The growing adoption of AI technologies is presenting new challenges to our customers’ data assets and applications. Most organizations expect to deploy AI workloads in hybrid environments that mix cloud, edge, and dedicated infrastructure, and privacy and security are more important than ever.
Microsoft SQL Server 2025, Previewing nowis an enterprise AI-enabled database from ground to cloud that solves these problems by applying AI to our customers’ data. This release continues SQL Server’s 30 years of performance and security innovation and adds new AI capabilities. Microsoft Fabric integration allows customers to bring their data into the next generation of data analytics. This release leverages Microsoft Azure innovations for customer databases to support hybrid environments across cloud, on-premises data centers, and the edge.
Over the years, SQL Server has gone far beyond traditional relational databases. The latest release of SQL Server enables customers to build AI applications that are tightly integrated with the SQL engine. SQL Server 2025 is transforming itself into a vector database with built-in filtering capabilities along with vector search, delivering superior performance and ease of use for developers using T-SQL.
Built-in AI
This new version has AI built in, leveraging familiar T-SQL syntax to simplify AI application development and Search Augmented Generation (RAG) patterns with secure, high-performance, and easy-to-use vector support. This new feature allows you to combine vectors with SQL data for hybrid AI vector search.
Building AI applications with enterprise databases
SQL Server 2025 is an enterprise-ready vector database with built-in security and compliance features that brings enterprise AI to your data. It features a native vector store and index powered by DiskANN, a vector search technology that uses disk storage to efficiently find similar data points in large data sets. These databases efficiently support chunking and enable accurate data retrieval through semantic search. This latest version of SQL Server takes AI models from the ground to the cloud with flexible AI model management within the engine using the Representational State Transfer (REST) interface.
Additionally, whether customers are working on data preprocessing, model training, or RAG patterns, our scalable low-code tools provide a flexible model interface within the SQL engine, supported by T-SQL and external REST endpoints. These tools enhance developers’ ability to create rich AI applications through seamless integration with popular AI frameworks such as LangChain, Semantic Kernel, and Entity Framework Core.
Increased developer productivity
When building data-intensive applications, such as AI applications, it is important to focus on scalability, frameworks, and data enrichment to increase developer productivity. We ensure that SQL provides the best experience for developers by incorporating features such as REST API support, GraphQL integration through Data API Builder, and regular expression activation. Additionally, native JSON support allows developers to more effectively handle frequently changing schemas and hierarchical data, making it easier to create more dynamic applications. Overall, we are making SQL development more scalable, performant, and user-friendly. All features are backed by the security provided by the SQL Server engine, making it a truly enterprise-grade platform for AI.
Best-in-class security and performance
SQL Server 2025 is the industry leader in database security and performance. Support for Microsoft Entra managed identities improves credential management, reduces potential vulnerabilities, and provides compliance and auditing capabilities. SQL Server 2025 introduces outbound authentication support for Managed Service Identity (MSI) for SQL Server supported by Azure Arc.
It also brings performance and availability improvements to SQL Server that have been extensively battle-tested in Microsoft Azure SQL. The new version delivers improved query optimization and query performance execution to increase workload performance and reduce troubleshooting. Optional Parameter Plan Optimization (OPPO) is designed to help SQL Server select the optimal execution plan based on customer-provided runtime parameter values and significantly reduce invalid parameter sniffing issues that may exist in your workload. Persistent statistics on the secondary replica prevents loss of statistics during restarts or failovers, preventing potential performance degradation. Improvements to batch mode processing and columnstore indexing for query execution further establish SQL Server as a mission-critical database for analytics workloads.
Optimized locking reduces lock memory consumption and minimizes blocking for concurrent transactions through Transaction ID (TID) locking and Lock After Qualification (LAQ). This feature allows customers to increase uptime and improve concurrency and scalability for their SQL Server applications.
Change Event Streaming for SQL Server provides an event-driven architecture, separation of command query responsibilities, and real-time intelligence and real-time application integration. It adds new database engine capabilities that capture incremental changes to data and schema in near real time and publish them to provided targets such as Azure Event Hubs and Kafka.
Microsoft Fabric and Azure Arc connected
In traditional data warehouse and data lake scenarios, any data integration involves designing, monitoring, and managing complex extract, transform, and load (ETL) processes to transfer operational data from SQL Server. These existing methods do not support real-time data transfer, resulting in latency that hinders the generation of real-time analytics. Microsoft Fabric provides comprehensive, integrated, AI-enhanced data analytics services designed to meet the modern needs of analytics workloads. Fabric’s mirrored SQL Server database is a fully managed, elastic process that simplifies replication of SQL Server data to Microsoft OneLake in near real-time. Mirroring allows customers to continuously replicate data from Azure virtual machines or SQL Server databases running outside of Azure to deliver online transaction processing (OLTP) or operational storage workloads directly to OneLake for analytics and insights on a unified fabric data platform. can promote.
Azure continues to be an important component of SQL Server. With Azure Arc, SQL Server 2025 will continue to deliver cloud capabilities to help customers better manage, secure, and govern their SQL assets at scale across on-premises and the cloud. Capabilities such as automatic patching, automatic backups, monitoring, and best practice evaluation provide customers with multiple ways to streamline day-to-day operations and further enhance business continuity. Azure Arc also offers pay-as-you-go options to simplify SQL Server licensing and provide customers with flexibility and license visibility.
Register for preview now
We are currently onboarding customers and partners to the preview of SQL Server 2025 ahead of general availability next year.
Register now for SQL Server 2025 Community Technology Preview (CTP).1 Stay up to date on SQL Server 2025 updates.
Microsoft just announced the release of SQL Server Management Studio (SSMS) 21 Preview 1. This release integrates Microsoft Copilot functionality into SSMS. The Copilot environment simplifies SQL development by providing real-time suggestions, code completion, and best practice recommendations. If you’d like to get involved and get some early hands-on experience with this new feature, please use this link to indicate your interest.
1Some of the new features covered in this blog may not be available in the first CTP version.
Asad Khan is Vice President of SQL Products and Services. He leads all engineering and product management for SQL Server and all of Azure SQL.
More articles by this author
