Amazon boosts S3 Vectors with bigger scale & faster queries
Amazon Web Services has announced the general availability of Amazon S3 Vectors, introducing expanded scale and improved performance for customers seeking to store and query vector data directly in cloud object storage.
Expanded scale
S3 Vectors now allows users to store and search across up to 2 billion vectors in a single index, raising the total possible capacity to 20 trillion vectors within a single vector bucket. This marks a fortyfold increase from the 50 million vectors per index that was supported during the preview phase. Organisations can now consolidate large-scale vector datasets into one index without the need for sharding or complex query federation, simplifying data management workflows.
Performance improvements
Amazon reports optimised query performance in the fully released version. Infrequent queries continue to produce results in under one second. Users running repeated or high-frequency queries will see latencies of around 100 milliseconds or less, supporting a range of interactive applications, including conversational AI and multi-agent workflows. The number of search results that can be retrieved per query rises to 100, an increase from 30 previously.
The system also offers enhanced write throughput, now supporting up to 1,000 PUT transactions per second for streaming single-vector updates into indexes. This improvement enables rapid ingestion of small datasets or concurrent writing from multiple data sources, ensuring new data becomes searchable almost immediately.
Cost model
S3 Vectors is available through a fully serverless architecture. There is no requirement to set up or provision infrastructure, and customers are billed for their storage and query usage. Amazon states that customers may be able to reduce the total cost of storing and querying vectors by up to 90% compared to specialised vector database solutions.
AI workloads
S3 Vectors is positioned as the underlying storage layer for a variety of artificial intelligence initiatives, including AI agent development, inference, semantic search, and retrieval augmented generation (RAG) applications. The platform is intended to support workflows from initial proof-of-concept projects to large-scale production deployments, enabling quick access to large quantities of vector data needed for developing and scaling AI models and services.
Integrations and regions
Amazon S3 Vectors is now integrated with both Amazon Bedrock Knowledge Base and Amazon OpenSearch. These integrations are generally available after a preview period. Users can employ S3 Vectors as a backing store for vector data, while leveraging OpenSearch for enhanced search and analytics functionality. The Bedrock Knowledge Base integration aims to facilitate the building of large-scale RAG applications using production-ready storage and querying capability.
The service's availability footprint has also expanded, moving from five AWS Regions during preview to fourteen AWS Regions for general availability. This gives more organisations global access to the platform's capabilities.
"You can now store and search across up to 2 billion vectors in a single index, that's up to 20 trillion vectors in a vector bucket and a 40x increase from 50 million per index during preview. This means that you can consolidate your entire vector dataset into one index, removing the need to shard across multiple smaller indexes or implement complex query federation logic," said Amazon Web Services.
