S3 Vectors

Amazon S3 Vectors is a native storage and query engine for vector embeddings that reduces infrastructure costs by 90% compared to standalone vector databases.

Amazon S3 Vectors turns standard object storage into a high-performance retrieval engine for AI agents and semantic search. It introduces specialized vector buckets that house billions of embeddings with sub-second query latency. By utilizing a serverless architecture, it removes the overhead of managing dedicated clusters while syncing directly with Amazon Bedrock for RAG (Retrieval-Augmented Generation) pipelines. Teams use it to index massive media archives (including millions of hours of video) at a fraction of the price of in-memory alternatives. The service supports up to 4,096 dimensions and provides dedicated APIs like QueryVectors for fast similarity matching.

https://aws.amazon.com/s3/features/vectors/