Industry story
AWS RDS adds SQL Server 2025 with native vector type and Bedrock integration
rag tool-use vector-embeddings
SQL Server 2025 on Amazon RDS is a quiet but meaningful unlock for teams already running SQL Server workloads who want to add AI without rebuilding around a new stack. The native vector type means embeddings live in the database itself, with no separate vector store bolted on. The REST endpoint support inside T-SQL lets you call Bedrock or Lambda directly from a query. For existing RDS customers, the path is a version upgrade. The tension to watch: whether keeping AI logic inside the database is a genuine architectural advantage or just technical debt with a better marketing story.
Full analysis
Amazon RDS for SQL Server now supports Microsoft SQL Server 2025 (Enterprise, Standard, and Developer editions), bringing two features directly relevant to agent builders. First, SQL Server 2025 lets developers call external REST endpoints from within T-SQL queries (T-SQL is the structured query language for SQL databases) without extra middleware, enabling direct integration with AWS services including Amazon Bedrock (AWS's managed AI/foundation-model service), SageMaker, S3, and Lambda. This opens paths to AI-powered query advisors and event-driven workflows without re-architecting existing database applications.
Second, SQL Server 2025 introduces a native vector data type. Vector embeddings are the numerical representations used in semantic search and retrieval-augmented generation (RAG) systems, and they can now be stored and queried directly inside the database engine. AWS also notes licensing improvements: a new free Standard Developer Edition and expanded Standard Edition capacity (up to 32 cores, 256 GB memory). Existing RDS customers can upgrade by changing the DB engine version; on-premises SQL Server users can migrate to take advantage of the managed infrastructure.
Comments