Microsoft announced its Azure HorizonDB for PostgreSQL in preview at the company’s Ignite conference this week.

With this new offering, Microsoft is joining the field of cloud native, managed Postgres services with a platform designed to compete directly with Amazon Aurora PostgreSQL and Google AlloyDB.

The preview release makes Microsoft a serious PostgreSQL vendor, where the open source database has become, as Andrew Brust, CEO of Blue Badge Insights and Microsoft MVP, put it, “the vanilla ice cream of databases: everybody wants some, and everyone takes its availability for granted.”

“We are amplifying Postgres goodness with cloud native architecture, leading on AI, performance and developers,” Shireesh Thota, corporate vice president for Azure databases at Microsoft, told The New Stack in an interview.

Disaggregated Architecture Drives Performance

Azure HorizonDB’s core innovation lies in its disaggregated architecture, which separates compute from storage — a departure from traditional PostgreSQL’s tightly coupled design, Thota said.

“With HorizonDB, we disaggregated. We move the compute outside of the storage so your compute can grow independently. Your storage can grow independently,” he noted. “You don’t have to worry about increasing both of them together and decreasing both together.”

With this architectural approach, the service can scale to 3,072 vCores across primary and replica nodes, with autoscaling shared storage supporting databases up to 128TB, the company said. Microsoft also claims sub-millisecond multizone commit latencies and up to three times more throughput compared with open source Postgres for transactional workloads.

The shared storage architecture also allows unlimited read replicas since data is not coupled to the primary node.

“If somebody has a workload that has a lot of reads coming in, you can keep adding read replicas because the data is moved away from the primary node, it’s shared across,” Thota said.

Vector Indexing for AI Workloads

Microsoft is betting heavily on AI use cases, introducing advanced filtering capabilities for DiskANN vector indexing that enable query predicate pushdowns directly into vector similarity searches.

“We have a lot of advanced industry-leading vector indexing capabilities inside HorizonDB that makes it very easy for you to write semantic search and the next-generation intelligent applications,” Thota said.

The service also includes built-in AI model management that integrates generative, embedding and reranking models from Azure AI Foundry with zero configuration — addressing what Microsoft sees as a gap in how databases support Retrieval-Augmented Generation (RAG) patterns.

Thota provided the example of employee benefits queries: “If an employee wants to go and ask the question about a specific health thing or a benefit question, LLMs [large language models] in ChatGPT or any LLM of your choice wouldn’t be able to access that because the data is locked in a database. But we can partner with LLM and a database if the database has some of the smarts.”

Alpha Life Sciences, an existing Azure Database for PostgreSQL customer, is already planning to build on Azure HorizonDB, Thota said.

“Its seamless support for Vector DB, RAG, and Agentic AI allows us to build intelligent features directly on a reliable Postgres foundation. With Azure HorizonDB, I can focus on advancing AI capabilities instead of managing infrastructure complexities,” said Pengcheng Xu, CTO of Alpha Life Sciences, in a statement.

Developer Tooling as Differentiator

Beyond the database engine itself, Microsoft is enhancing the developer experience through its PostgreSQL Extension for VS Code, which became generally available along with the HorizonDB preview, Thota said.

The extension has already garnered approximately 250,000 downloads since its preview release months ago, he added. It integrates GitHub Copilot with the context awareness of the PostgreSQL database and adds live monitoring with one-click debugging, where Agent mode can launch directly from performance dashboards, he said.

“The tooling is not that great for Postgres, so Microsoft has contributed massively into this by creating a VS Code extension,” Thota said.

The extension also includes a preview of GitHub Copilot-powered Oracle migration capabilities, enabling users to automate end-to-end conversion of complex database code within VS Code’s integrated development environment.

Brad Shimmin, an analyst at The Futurum Group, said he sees this developer ecosystem integration as Microsoft’s key advantage.

“The market is already flooded with managed Postgres options, so basic capabilities aren’t enough to turn many heads,” Shimmin told The New Stack. “Microsoft’s real differentiator here is its developer ecosystem. By integrating this database with tools like Visual Studio and GitHub Copilot, they aren’t just selling a database; they are selling a seamless, AI-assisted platform upon which to build apps.”

Enterprise Features and Security

Azure HorizonDB includes enterprise-grade security features from launch, including native Entra ID authentication, Private Endpoints and data encryption. All data is replicated across availability zones by default, with automated backups and Azure Defender for cloud integration, Thota said.

“A great example is Microsoft has a very strong authentication mechanism in Entra, and so we brought the Entra authentication to Postgres, which wasn’t existing earlier, so you really have a robust security mechanism,” Thota said.

Microsoft’s PostgreSQL Commitment

Meanwhile, Thota said Microsoft is a major contributor to the open source PostgreSQL project, with nearly 20 PostgreSQL project contributors employed by the company.

“We are one of the biggest contributors into the Postgres source code,” he said. “We’ve contributed, in fact, bigger than most of the hyperscale competition in terms of getting there.”

The company contributes engine innovations like Async I/O back to the upstream project.

“We want to make sure that this product really thrives. It’s not just us; anybody who uses Postgres anywhere should get those benefits, and we are committed to doing that,” Thota added.

Microsoft’s team is already working on contributions to PostgreSQL 19, scheduled for release next year.

Competitive Landscape

As mentioned, the new offering puts Microsoft in direct competition with Amazon Aurora PostgreSQL and Google AlloyDB in the cloud native managed PostgreSQL space.

“While Azure had its more pedestrian Azure Database for PostgreSQL and its specialized Cosmos DB for PostgreSQL, it lacked a big, distributed, all-you-can-eat, vanilla Postgres,” Brust said. “Now that gap is filled and equilibrium is established. Aurora PostgreSQL and AlloyDB, move over. This bench seats three.”

Meanwhile, Shimmin noted that Azure HorizonDB for PostgreSQL serves multiple strategic purposes for Microsoft.

“It shores up Microsoft Fabric with a solid operational layer and gives the company a nice, cloud native answer to other relational databases like Google AlloyDB,” he said. “It’s a smart evolution beyond just SQL Server, specifically engineered to handle the increasingly demanding vector indexing needs of today’s AI and agentic AI workloads.”

Azure HorizonDB is initially available in Central US, West US3, UK South, and Australia East regions. Customers can apply for early preview access at aka.ms/PreviewHorizonDB, though participation is limited.

The service complements Microsoft’s existing Azure Database for PostgreSQL flexible server, providing what Thota described as options for customers “who want 100% Postgres” versus those who “want to scale it and make it cloud native.”

TRENDING STORIES