop reasoning; (3) Data Agents for conversational analytics; (4) Operations Agents for real-time autonomous action; and (5) Ontologies, the only brand-new component. Part of Microsoft's IQ layer (alongside Work IQ for M365, Foundry IQ for AI Foundry).
Domain Entity Modeling
New workload for domain-level entity modeling beyond what semantic models offer. Includes operations agents that monitor the business in real time. Defines the conceptual language Fabric IQ uses to train agents.
Conversational Analytics at F2+
Conversational Q&A over warehouses, lakehouses, semantic models. Domain-specific virtual analysts.
Enterprise Graph Database
Scalable graph DB for customers, supply chains, knowledge graphs. Powers Fabric IQ's ontology layer.
Local MCP Server
Manage permissions, create workspaces, leverage item definitions from AI coding assistants.
Remote MCP Server
Authenticated actions in Fabric from any AI agent platform. RBAC-compliant.
Notebook Custom Agent
Unlike generic coding assistants, this agent operates with full awareness of your Fabric workspace, runtime, environments, and Lakehouse resources. Available in the Fabric Data Engineering VS Code extension.
Data Engineering & Compute
The biggest compute upgrade since Fabric GA. Runtime 2.0 and Materialized Lake Views fundamentally change pipeline architecture.
Runtime 2.0 (Experimental Public Preview)
Next-gen compute foundation. Spark 4.0 brings significant performance improvements and new APIs. Note: this is an Experimental Public Preview (EPP), a lower maturity tier than regular preview. Not recommended for production workloads.
Materialized Lake Views
Define bronze→silver→gold declaratively. System keeps views current with no manual orchestration. Eliminates massive boilerplate ETL.
Notebook APIs
CI/CD integration, automated testing, IaC patterns for data engineering teams.
Eventstreams Deltaflow
Event-driven architecture on operational DB changes. Also: stream processing with Spark notebooks (PV).
Breaking changes
Scala 2.13 is not binary-compatible with 2.12, and Java 21 replaces Java 11. Test all library dependencies before migrating production workloads.
Security & Governance
OneLake Security
Unified row/column-level ACLs across Spark, T-SQL, Power BI, agents. Define once, enforce everywhere.
DLP for Structured Data
Purview DLP detects sensitive data in structured OneLake tables. Auto-restrict access on detection.
DSPM for AI
Govern AI interactions: visibility into prompts/responses, detect sensitive data risks in agent workflows.
Workspace Tags
Tag workspaces. Filterable in lists + OneLake Catalog. API-managed at scale.
Purview IRM PAYG Report
Transparency: see CU billing distribution across Fabric workloads (Power BI, Lakehouse). Tune policies to control costs.
Fabric Data Factory
23B+ orchestration runs/month. FabCon Atlanta focused on enterprise hardening and the critical ADF migration path.
Outbound Access Protection
Pipelines, Dataflows Gen2, Copy jobs connect only to approved destinations. Defense-in-depth with private endpoints + workspace policies.
Mapping Data Flows → Fabric (June 2026)
Teams relying on MDF in Azure Data Factory can modernize to Fabric without rewriting transform logic. This is the #1 blocker for many ADF-to-Fabric migrations.
Dataflow Gen2 Enhancements
New destinations: Snowflake, Excel Files (PV). Save As Dataflow Gen2 from Gen1. Recent Tables for faster source selection. Streaming API support.
First-ever SQLCon co-located with FabCon
50+ dedicated SQL sessions, 4 workshops, 30+ Microsoft SQL product team members on-site. The message: the boundary between OLTP and OLAP is becoming less meaningful. Databases now participate directly in AI and analytics workflows.
Unified Database Management
Collapses fragmented Azure Portal blades into a single view. Covers Azure SQL, Cosmos DB, PostgreSQL, SQL Server (via Azure Arc), and Fabric Databases. Built-in observability, delegated governance, human-in-the-loop AI agents.
EXPOSE TO FABRIC
New T-SQL extension creating virtualized representations of database objects in OneLake, without moving data. Intelligent Query Processing works with Fabric's compute engine. On-premises databases participate via Azure Arc without cloud migration.
SQL Database in Fabric: Next Wave
Reached GA at Ignite November 2025. FabCon Atlanta builds on: Simplified (easier provisioning), Autonomous & Secure (self-tuning), Optimized for AI (closer to agents). Adoption accelerating since GA.
Savings Plans for Serverless
Azure SQL Database serverless customers can now access savings plans with estimated savings between 0–35% for 1-year commitments.
Power BI at FabCon Atlanta
Power BI serves 35M+ active users. The FabCon theme for PBI: semantic models are no longer just BI infrastructure; they're AI agent grounding infrastructure.
Direct Lake on OneLake
Tables read directly from OneLake. BI teams get import-mode speed on lakehouse-scale data without scheduled refreshes.
Translytical Task Flows
Business users act without submitting requests or leaving the BI surface. Closes the insight-to-action gap.
