Announced at Microsoft Ignite two weeks ago were many new product features related to the data platform. Check out the Major announcements and Book of News. I went through the many announcements and picked out the most interesting ones, so you don’t have to:

Microsoft SQL Server 2025, with built-in AI and developer-first enhancements, is now generally available. The platform enables customers to securely use data they already have and work in the familiar T-SQL language. It provides:

• A way to access AI models of choice, hosted locally or in the cloud, and to securely use data to best fit business needs.
• Simplified data processing with native JSON support, built-in REST APIs and change event streaming.
• Near real-time analytics by replicating SQL Server data to Microsoft OneLake with database mirroring in Microsoft Fabric.
• Increased workload performance, uptime and concurrency for SQL Server apps with enhanced query optimization, optimized locking and improved failover reliability.
• Improved credential management and fewer potential vulnerabilities with Microsoft Entra ID for authentication through Microsoft Azure Arc.
• GitHub Copilot integration in Visual Studio Code and SQL Server Management Studio 22 for better productivity.
• A new Microsoft Python driver for SQL Server (mssql-python) for a fast and developer-friendly experience in Windows, macOS and Linux.

(more info) (Bob Ward post)

Microsoft Azure DocumentDB now generally available
Microsoft Azure DocumentDB, the first managed service built on the open-source DocumentDB standard, is generally available. Now governed by the Linux Foundation, DocumentDB delivers an open and community-driven MongoDB-compatible engine with multicloud flexibility, running consistently across Azure, other clouds and on premises. This gives organizations freedom from proprietary lock-in and the ability to standardize on open source while operating at a global scale. Azure DocumentDB was previously known as Azure Cosmos DB for MongoDB (vCore). Note that this new Azure DocumentDB is not the same product as the original DocumentDB that became Cosmos DB—it’s a newly branded, separate vCore-based MongoDB service. (more info)

Azure HorizonDB, a new PostgreSQL database, in private preview
Microsoft Azure HorizonDB, a new PostgreSQL cloud database service for building or modernizing mission-critical apps, is now in private preview. Integrated with Microsoft Foundry, Microsoft Fabric, Visual Studio Code and more, Azure HorizonDB streamlines development with the following features:

• Transactions and vector search up to three times faster than open-source PostgreSQL, based upon internal benchmarking.
• Scale-out compute to 15 replicas with 192 vCores each.
• Auto-scaling storage up to 128 TB.
• Advanced DiskANN vector indexing for AI workloads and native semantic operators.
• AI-readiness with pre-provisioned models.

Organizations can right-size consumption to their workloads’ needs and save capacity for future requirements by independently scaling compute and storage scale. Modern authentication with Microsoft Entra ID and security features like Microsoft Defender and private endpoints support enterprise-grade protection. It is optimized for performance, scale, and AI apps and designed to compete directly with AWS Aurora and GCP AlloyDB. (more info)

Microsoft Fabric databases, now generally available, bring together SQL database and Cosmos DB in a new, unified software-as-a-service (SaaS) experience for organizations to manage, analyze and activate their data. Fabric databases provide instant provisioning, autonomous architecture, enterprise-grade security and native AI integration — including support for vector data and retrieval-augmented generation (RAG) patterns — to help teams build intelligent, real-time apps. (more info)

Announcing Microsoft Fabric IQ: The Semantic Intelligence Platform
Microsoft Fabric IQ is the new semantic intelligence layer that elevates Fabric from a unified data platform to a unified intelligence platform. It turns your unified data estate, already consolidated in OneLake, into a live, structured, connected model of how your business operates. It bridges the gap between where your data lives and how your teams and AI reason, decide, and act.

