AWS Launches S3 Files to Ease Agentic AI Data Access Bottlenecks

Amazon S3 has long managed data as objects through APIs. While it offers exabyte-scale capacity and durability, it does not fit the file paths and standard read-write operations used by most AI agents and machine-learning tools. AWS is bridging the two interfaces with S3 Files, allowing the same data to combine the scale of object storage with shared-file-system access while reducing the cost of data transfers, duplication and workflow rewrites.

Amazon Web Services (AWS) launched S3 Files on April 7, 2026, allowing existing general-purpose S3 buckets to be mounted on EC2, ECS, EKS and Lambda using NFS 4.1 or later. Built on Amazon EFS, the service delivers latency of about 1 millisecond for active data and supports up to 25,000 connections. File changes are synchronized back to S3 within minutes, and AWS says costs can be reduced by as much as 90%.