Microsoft Fabric June 2026 Update: SQL Endpoint Refresh & Semantic Model Agent API Introduced
microsoft-fabric power-bi sql-endpoint semantic-model-api
trending_up Trend: microsoft-fabric

Microsoft Fabric June 2026 Update: SQL Endpoint Refresh & Semantic Model Agent API Introduced

calendar_month June 25, 2026

Summary

The June 2026 update for Microsoft Fabric and Power BI introduces critical enhancements for data architects and AI developers. Key among these are the introduction of a native “Refresh SQL Endpoint” activity in Data Factory pipelines, resolving the unpredictable “wait-and-hope” lag for metadata synchronization between Lakehouses and SQL analytics endpoints. Additionally, Power BI semantic models are now accessible as direct, governed APIs for AI agents (such as Microsoft 365 Copilot and custom Fabric apps), facilitated by the general availability of DAX User-Defined Functions (UDFs).

What happened?

The June 2026 release addresses key orchestration and integration bottlenecks across the Fabric ecosystem:

  • “Refresh SQL Endpoint” Activity GA: Data Factory pipelines can now trigger an explicit refresh of the Lakehouse SQL analytics endpoint. This replaces the previous asynchronous background sync that forced data teams to insert arbitrary “Wait” activities. The sync now operates under a 30-second Service Level Objective (SLO) for Delta log metadata.
  • Semantic Models as Agent APIs: Power BI semantic models are exposed as secure, structured answer engines for AI agents. Agents can query business logic, metrics, and relationships directly without querying raw data or fabricating calculations.
  • DAX UDFs Reach General Availability: Developers can now define reusable, typed, and parameterized functions within DAX, making code maintainable and providing a structured framework for agents.
  • Resumed PBIR Format Rollout: Microsoft has restarted the default-on rollout of the Power BI Report (PBIR) format in the Service, storing reports as machine-readable code for clean Git integration and CI/CD pipelines.

Why it matters

These features bridge the gap between traditional Business Intelligence and agentic AI integration, introducing three major shifts:

  1. Deterministic Orchestration: Replacing arbitrary pipeline wait steps with direct endpoint refresh calls creates faster and more reliable end-to-end data loads.
  2. Governed AI Access (Trust): Instead of allowing LLMs to run raw SQL queries and hallucinate calculations, AI agents can leverage verified, certified measures defined in semantic models.
  3. True Code-First BI: The convergence of TMDL, UDFs, and the PBIR format ensures that BI assets behave like standard software engineering artifacts, complete with Git versioning.

Evidence

  • Climber BI Blog: Explains how the new “Refresh SQL Endpoint” activity solves the Lakehouse-to-SQL sync lag and details the 30-second metadata synchronization SLO.
  • Microsoft Fabric Updates Blog: Features detailed announcements regarding Data Factory enhancements, including the preview of the “Refresh Materialized Lake View” activity and the sp_get_table_health_metrics stored procedure.
  • LinkedIn Masterclass / Developer Community: Discussions detailing the utilization of DAX UDFs and how developers are utilizing semantic models as governed API endpoints for Microsoft 365 Copilot.

Analysis

Treating semantic models as APIs represents a major evolution. Historically, BI models were visual-first endpoints designed solely for human dashboards. By transforming them into queryable API layers, they become the central “semantic layer” and truth engine for autonomous AI systems. However, developers must proceed with caution regarding security. The interaction between Object-Level Security (OLS) and DAX UDFs has documented gaps where OLS restrictions do not automatically apply to the UDF layer. This presents a critical vulnerability if sensitive datasets are queried via agent APIs without proper validation.

Practical Takeaways

  1. Remove Pipeline Wait Steps: Refactor Data Factory pipelines by removing arbitrary “Wait” activities and inserting the native “Refresh SQL Endpoint” activity immediately following Lakehouse writes.
  2. Adopt DAX UDFs: Standardize business math using DAX UDFs. Document them with /// comments to enable self-discovery by both human developers and AI models.
  3. Audit Object-Level Security: Before exposing semantic models to Microsoft 365 Copilot or external APIs, verify that Object-Level Security (OLS) and Row-Level Security (RLS) rules execute correctly under UDF computations.
  4. Transition to PBIR: Save reports in the PBIR format to establish clean source control, merge conflict resolution, and automated deployment pipelines.

Open Questions

  • How quickly will Microsoft resolve the Object-Level Security (OLS) inheritance gaps in DAX UDFs to ensure full enterprise compliance?
  • When will the PBIR report format become the default save option in Power BI Desktop without needing preview flags?
  • What are the performance and cost implications on Fabric Capacity (F-SKUs) when autonomous agents begin making high-frequency API queries to semantic models?

Sources

  1. Power BI June 2026 Update: Turning Semantic Models into Agent APIs
  2. Climber BI Blog: Removing the “Wait and Hope” Pipeline Step
  3. Microsoft Community: Power BI June 2026 Feature Summary
  4. Reddit: Discussion on Semantic Model Agent APIs
  5. Microsoft Learn: What’s New in Microsoft Fabric Fundamentals