Blog Post

Microsoft Blog for PostgreSQL
9 MIN READ

HorizonDB: PostgreSQL for enterprises, developers, and agents on Azure and Fabric

charlesfeddersenMS's avatar
Nov 17, 2026

Today at Microsoft Ignite, we’re thrilled to announce the general availability of Azure HorizonDB and the public preview of Microsoft Fabric HorizonDB, expanding what’s possible with PostgreSQL on Azure. HorizonDB brings a new cloud-native architecture to PostgreSQL, delivering new levels of performance, scalability, and availability for building AI-powered applications or modernizing demanding enterprise workloads.

The role of PostgreSQL is growing, and so is the range of users and companies that depend on it. Enterprises are standardizing on it for critical workloads, developers are choosing it to build new applications, and AI agents are increasingly using it to create, retrieve, and act on data. Each brings different needs, from resilience and security to instant provisioning, developer flexibility, and agent-driven workflows. We built HorizonDB to meet those needs, on Azure and in Microsoft Fabric.

Built for mission-critical enterprise workloads
Azure HorizonDB is built for the most demanding PostgreSQL workloads, where performance and availability matter most. It delivers up to 8x faster OLTP performance than self-managed PostgreSQL* and supports databases up to 128 TB. Every database is resilient across availability zones by default, with sub-millisecond commit latency and failover in under five seconds, providing high availability without impacting application performance.

* Based upon internal, standardized benchmarking, running on high availability and on comparable hardware. HammerDB TPROC-C workload, 95K Warehouses, 512 Users, Azure Edsv6 Virtual Machines with 64 vCores, PostgreSQL 17.11, Korea Central, Availability Zone 1 and 2

HorizonDB integrates with Azure security controls for identity, networking, and data protection. Microsoft Entra ID authentication centralizes identity-based access to the database, helping teams apply consistent identity governance across their Azure environment. Private endpoints provide database connectivity through Azure virtual networks, keeping those connections within the enterprise network boundary. Customer-managed keys give organizations control over the encryption keys that protect their data, in line with their key management policies. Together, these capabilities support sensitive workloads on Azure while preserving the security practices already in place.

“Our AI ambitions depend on secure, reliable, high-performance API access to LSEG’s data. Azure HorizonDB meets our enterprise security requirements while providing a PostgreSQL foundation for our existing API services and the MCP layer powering new AI experiences.” – Grzegorz Giedrojc, Director of Engineering, London Stock Exchange Group

Serverless PostgreSQL in Fabric, from first query to OneLake
Developers have always built applications on databases, but agents are now building them at an entirely different speed. Agents spin up databases for prototypes, experiments, and short-lived tasks, then drop them when the work is done. Most of these workloads start small, run in bursts, and sit idle between uses, which makes their compute needs hard to predict and inefficient to run continually.

Microsoft Fabric HorizonDB is serverless and built for exactly these scenarios, with:

  • Rapid provisioning: Databases provision in seconds so developers and agents can start building immediately.

  • Automatic pause: Compute pauses after five minutes of inactivity to avoid keeping resources running between tasks.

  • On-demand resume: Compute resumes when the next query arrives without manual intervention.

Fabric HorizonDB is also the default database for Fabric apps, where the Rayfin SDK brings PostgreSQL, data APIs, authentication, functions, storage, and observability together in a managed application backend. Instead of assembling and connecting these services separately, developers and agents can focus on application logic and the data it needs, moving from first idea to a running application without operating the backend underneath it.

Once those applications are running, they need more than their own operational data. Fabric HorizonDB now supports translytical capabilities on OneLake to operationalize lake data, bringing the scale and performance of Fabric's analytical engines to live transactions through:

  • Translytical queries: PostgreSQL queries are transparently converted to run on Fabric Data Warehouse, using distributed, scale-out execution on CPU- or GPU-powered processing. Analytical queries can run an order of magnitude faster than they would natively in PostgreSQL. Full and partial pushdowns are supported, so a single query can efficiently join PostgreSQL tables with data in OneLake.

  • Mirroring: Fabric HorizonDB tables are replicated into OneLake by default and are automatically converted to Delta format, so data is ready for translytical application workloads and large-scale analytics across Fabric without building data integration pipelines.

