Amazon just dropped S3 Vectors - and this could actually shake up the entire AI dev space
Yeah, sounds like another AWS release, but there's something real here. Remember how S3 in 2006 killed the entire file storage industry? Just gave you an API instead of hardware. Same thing might happen to vector databases now.
What caught my attention:
Money talks
They're promising 90% cost reduction. $0.06 per GB storage, $0.004 per TB queries. Some hosted vector DBs cost 400x more - that's just insane. Plus AWS is often free for startups, making AI experiments accessible to everyone.
Scaling without the headache
Billions of vectors, sub-second queries, zero infrastructure. Create a bucket, get an API - you're done. Up to 10k indexes per bucket, each handling tens of millions of vectors.
Everything in one ecosystem
S3 is already everywhere. Now you get native integration with Bedrock for RAG, OpenSearch for hybrid storage, SageMaker. No need to glue different services together.
Where to use this:
- RAG applications just got way cheaper (we have a project with a massive vector DB - thinking of migrating)
- Semantic search across any content at scale
- AI agents with long-term memory through vector-stored history
- Search in medical images, legal docs, code
- Video archive indexing for scene search
- Corporate search by meaning, not keywords
Basically everything you'd use vector DBs for, but cheaper and with proper scaling. For many, it's also more convenient - documents are already in S3 anyway.
Community reaction is mostly positive so far. Query times around 250ms in tests - good enough for most use cases.
Available in preview in US East/West, Europe (Frankfurt) and Asia Pacific (Sydney). If you're testing it - drop your results in the comments.