Fabric IQ combines five integrated capabilities into one semantic intelligence system:

• Ontology (preview): shared model of business entities, relationships, rules, and objectives. This was just released at Ignite.
• Semantic Model: trusted BI definitions, now extended beyond analytics into operations and AI. Semantic models have been available for a long time.
• Graph model (preview): native graph engine for multi-hop reasoning and system-wide insights. This was released a couple of months ago.
• Data Agent (preview): virtual analysts that answer business questions using structured business meaning. This has been available for about a year.
• Operations Agent (preview): autonomous agents that reason, learn, and act in real time to advance outcomes. This has been available for about a year.

So to clarify: Fabric IQ is a new “umbrella” name for five features, only one of which is brand new – Ontology.

More about Ontology:

An ontology in Microsoft Fabric is a shared, machine-understandable vocabulary that defines the core business concepts that exist within an organization—things like customers, products, orders, and assets, rather than just the raw tables that store data. It provides a business-level semantic layer that standardizes terminology across domains, ensuring all teams and tools refer to the same entity names, properties, and relationships. This eliminates inconsistencies that commonly arise when different teams model the same concepts in different ways.

Once created, the ontology is bound to actual data sources in Fabric, including lakehouse tables, event streams, and semantic models. This binding process maps columns to properties, links identifiers to relationships, and transforms raw table rows into typed entity instances enriched with consistent semantics, metadata, provenance, and timestamps. The ontology becomes the layer through which physical data is interpreted, giving Fabric a unified view of business meaning across disparate systems.

When Graph in Microsoft Fabric is enabled, the ontology is materialized as a graph where each entity instance becomes a node and each relationship becomes an edge. This graph representation allows visual exploration (e.g. browsing relationships, lineage, dependencies), graph-style queries or algorithms (e.g. pathfinding, impact analysis) and also semantic queries: you ask in terms of business concepts (not tables) — e.g. “Find all shipments that are delayed and associated with high-risk routes” — and Fabric handles the underlying joins, filtering, data reconciliation. This enables a far richer and more intuitive way to navigate enterprise data.

Using an ontology brings multiple benefits over traditional table-centric models. It enforces cross-domain consistency and governance by providing a single semantic standard for the entire organization. It simplifies integration by allowing data from different sources—batch, real-time, structured, or semantic—to be unified under the same conceptual model. Its semantic richness enables deeper context and expressiveness, supporting advanced analytics and reasoning that are difficult or impossible with flat tables.

Finally, because the ontology abstracts business meaning away from physical schemas, teams gain flexibility and agility. Changes to underlying tables or new data sources can often be accommodated simply by rebinding them to the ontology, without rewriting analytics or business logic. Combined with Fabric’s graph and AI capabilities, the ontology becomes a powerful foundation for agents, copilots, and intelligent applications that interact with data in terms of real business concepts.

To create an ontology in Microsoft Fabric, you begin by defining the ontology itself—either manually or by generating it from an existing semantic model. When generated, each table is converted into an entity type, its columns become properties, and any existing relationships are turned into ontology relationships. After the initial creation, the next step is to refine and rename these entity types so they reflect business-friendly concepts rather than raw technical table names, such as transforming “dimproduct” into “Product” or “factsales” into “SaleEvent.”

With the conceptual model established, you then bind your actual data sources to the ontology. This involves mapping tables or event streams to the appropriate properties and relationships, defining identifiers and keys, and handling differences between time-series and static data. If desired, you can then enable Graph support, which turns the ontology into a first-class graph structure. This unlocks graph traversal, lineage insights, dependency analysis, and more advanced graph operations.

Finally, once the ontology is bound and optionally graph-enabled, you can query your data through the business-level semantic layer rather than through raw tables. Queries are expressed using business concepts—such as “Show all Orders for Customer X between date Y and date Z”—and Fabric handles the necessary joins, filters, and reconciliations. If natural-language querying is enabled, users can interact with the data even more intuitively, relying on the ontology to interpret their intent.

