Amazon S3 Vectors Reaches GA, Introducing "Storage-First" Architecture for RAG

2026-01-02 · Source: InfoQ · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Data Science & Analytics · Depth: Intermediate, quick

Summary

Amazon S3 Vectors has reached general availability (GA), introducing a "storage-first" architecture for retrieval-augmented generation (RAG) applications. This AWS cloud object storage service now natively supports storing and querying vector data, increasing per-index capacity forty-fold to 2 billion vectors. It also achieves sub-100ms query latencies for frequent queries and under 1 second for infrequent queries, with up to 100 search results per query. Write performance supports up to 1,000 PUT transactions per second for single-vector updates. The service, which saw over 250,000 vector indexes and 40 billion vectors ingested during its preview, is now available in 14 AWS regions. Key integrations with Amazon Bedrock Knowledge Base and Amazon OpenSearch are also generally available, allowing S3 Vectors to serve as the vector storage layer.

Key takeaway

For CTOs and VPs of Engineering evaluating vector database solutions for RAG, S3 Vectors presents a compelling "storage-first" alternative. You can significantly reduce total ownership costs by up to 90% by eliminating idle compute and cluster management overhead, paying only for storage and query fees. Consider S3 Vectors as a reliable, scalable trunk for your vector data, especially for applications not requiring a "Ferrari" vector database.

Key insights

S3 Vectors offers a "storage-first" approach for RAG, simplifying vector data management and scaling.

Principles

• Consolidate vector datasets into a single index.
• Treat vectors like any other S3 object for serverless management.

Method

S3 Vectors enables vector search by storing vector data, metadata, and keys, with costs based on logical GB uploaded, total logical storage, and per-API/TB query charges.

In practice

• Use S3 Vectors for internal RAG applications.
• Integrate with Amazon Bedrock Knowledge Base.
• Utilize with Amazon OpenSearch for search and analytics.

Topics

• S3 Vectors
• Retrieval-Augmented Generation
• Vector Storage
• Amazon Bedrock
• Serverless Architecture

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Machine Learning Engineer, AI Architect, MLOps Engineer

Related on AIssential

• You probably don’t need a Vector Database (Yet) for your RAG — For small-to-medium RAG applications, NumPy and SciKit-Learn can replace dedicated vector databases.
• Architectural patterns for graph-enhanced RAG: Moving beyond vector search in production — Graph-enhanced RAG improves LLM accuracy by integrating structural data from graph databases with semantic vector search.
• Vectorless RAG - Local Financial RAG Without Vector Database | Tree-Based Indexing with Ollama — Vectorless RAG uses document structure and LLMs for retrieval, eliminating vector databases for structured data.
• Vectorless Agentic RAG Is Quietly Replacing Traditional RAG Architectures — Retrieval Augmented Generation (RAG) is evolving beyond vector databases, driven by the limitations of semantic similarity.
• RyanCodrai / turbovec — TurboQuant provides a data-oblivious vector quantization method for fast, memory-efficient, and accurate approximate nearest neighbor search.
• The Open Source RAG Stack: A Complete Guide to Building Retrieval-Augmented Generation Systems — The open-source RAG stack offers a modular, customizable approach to building context-rich AI systems.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by InfoQ.