Semantic Model AI Enhancements
New features for Copilot in PBI and data agents: verified answers (curated Q&A pairs), AI instructions (model-level guidance for agent behavior). Foundation for Fabric IQ.
Copilot & AI: March Improvements
Improvements across Copilot, reporting, modeling, and data connectivity. Focus on making everyday authoring smoother and visuals more polished.
Maps in Fabric
First native geospatial capability in the platform. Location-aware analytics without third-party tools.
PBI DataViz World Championship
Four finalists competed live on stage. Winner: Gustavo Bavia. Community celebration of the craft of data visualization.
Deprecation: PBI Scorecard Hierarchies removed April 15, 2026
The Hierarchies feature in Power BI Scorecards (multi-level structure: region, division, team) will be removed. Plan migration if you depend on auto-generated filtered scorecard views.
Ecosystem, ISV & Migration
Microsoft's thesis: platforms win when ecosystems thrive. FabCon Atlanta expanded the ISV, interoperability, and migration stories significantly.
Self-Service Workload Publishing
ISV partners can now deliver custom workloads inside Fabric without gatekeeping. GA workloads: ESRI, Lumel, Neo4j, Osmos, Profisee, PowerBI.tips. Previews: Celonis, CluedIn, Statsig, Striim.
Azure Databricks at FabCon
Lakebase GA (DB for AI agents), Lakeflow Connect Free Tier (100 free DBUs/day), Genie Code for agentic data engineering, new Excel add-in and Teams integration. Unity Catalog → OneLake now in preview.
Auger: Supply Chain on Fabric
Auger built its entire supply chain platform on Fabric with all data in OneLake. Customers access operations data via OneLake shortcuts within their own Fabric environments, including Power BI, agents, and full platform capabilities.
Snowflake Interop
Snowflake interoperability generally available. Mirroring with views and Change Data Feed now in preview for Snowflake specifically.
Persona x Announcement Matrix
Every announcement mapped to the persona it impacts, the use case it unlocks, and the prior pain point it resolves.
Data Engineer
Runtime 2.0 (EPP), MLVs GA, Notebook APIs, Deltaflow, CLI v1.5, Variable Library Item Ref, Notebook Custom Agent
Declarative medallion. CI/CD-driven notebook deployment with OIDC. Event-driven on DB changes. Spark 4.0 perf (EPP). Typed item refs.
Manual orchestration. No notebook APIs. Stuck on Spark 3.5. CLI lacked deploy. GUIDs as strings.
Analytics/BI Engineer
Direct Lake on OneLake GA, Fabric IQ (PV), Semantic model AI, Maps GA, Translytical Task Flows GA, Ontologies
Import-speed on lakehouse data. Semantic models as agent grounding. Geospatial analytics. Users act from reports.
Direct Lake limited. Models served BI only. No maps. Read-only reports. No ontology layer.
DBA / Platform Admin
Database Hub EA, OneLake Security, SQL Server 2025 EXPOSE TO FABRIC, Mirroring expansion, SQL DB updates, Azure SQL savings
Single-pane DB observability. Unified security across all engines. Zero-copy SQL Server → OneLake. Zero-ETL from Oracle/SAP.
Separate portals per DB. Per-engine security. ETL required. No savings plans.
Data Scientist
Runtime 2.0 (EPP), Graph in Fabric, Remote MCP, AI functions, Notebook Custom Agent, Ontologies
Graph-based feature engineering. Multi-agent execution in cloud. LLM transforms on OneLake data. Workspace-aware code gen.
No graph engine. No remote agent exec. AI transforms required external services. Basic Copilot.
Business Analyst
Data Agents GA (F2+), Fabric IQ, File Explorer, Copilot in PBI, Translytical Task Flows, Catalog Auto-Summary
Conversational Q&A in natural language. Browse data via Windows. AI-powered insights without DAX/SQL. Act from reports.
Data access required SQL. No conversational interface (or limited to F64+). Portal-only. Read-only reports.
Security / Governance
OneLake Security GA, DLP structured data, DSPM for AI, Workspace Tags, OAP, Purview IRM PAYG report
Unified ACLs across engines. Auto-restrict sensitive data. Govern AI prompts/responses. Tag-based discovery. CU billing transparency.
Per-engine ACLs. No structured DLP. No AI governance. Manual classification. No billing breakdown.
DevOps / Platform Eng
CLI v1.5 deploy + OIDC, Local MCP GA, SPN/WI GA, Notebook APIs, Variable Library Item Ref, Migration Assistants
Zero-touch Git deployments. AI assistants connected to Fabric. Identity-based access everywhere. IaC for notebooks.
CLI lacked CI/CD. No MCP. Manual identity per activity. Manual portal management. String GUIDs.
ISV / Partner
Self-Service Workload Publishing GA, Remote MCP, OneLake interop (Databricks, Snowflake), Auger case study
Publish custom workloads without gatekeeping. Connect agents to Fabric from any platform. Bi-directional data access.