  • Synced Tables: Enriched data from Fabric Data Warehouse flows back into HorizonDB (reverse ETL) for low-latency, high-concurrency serving to applications and agents. Snapshot sync handles periodic refreshes such as daily scores or aggregates, while the continuous sync option keeps tables current as source data changes.

Together, serverless compute and translytical capabilities on OneLake enable developers and agents to build Fabric apps that start small, scale with demand, and act on live transactions and analytical insights in the same application, all through one PostgreSQL API. With no pipelines to build or infrastructure to manage, every application can use the full value of data in OneLake from day one.

“As settlement cycles shorten and financial services become more automated, resolving payment and trade exceptions requires more than the current transaction state. Fabric HorizonDB enables us to connect operational data with historical insights in OneLake to provide firms with the broader business context to resolve exceptions faster and make more informed decisions.” – Jennings Balavari, Founder and CEO of Opsen AI

Build retrieval, reasoning, and AI workflows, all in PostgreSQL

AI agents need more than a place to store data. They need a data platform that combines relevant context and relationship-based reasoning with integrated AI workflows and tools designed for agents. HorizonDB consolidates these capabilities in PostgreSQL so applications can do more with operational data without stitching together specialized databases and disparate services.

Retrieve the right context. HorizonDB combines vector and full-text search to capture both semantic meaning and precise keyword matches. Applications can retrieve relevant context directly in PostgreSQL without separate search or vector databases.

  • pgvector support: Existing vector applications work as they are.

  • DiskANN indexing: Embedding search scales with up to 1.5x higher throughput than HNSW at comparable recall and up to 5x smaller indexes, reducing storage and memory requirements.

  • BM25 ranking: pg_textsearch ranks text results by keyword relevance to find exact names, identifiers, and technical terms.
  • Hybrid search with RRF: Reciprocal Rank Fusion functions combine vector and keyword results based on their positions in each ranked list. This helps retrieve relevant context that either search method alone might miss.

“HorizonDB is changing how we think about the boundary between operational data and intelligence. Instead of creating another datastore for every new capability, we’re exploring how relational state, relationships, semantic retrieval, and AI can coexist around the same trusted operational truth.” - Dylan Addington, Enterprise Architect, Pediatric Associates Family of Companies

Reason across relationships. HorizonDB supports graph modeling and traversal through Apache AGE, with significant new optimizations to this extension for faster queries and more efficient graph updates. The AGE extension API is unchanged, so existing applications can run on HorizonDB and just work without code changes. Compared with community Apache AGE:

  • Simple reads: Resolve direct dependencies, linked entities, and ownership up to 3,900x faster (LDBC SNB Interactive workload).

  • Complex reads: Run multi-hop queries across connected data up to 512x faster (LDBC SNB Interactive workload).

  • Graph builds and refreshes: Run MERGE and batch upserts up to 190x faster with about 6x lower peak memory (internal Microsoft testing).

Faster graph queries reduce the time agents spend following relationships across connected data. A supply chain agent can trace a delayed component through suppliers, shipments, and affected orders. A support agent can connect a customer’s case to related accounts, products, and known issues. By combining graph traversal with vector search and relational queries, agents can reason over how data is connected as well as what it contains.

Integrate models and scale AI workflows. Building scalable AI applications requires models and data processing to work together without a separate orchestration layer. HorizonDB integrates model access and AI execution with operational data through:

  • AI Model Management: Connect applications to the models they need and manage model access from the database.
  • AI Functions: Invoke models from PostgreSQL for tasks such as embedding generation, extraction, and content generation.

  • AI Pipelines: Compose these functions into durable, asynchronous workflows that scale AI processing across data. For example, a pipeline can call external models to enrich new rows as they arrive or keep a knowledge graph current as the underlying data changes.

“Our team built an entire RAG retrieval pipeline in Azure HorizonDB without an application tier. Chunking through reranking all ran in SQL.” - Brian Steele, Chief AI Officer – Field, Coretek

Connect agents to HorizonDB with tools and skills: Microsoft’s PostgreSQL MCP Server and PostgreSQL Skills combine database connectivity with task-specific guidance for agents.