In short, Fabric ontology is a business-level semantic layer you create over your data that standardizes meaning across the organization and makes it far easier for users—and AI agents—to ask questions and get answers in business terms rather than table structures. It unifies data from disparate sources, represents it as a connected graph when enabled, and provides a consistent, intelligent foundation for analytics, exploration, and automation. It differs from a sematic model in that a semantic model describes how your data is stored, while an ontology describes what your business actually is.

Fabric OneLake and Databricks integration announcements: Mirroring data into OneLake – already generally available; By the end of 2025, Azure Databricks will enable native reading from OneLake through Unity Catalog in preview, allowing users to seamlessly access data stored in OneLake without duplication or complex pipelines; Looking ahead, Azure Databricks will support writing and storing data directly in OneLake, without any additional storage resources to manage. 

Preview of OneLake shortcuts to SharePoint and OneDrive, allowing you to bring your unstructured, productivity data into OneLake without copying files or building custom ETL flows. 

Fabric capacity overage/expanding surge protection – To help you gain control over the jobs running on your Fabric capacities, Microsoft is expanding surge protection and introducing a new tool called Fabric capacity overage—both of which will be released into preview in Q1 2026—and adding Fabric capacity events in the Real-Time hub. First, surge protection will now let you set limits on specific workspace activity to protect your capacities from unexpected surges from non-critical workspaces.  Also to be released is Fabric capacity overage which admins can turn on for specific capacities, allowing them to automatically pay for excess consumption and avoid throttling whenever high-traffic periods occur. Rather than over-provisioning for rare spikes, you can right-size your capacity for typical usage and enable overage only when needed. Admins can even set a 24-hour limit so you don’t break your budget, and the feature can be toggled on or off in seconds. These tools are designed to work together to help you prevent over-use and maintain smooth, uninterrupted operations even during peak demand.

Data Clustering in Fabric Data Warehouse (Preview) – Fabric Data Warehouse introduces data clustering capabilities to optimize query performance and reduce storage costs through intelligent data organization

IDENTITY columns (Preview) in Fabric Data Warehouse, a long-awaited feature that simplifies surrogate key generation during data ingestion. IDENTITY columns automatically produce unique values for each new row, eliminating the need for manual key assignments and eliminating the risk of key duplication and key integrity issues. (more info)

Fabric Warehouse Snapshots have GA’d. Create read-only views of your warehouse at a specific point in time. (more info)

Mirroring for SQL Server in Fabric for all in-market versions of SQL Server from SQL Server 2016 to SQL Server 2025 is Generally Available. (more info)

Fabric Capacity Events in Real-Time Hub (Preview) – Real-Time Hub now streams Fabric capacity events, enabling proactive monitoring and management of compute resources and workload performance.

Cosmos DB in Microsoft Fabric is now Generally Available – Cosmos DB integrates natively with Microsoft Fabric, enabling seamless NoSQL workloads alongside analytics and AI within a unified data platform.

What’s New for Fabric Data Agents at Ignite 2025 – Fabric Data Agents gain enhanced reasoning capabilities and improved AI interoperability, enabling more sophisticated data analysis and automated insights.

ReadWrite access controls within lakehouse (preview) is now supported for items via OneLake security. This enhancement gives data owners the ability to grant precise write permissions to users—without requiring elevated workspace roles like Admin or Member. With ReadWrite access, workspace viewers or users with only Read access can now write data to specific tables and folders in a lakehouse, while remaining restricted from creating or managing Fabric items. (more info)

Some name change history: Azure AI Studio was rebranded as Azure AI Foundry at Microsoft Ignite 2024, introducing a unified platform for building and managing AI applications. In Ignite 2025, it was again renamed, this time to Microsoft Foundry, along with the announcement of many new features in Foundry (workflow, direct integration with m365).

More info:

Azure at Microsoft Ignite 2025: All the intelligent cloud news explained

Reflections from Microsoft Ignite 2025 – Podcast

Microsoft Databases and Microsoft Fabric: Your unified and AI-powered data estate

Fabric November 2025 Feature Summary