SQL Server 2025 takes vectors from proof-of-concept into production with new features designed for real-world AI workloads. In this advanced session, we’ll explore how the new vector index (DiskANN) enables millisecond-level semantic searches over millions of embeddings, making AI-powered search practical inside SQL Server itself. We’ll look at half-precision vector support, which cuts storage requirements and speeds up similarity calculations — a critical improvement when dealing with large-scale embeddings. And we’ll dive into AI_Generate_Chunks, a function that automates splitting content into semantically meaningful passages, a building block for building retrieval-augmented generation (RAG) pipelines directly within the database.

You’ll learn not just how these features work, but why they matter: how to reduce infrastructure costs, accelerate semantic queries, and simplify the pipeline between your SQL data and your AI models. By the end of the session, you’ll see how SQL Server 2025 transforms vector support from “possible” to production-ready, enabling developers and data professionals to deliver faster, more scalable, and more secure AI-driven applications.

This is not an introduction to vectors — it’s a forward-looking session for attendees who already know the basics and want to see how SQL Server 2025’s latest innovations can elevate their AI architectures.