At AWS's recent summit in New York City, the company announced the preview version of Amazon S3 Vectors, positioning it as the first cloud object storage solution designed to handle large-scale vector datasets. This new offering delivers sub-second query performance while significantly reducing storage costs for AI workloads compared to traditional vector databases. S3 Vectors introduces a novel concept called vector buckets - a specialized bucket type with dedicated APIs distinct from standard S3 buckets. As explained by AWS senior developer advocate Channy Yun: "This innovation allows developers to economically store vector embeddings representing unstructured data like images, videos, and audio files, enabling scalable generative AI applications including semantic search, RAG systems, and agent memory construction." Upon creating a vector bucket, developers can organize vector data within vector indexes and execute similarity search queries. Documentation reveals each vector bucket supports up to 10,000 vector indexes, with each index capable of storing tens of millions of vectors. The platform enables metadata attachment through key-value pairs, with all metadata filters enabled by default. Vector indexes support string, numeric, boolean, and list data types. While designed for infrequently queried workloads, S3 Vectors integrates with Bedrock Knowledge Bases and OpenSearch for RAG application development. AWS VP and distinguished engineer Andrew Warfield clarified the positioning: "For S3, this marks an exciting evolution as we observe workload patterns shifting. S3 Vectors offers dramatically lower storage costs than traditional vector stores, though it currently doesn't match the high TPS and low latency of DRAM-based solutions. Our approach reflects how builders value durable, cost-effective foundation layers for vectors - with flexibility to migrate data to higher-performance tiers when needed." To streamline vector embedding management, AWS released the S3 Vectors Embed CLI library - a standalone command-line tool for creating, storing and querying vector embeddings on S3. The provider also published a Python SDK tutorial titled "Getting Started with S3 Vectors." Community reactions show strong interest in this development. On Reddit, developer Travis Cunningham noted: "This innovation effectively transforms every S3 bucket into a mini vector store, maximizing hardware profit margins already secured by AWS while closing competitive avenues for AI workload diversification." Meanwhile, Hacker News user bob1029 raised valid concerns: "Many users still prioritize traditional full-text search over adding another 'black box' layer after LLMs. When you already have a model with strong semantic understanding, why should document storage also become 'smart'? Models could naturally map multiple OR clauses to search terms based on contextual understanding." Separately, S3 now offers real-time inventory tables for bucket metadata, providing complete snapshots of objects and their metadata through managed Apache Iceberg tables. The S3 Vectors preview was accidentally leaked days before the official announcement and is currently available in regions including Northern Virginia, Ohio, and Frankfurt.