  • PostgreSQL MCP Server: Provides a standard way to discover and invoke HorizonDB capabilities through the Model Context Protocol.
  • PostgreSQL Skills: Supply reusable knowledge and instructions for common database tasks.

Modern developer tools for HorizonDB, powered by VS Code

More than 70% of developers already work in Visual Studio Code, according to the 2026 Stack Overflow Developer Survey. The PostgreSQL extension for VS Code brings HorizonDB development, management, and troubleshooting into the editor, so teams can understand their cluster, tune queries, and inspect data and AI workflows without switching tools.

  • Cluster topology: A clear view of each HorizonDB cluster shows its pools and nodes, including the writer and read-only nodes, their availability zones, compute configuration, high availability status, and parameter group sync state.
  • AI-assisted query optimization: Query plans render as interactive graphs with performance hotspots highlighted. Developers can ask GitHub Copilot to analyze a plan or individual operator and suggest optimizations, using live database schema context.

  • Monitoring and debugging: Metrics, wait events, top queries, and active sessions are available alongside cluster topology, helping developers investigate slow queries and blocking relationships. Queries can be opened directly in the SQL editor or plan visualizer for deeper analysis.

  • Knowledge graph visualization: Apache AGE Cypher results render as interactive node-edge graphs, making relationships in the data easier to explore.

  • AI pipeline visualization: Developers can follow AI workflows from source data through chunking, embedding, extraction, generation, and ranking.

Modernize Oracle and PostgreSQL workloads on HorizonDB

For existing PostgreSQL workloads, the path to HorizonDB is direct. The new Azure Database Migration Service moves PostgreSQL databases from AWS, Google Cloud, on-premises, and Azure virtual machines to HorizonDB, with online migration of schema and data plus built-in validation and monitoring.

Oracle migrations are a different challenge. Beyond moving the data, teams have to convert schemas, PL/SQL packages, and the application code that depends on them, then prove the result behaves the same. That makes an Oracle migration a software engineering project, not a database task. That's why we built the tooling into the PostgreSQL extension for VS Code, where migration teams already have version control, code review, and CI/CD pipelines. Schema and application changes stay in lockstep, with every conversion versioned, reviewed, and regression-tested on each commit.

The extension brings AI-assisted conversion into that workflow with:

  • Schema conversion: Reads the Oracle data dictionary and uses models in Microsoft Foundry to generate PostgreSQL DDL. It then compiles the results against a PostgreSQL database and flags objects that need manual changes in a conversion report.
  • Application conversion: AI-assisted conversion identifies Oracle-specific dependencies in SQL scripts, stored procedure calls, and Java code, then proposes changes for PostgreSQL compatibility. It builds on the output of schema conversion, so the updated code matches the converted object names and types. Developers review and validate each change before accepting it in VS Code, GitHub Copilot CLI, or existing modernization workflows such as Modernize CLI.

  • Data movement: Once the converted schema is deployed, Azure Database Migration Service moves the Oracle data into HorizonDB. Offline migration is now in preview, and online migration is coming soon.

The tools handle the repetitive conversion work, so engineers can focus on review, testing, and the business logic that needs human judgment.

Ready to get started with HorizonDB?

Azure HorizonDB is generally available today in 13 Azure regions across the Americas, Europe, and Asia Pacific, with more regions coming online soon. Microsoft Fabric HorizonDB is available in public preview across Fabric regions worldwide.

You can get started today by creating a new HorizonDB instance using the Azure portal, API’s, or the Visual Studio Code extension for PostgreSQL to begin exploring these capabilities firsthand.

Azure HorizonDB Documentation: Microsoft Azure HorizonDB Documentation

Additional Ignite Content

For a closer look at the capabilities in this post, see our companion blogs on AI with Azure HorizonDB, covering hybrid search, graph, and AI pipelines, and migrating to HorizonDB, covering the new Azure Database Migration Service and Oracle migration in VS Code.

HorizonDB is one part of our PostgreSQL investment on Azure. To see what else is new, read the Azure Database for PostgreSQL Ignite blog, and for the full set of database announcements, read the Azure Databases Ignite blog.

 

Updated Nov 16, 2026
Version 1.0
No CommentsBe the first to comment