AWS exposes S3 as files

For years, Amazon Simple Storage Service was the cheap, durable warehouse in the cloud, but it stored data as whole objects, not as editable files. If an old application expected folders, file locks, and normal save operations, teams usually had to rewrite code or add a separate file system in front of Amazon Simple Storage Service. (docs.aws.amazon.com) Amazon Web Services just changed that with Amazon Simple Storage Service Files, announced on April 7, 2026. The new service lets a bucket in Amazon Simple Storage Service show up to compute as a shared file system instead of only as an object store. (aws.amazon.com) The simplest way to picture the difference is this: object storage is like storing sealed boxes on warehouse shelves, while a file system is like keeping papers in desk drawers. With sealed boxes, you usually replace the whole box; with desk drawers, you open one folder, edit one page, and save it back in place. (aws.amazon.com) Amazon Simple Storage Service Files sits between those two worlds. Amazon says it gives “full file system semantics,” which means applications can do normal file operations like reading, writing, organizing directories, and locking files while the underlying data stays in Amazon Simple Storage Service. (docs.aws.amazon.com) Amazon built the new layer using Amazon Elastic File System, which is the company’s managed network file system service. The result is a shared file system that can connect Amazon Web Services compute directly to existing bucket data without copying that data into a second storage product first. (docs.aws.amazon.com) That “without copying” part is the whole pitch. Amazon says new and existing data in buckets works with no migration required, so a team can create a file system, mount it on compute, and start working against the same data already sitting in Amazon Simple Storage Service. (aws.amazon.com) The service does not turn every byte into a local file all at once. Amazon says it imports directory metadata on access, pulls in small files under a default 128 kilobyte threshold, and serves large reads of 1 megabyte or more directly from the bucket. (docs.aws.amazon.com 1) (docs.aws.amazon.com 2) That is how Amazon gets both speed and scale into the same product. The company says actively used data is cached for low-latency access, while large sequential reads can come straight from Amazon Simple Storage Service, and aggregate throughput can reach multiple terabytes per second. (aws.amazon.com) (docs.aws.amazon.com) Writes are not mirrored back to the bucket instantly. Amazon says the system usually waits about 60 seconds to batch successive file changes before copying them to Amazon Simple Storage Service, and changes made directly in the bucket typically appear in the file system within seconds. (docs.aws.amazon.com) That means this is not pretending object storage and file storage are identical. Amazon’s own troubleshooting guide says the bucket is the “source of truth” during conflicts, and if the file system and the bucket disagree, the conflicting file can be moved to a lost and found directory while the latest bucket version is imported. (docs.aws.amazon.com) The first users Amazon is aiming at are the ones with old software and shared data pipelines. Amazon says every file-based application, agent, and team can use existing tools, and Amazon Elastic Container Service already documents mounting Amazon Simple Storage Service Files volumes into container tasks. (docs.aws.amazon.com 1) (docs.aws.amazon.com 2) The launch is already broad by Amazon standards. Amazon says Amazon Simple Storage Service Files is generally available in 34 Amazon Web Services Regions, which turns a long-running storage tradeoff into a product feature instead of a migration project. (aws.amazon.com)