AWS piles on S3 upgrades from vector search to 50 TB objects

AWS has boosted its S3 offerings with support for larger, faster-accessed, and more AI data sources, including NetApp, and has announced a Cohesity European Sovereign Cloud partnership.

The news came at AWS re:Invent 2025 in Las Vegas. AWS says S3 now stores more than 500 trillion objects taking up hundreds of exabytes of capacity. There are six S3 updates with S3 Vectors scaling up and becoming generally available, S3 object size increased, S3 batch jobs boosted, S3 Tables getting tiered, FSx for NetApp ONTAP becoming S3-accessible, and S3 Storage Lens getting new performance measures.

S3 Vectors were announced in July, with AI vector-storing buckets available for retrieval-augmented generation (RAG) and agent use. They are now generally available, scale up to two billion vectors per index (40x preview capacity), support up to 20 trillion vectors per bucket, deliver 2-3x faster frequent-query performance, and reduce costs by up to 90 percent over alternatives. By “alternatives,” AWS appears to be referring to startups such as Pinecone, Weaviate, and Zilliz.

AWS applications such as Bedrock Knowledge Bases and its OpenSearch Service can integrate with S3 Vectors as a vector data source.

AWS is increasing the maximum S3 object size to 50 TB from the current 5 TB to cope with high-res videos, large seismic data files. and AI training datasets.

Some customers need to move S3 datasets across AWS regions or tag objects in their buckets for lifecycle management and do this using batch jobs. Such S3 Batch Operations will now be up to 10x faster at a scale of up to 20 billion objects in a job, AWS says.

AWS introduced an Iceberg table bucket type to S3 a year ago, and now stores more than 400,000 such tables. It’s adding automatic replication across regions and three cost-lowering access tiers by supporting an Intelligent Tiering class.

This class automatically tiers table data across Frequent Access, Infrequent Access, and Archive Instant Access tiers. This is based on access patterns and can deliver up to 80 percent cost savings. AWS says there is no operational overhead or performance impact from this tiering.

The replication addition saves the need for equivalent manual procedures and syncing.

There are also three new Storage Lens features to reveal performance metrics, support for analyzing billions of S3 prefixes, and export to S3 Tables. Learn more about these features here

NetApp

NetApp’s fully managed FSx for ONTAP is becoming an effective S3 data source. By using Amazon S3 Access Points customers can connect S3-based AWS services and ISV applications to FSx for ONTAP file systems as if they were S3 buckets. The AWS services include artificial intelligence, machine learning, serverless compute, and analytics services and others that work with S3. Think Bedrock Knowledge Bases, SageMaker, and Athena.

Pravjit Tiwana, SVP and GM, Cloud Storage and Services at NetApp, said: “By connecting FSx for ONTAP data natively to AWS’s wide range of AI, ML, and analytics services, the new integration with Amazon S3 Access Points unleashes the potential to connect to more than 100 exabytes of enterprise data stored on NetApp systems for transformative use cases like generative AI and analytics use cases built on AWS.”

NetApp says built-in replication capabilities in ONTAP enable customers to move their data within hybrid cloud environments, allowing AWS FSx for ONTAP customers to connect to their ONTAP data stored on-premises.

Cohesity

Cohesity has been selected as a launch partner for the new AWS European Sovereign Cloud, a strategic move from AWS designed to meet Europe’s sovereignty and regulatory requirements. Cohesity says this cloud, expected to launch by the end of the year, is “creating a fully isolated European environment for regulated industries.” The crux of this is whether AWS will give US government functions access to the data in this cloud. If AWS can credibly guarantee restricted access, it will have a much better European market fit than in the absence of such a guarantee.

This cloud will have Cohesity’s data cloud platform capabilities, providing all the functionality of the integrated Helios platform and Data Protect. This includes immutability, multi-role access control, quorum approvals, and air-gapped protection. Find more information here.

There is plenty of S3 background reading in AWS blogs, which can be found here. An overall summary of AWS re:Invent 2025 announcements, with deep dive links, can be found here.