- Vector embeddings power semantic search, recommendation, and retrieval-augmented generation (RAG). Storing and querying embeddings natively in S3 reduces data movement, simplifies security and lifecycle management, and leverages S3’s serverless scale.
- A regular S3 bucket is like a giant filing cabinet where you store files and retrieve them by filename or key.
- S3 Vector Buckets extend that cabinet so it can “understand” the content and let you search by similarity rather than just name or path.
S3 Vector Buckets provide serverless, scalable vector storage and similarity search directly within S3, reducing operational overhead and often lowering costs compared to dedicated vector databases.
- Native vector storage and similarity search inside S3.
- Serverless operation: no infrastructure to provision or manage.
- Integration with AWS ecosystem (e.g., Bedrock Knowledge Bases) and export paths to analytics/search services like OpenSearch.

Preview, GA and availability timeline
S3 Vector Buckets progressed quickly from preview to general availability and expanded both features and regional reach.
- Preview (July 14, 2025): initial capabilities, integration with Bedrock Knowledge Bases, and export support to OpenSearch.
- GA (December 2, 2025): increased limits and performance, expanded to 14 AWS Regions, added CloudFormation support, PrivateLink, tagging, and higher write throughput.

Early traction and ecosystem adoption
- During preview, customers and partners tested large workloads: hundreds of thousands of vector indexes and tens of billions of vectors.
- By November 2025 some adopters executed over a billion similarity queries against S3 Vector Buckets.
- Ecosystem integrations (for example, Supabase) began appearing quickly, signaling strong interest from the developer community.
- Simplify architecture by co-locating objects and embeddings in the same store.
- Reduce operational overhead with serverless scaling.
- Leverage AWS integrations for production ML and search workflows.