AWS has introduced a new capability called S3 Files that enables data stored in Amazon S3 to be accessed using standard file system operations, addressing a long-standing limitation of object storage.
For nearly two decades, S3 has served as a core repository for cloud data, particularly for analytics, archives and data lakes. Yet its object-based architecture has required specialized APIs, forcing developers to build workarounds when using applications that expect traditional file systems. The result has often been duplicated data, added infrastructure and complex synchronization pipelines.
S3 Files is intended to remove that friction. It allows applications to read and write data in S3 using familiar file system semantics, without requiring code changes. This means that workloads ranging from machine learning pipelines to enterprise applications can operate directly on data stored in S3, rather than copying it into a separate file-based environment.
Unifying File and Object Storage
The service is built on Amazon’s EFS infrastructure but maintains S3 as the underlying source of truth. Rather than converting objects into files outright, S3 Files presents a file system view of the data. Changes made through the file interface are aggregated and committed back to S3, preserving the integrity and behavior that existing applications depend on.
This architectural decision addresses a key challenge in unifying file and object storage. Files are typically mutable and accessed incrementally, while objects are stored and retrieved as complete units. AWS engineers opted not to erase that distinction but to manage it explicitly, allowing each model to operate within its strengths.
Early user feedback indicates that the system handles concurrency and synchronization reliably, even under conflicting operations between file and object interfaces. Updates to existing data propagate quickly, while newly created objects may take longer to appear in the file system view due to event propagation delays. These trade-offs highlight the complexity of bridging two fundamentally different storage paradigms.
For AI and data-intensive workloads, the implications are major. Machine learning teams can run training jobs directly against S3 data without staging it elsewhere. AI agents and data pipelines can interact with shared datasets using standard tools, reducing latency and operational overhead.
AWS also notes that thousands of compute instances can simultaneously access the same data, supporting distributed workloads without duplication.
Pricing Based on Usage
The pricing model is designed to align with AWS’s existing storage services. File system rates primarily apply to data that is actively accessed through the file interface, while the majority of stored data continues to be priced under standard S3 rates. This approach reflects typical usage patterns, where only a small portion of large datasets is frequently in use.
AWS’s move here is in keeping with an industry trend toward attempting to streamline infrastructure for AI-driven applications. While other vendors offer methods to connect object storage to file systems, AWS promotes S3 Files as a more integrated approach, embedded within its broader storage ecosystem.
S3 Files is now available across dozens of AWS regions, creating a broad rollout of a feature that demonstrates a shift in how cloud storage is evolving.