Required Microsoft involvement. No MCP standard. One-way interop. ISV patterns not documented.
| Persona | Key Announcements | Use Cases Unlocked | Previous Pain |
|---|---|---|---|
| Data Engineer | Runtime 2.0 (EPP), MLVs GA, Notebook APIs, Deltaflow, CLI v1.5, Variable Library Item Ref, Notebook Custom Agent | Declarative medallion. CI/CD-driven notebook deployment with OIDC. Event-driven on DB changes. Spark 4.0 perf (EPP). Typed item refs. | Manual orchestration. No notebook APIs. Stuck on Spark 3.5. CLI lacked deploy. GUIDs as strings. |
| Analytics/BI Engineer | Direct Lake on OneLake GA, Fabric IQ (PV), Semantic model AI, Maps GA, Translytical Task Flows GA, Ontologies | Import-speed on lakehouse data. Semantic models as agent grounding. Geospatial analytics. Users act from reports. | Direct Lake limited. Models served BI only. No maps. Read-only reports. No ontology layer. |
| DBA / Platform Admin | Database Hub EA, OneLake Security, SQL Server 2025 EXPOSE TO FABRIC, Mirroring expansion, SQL DB updates, Azure SQL savings | Single-pane DB observability. Unified security across all engines. Zero-copy SQL Server → OneLake. Zero-ETL from Oracle/SAP. | Separate portals per DB. Per-engine security. ETL required. No savings plans. |
| Data Scientist | Runtime 2.0 (EPP), Graph in Fabric, Remote MCP, AI functions, Notebook Custom Agent, Ontologies | Graph-based feature engineering. Multi-agent execution in cloud. LLM transforms on OneLake data. Workspace-aware code gen. | No graph engine. No remote agent exec. AI transforms required external services. Basic Copilot. |
| Business Analyst | Data Agents GA (F2+), Fabric IQ, File Explorer, Copilot in PBI, Translytical Task Flows, Catalog Auto-Summary | Conversational Q&A in natural language. Browse data via Windows. AI-powered insights without DAX/SQL. Act from reports. | Data access required SQL. No conversational interface (or limited to F64+). Portal-only. Read-only reports. |
| Security / Governance | OneLake Security GA, DLP structured data, DSPM for AI, Workspace Tags, OAP, Purview IRM PAYG report | Unified ACLs across engines. Auto-restrict sensitive data. Govern AI prompts/responses. Tag-based discovery. CU billing transparency. | Per-engine ACLs. No structured DLP. No AI governance. Manual classification. No billing breakdown. |
| DevOps / Platform Eng | CLI v1.5 deploy + OIDC, Local MCP GA, SPN/WI GA, Notebook APIs, Variable Library Item Ref, Migration Assistants | Zero-touch Git deployments. AI assistants connected to Fabric. Identity-based access everywhere. IaC for notebooks. | CLI lacked CI/CD. No MCP. Manual identity per activity. Manual portal management. String GUIDs. |
| ISV / Partner | Self-Service Workload Publishing GA, Remote MCP, OneLake interop (Databricks, Snowflake), Auger case study | Publish custom workloads without gatekeeping. Connect agents to Fabric from any platform. Bi-directional data access. | Required Microsoft involvement. No MCP standard. One-way interop. ISV patterns not documented. |
Top Actions by Priority
| Action | Persona | Priority |
|---|---|---|
| Enable OneLake Security when GA to replace per-engine ACLs | Security, DBA | Critical |
| Migrate manual medallion pipelines to Materialized Lake Views | Data Engineer | High (GA) |
| Test Runtime 2.0 (EPP): validate Scala 2.13 + Java 21 library compat | Data Engineer | High |
| Pilot Fabric IQ (preview) + Ontologies with existing semantic models | BI Engineer, Data Scientist | High |
| Adopt CLI v1.5 deploy + OIDC in CI/CD pipelines | DevOps | High (GA) |
| Plan ADF migration: evaluate Mapping Data Flows → Fabric (June '26) | Data Engineer, DevOps | High, plan now |
| Request Database Hub Early Access and inventory estate | DBA | Medium |
| Deploy Local MCP to connect GitHub Copilot to Fabric | DevOps, Eng | Medium |
| Evaluate EXPOSE TO FABRIC for SQL Server 2025 workloads | DBA | Medium |
| Test Direct Lake on OneLake for existing PBI models | BI Engineer | Medium (GA) |
| Note PBI Scorecard Hierarchies deprecation (April 15) | BI Engineer | Action needed |
The pattern across all personas
Unify → Govern → Ground Agents. Unify data (mirroring, shortcuts, Database Hub). Govern centrally (OneLake Security, DLP, DSPM). Ground AI agents (Fabric IQ, semantic models, Ontologies, MCP). This isn't a feature list. It's an architecture thesis for the AI-ready enterprise.