Describes Amazon S3 Vector Buckets for storing and querying vector embeddings natively in S3 to enable serverless similarity search, lower costs, and simplify ML workflows.
Hello and welcome back.In this lesson we’ll explore S3 Vector Buckets — Amazon S3’s native support for storing and querying vector embeddings directly inside S3. This capability lets you keep embeddings and perform similarity search without deploying a separate vector database, simplifying architecture and lowering operational cost.Why this matters
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.
Analogy
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.
Overview: what S3 Vector Buckets provide
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.
Key benefits (at-a-glance)
Benefit
Why it matters
Typical outcome
Cost savings
Avoids separate vector DB storage/compute costs
Up to ~90% cost reductions in many scenarios
Serverless
Fully managed by AWS, auto-scaling
Less ops overhead and simpler deployments
Massive scale
High per-index and per-bucket vector limits
Supports billions–trillions of vectors
Low latency
Optimized for sub-second similarity queries
Fast retrieval for real-time apps
Ecosystem integration
Works with Bedrock, OpenSearch, IAM, PrivateLink
Easier integration into ML and app pipelines
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.
Preview vs GA — highlights and limits
Area
Preview
General Availability (GA)
Launch date
July 14, 2025
December 2, 2025
Per-index limit
Initial limits in preview
Up to 2 billion vectors per index
Per-bucket scale
Early scale tests
Up to 20 trillion vectors per bucket
Query latency
Preview optimizations
Sub-100 ms typical / sub-second at scale
Features added at GA
Basic integrations
CloudFormation, PrivateLink, tagging, 1,000 PUT TPS
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.
Final thoughts
S3 Vector Buckets bring scalable, cost-effective vector search into the S3 ecosystem. They are an attractive option when you want to:
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.
Next steps / hands-on demo
In the next lesson we’ll walk through a hands-on demo: creating an S3 Vector Bucket, ingesting embeddings, and running similarity queries so you can see the end-to-end workflow.References and further reading
