AWS launches S3 Files GA
AWS put S3 Files into general availability so you can mount S3 as a file system directly and skip copying to EFS for many workflows (x.com). AWS published performance claims tied to the launch — more than 10 million IOPS, up to 4TB/s throughput — and said the feature can be up to 90% cheaper for certain machine‑learning workloads ( ).
Amazon Web Services has made Amazon Simple Storage Service Files generally available, letting customers mount Simple Storage Service buckets as file systems instead of copying data into a separate file service first. (aws.amazon.com) Object storage stores data as separate objects, while a file system exposes folders, filenames, and in-place edits. Amazon Web Services said Simple Storage Service Files now gives any Amazon Web Services compute resource direct file-style access to data that stays in Amazon Simple Storage Service. (docs.aws.amazon.com) Amazon Web Services announced the launch on April 7, 2026 and said the feature is now generally available in 34 Amazon Web Services Regions. The company said customers can mount it from Amazon Elastic Compute Cloud, Amazon Elastic Container Service, Amazon Elastic Kubernetes Service, and Amazon Web Services Lambda. (aws.amazon.com; aws.amazon.com; docs.aws.amazon.com) The immediate change is architectural: teams that kept one copy of data in Amazon Simple Storage Service and another in Amazon Elastic File System can now point file-based tools at the Simple Storage Service copy. Amazon Web Services said changes made through the file system are synchronized back to the bucket automatically. (aws.amazon.com; docs.aws.amazon.com) That setup targets machine learning, analytics, and agent software that still expects files and folders even when the source data already lives in object storage. Amazon Web Services said Python libraries, machine-learning frameworks, command-line tools, and shell scripts can work with Simple Storage Service data without new application programming interfaces or custom connectors. (aws.amazon.com) Under the hood, Simple Storage Service Files uses two layers. Frequently used metadata and small-file content are pulled onto a low-latency storage layer, while large reads can stream straight from Amazon Simple Storage Service to favor throughput over caching everything. (docs.aws.amazon.com; docs.aws.amazon.com) Amazon Web Services said the service supports up to 250,000 read input-output operations per second and 50,000 write input-output operations per second per file system, with up to terabytes per second of aggregate read throughput and 1 to 5 gibibytes per second of aggregate write throughput depending on Region. The company also said customers can attach multiple file systems to the same bucket, which removes a bucket-level cap on read and write input-output operations per second. (docs.aws.amazon.com) Writes land on the high-performance layer first and are durable immediately, then sync back to the bucket after about 60 seconds by default. Amazon Web Services said that batching can reduce Simple Storage Service PUT requests and avoid creating a new object version for every rapid edit. (docs.aws.amazon.com) Amazon Web Services is also framing the launch as a cost play. Its product page says customers no longer need to duplicate datasets or run synchronization pipelines between file and object systems, and its pricing page says Simple Storage Service pricing remains pay-as-you-go with no minimum charge, though workload-level savings claims depend on how much data stays on the high-performance layer and how often it is accessed. (aws.amazon.com; aws.amazon.com) The pitch is simple: keep the authoritative copy in Amazon Simple Storage Service, present it as files when applications need files, and stop moving the same dataset between storage products just to satisfy older software expectations. (aws.amazon.com; docs.aws.amazon.com)
