What’s New in SQL Server 2025? AI Vectors, Change Event Streaming, and More

AI Enhancements: Vector Data Type

SQL Server 2025 adds a new Vector data type designed to store high‑dimensional numeric arrays for AI applications. Vectors are saved in an optimized binary format that can be accessed as JSON arrays, simplifying developer interaction.

Storage format: Optimized binary, JSON‑compatible access.

Dimension limit: Up to 1,998 dimensions per vector.

Several built‑in vector functions are provided, including VECTOR_DISTANCE, VECTOR_NORM, and VECTOR_NORMALIZE, enabling distance calculations, norm computation, and normalization. A vector index feature accelerates similarity searches.

CREATE TABLE dbo.vectors (
    id INT PRIMARY KEY,
    v VECTOR(3) NOT NULL
);

INSERT INTO dbo.vectors (id, v) VALUES 
    (1, '[0.1, 2, 30]'),
    (2, '[-100.2, 0.123, 9.876]');

Change Event Streaming (CES)

The new Change Event Streaming (CES) expands on traditional CDC by emitting change events to cloud services such as Azure Event Hubs and Kafka. This enables real‑time propagation of DML operations to downstream systems.

SQL AI Copilot in SSMS

SQL Server 2025 integrates an AI assistant—SQL Copilot—directly into SSMS 21. Developers can write queries using natural English, and the assistant also offers query‑performance recommendations. To enable it, install SSMS 21 and configure the Azure OpenAI endpoint and deployment information.

GraphQL Support via Data API Builder

SQL Server 2025’s Data API Builder can automatically generate a GraphQL schema, allowing seamless front‑end data access without manual schema definition.

Runtime Parameter Optimization (OPPO)

The new OPPO feature continuously monitors runtime parameters, collects performance data, and suggests adjustments to improve query execution without requiring manual tuning.

Azure Cloud Integration and Fabric

SQL Server 2025 is positioned as a data‑analysis platform, supporting data extraction, transformation, and loading (ETL) within Microsoft Fabric. It also offers Azure Arc integration for hybrid deployments, enabling unified management of on‑premises and cloud databases.

Conclusion

Despite being a legacy product, SQL Server continues to evolve with AI‑centric features, cloud‑native capabilities, and developer‑friendly tools, making it a competitive choice for modern data workloads.