Amazon Simple Storage Service (S3) is AWS’s object storage service. AWS launched it in March 2006, and it remains one of the most widely used services on the platform. You store files as objects in buckets, pay for what you store and what you do with it, and reach the data over HTTPS. This guide covers the core concepts, storage classes, default security, data protection and costs.
What Amazon S3 is
S3 holds any kind of data, from text files to video and database backups. Teams use it for application assets, backups, log archives, static websites, data lakes and machine learning datasets. You don’t provision servers or disks for it. You create buckets, upload objects and set access policies.
Buckets, objects and keys
- A bucket is a container for objects. Each bucket is created in one AWS Region, and its data stays in that Region unless you copy it.
- An object is the data plus its metadata. A single object can be up to 50 TB.
- A key is the object’s name inside the bucket, such as
reports/2026/q3.pdf.
Bucket names are globally unique. They must be 3 to 63 characters long and can contain lowercase letters, numbers, dots and hyphens. While a bucket exists, no other AWS account can use its name.
S3 has a flat namespace. Keys that contain slashes look like folders in the console, but the folders are only prefixes. Renaming a “folder” means copying and deleting every object under that prefix.
For a single upload, the PUT operation accepts up to 5 GB. For larger objects, use multipart upload. AWS recommends multipart upload for objects over 100 MB.
Durability and availability
S3 Standard is designed for 99.999999999% durability (11 nines). It stores data across at least three Availability Zones in a Region. AWS designs S3 Standard for 99.99% availability and offers a 99.9% service level agreement. These are design and contract targets, not promises against every failure, so backups and replication still matter.
S3 provides strong read-after-write consistency. After a successful write or delete, reads return the latest version. This has been the case since December 2020.
S3 storage classes
Choose a storage class based on how often you read the data and how quickly you need it back.
| Storage class | Typical use | Retrieval |
|---|---|---|
| S3 Standard | Frequently accessed data | Milliseconds |
| S3 Intelligent-Tiering | Data with changing or unknown access patterns | Milliseconds in the default tiers |
| S3 Standard-IA | Infrequently accessed data that still needs fast access | Milliseconds, with a retrieval fee |
| S3 One Zone-IA | Infrequently accessed data that can be re-created | Milliseconds, stored in one AZ |
| S3 Glacier Instant Retrieval | Rarely accessed data that needs millisecond access | Milliseconds |
| S3 Glacier Flexible Retrieval | Archives where waiting is acceptable | Minutes to 12 hours |
| S3 Glacier Deep Archive | Long-term retention of data that is rarely read | Within 12 hours (standard) or 48 hours (bulk) |
| S3 Express One Zone | Latency-sensitive workloads in one AZ, using directory buckets | Single-digit milliseconds |
Some classes have minimum storage durations. Deleting an object early still bills for the rest of the minimum. Standard-IA and One Zone-IA have a 30-day minimum, Glacier Instant Retrieval and Glacier Flexible Retrieval have 90 days, and Glacier Deep Archive has 180 days. Standard-IA and One Zone-IA also bill smaller objects as if they were 128 KB.
Security defaults and access control
S3 security defaults have improved over the years, and current defaults are more protective than many older guides describe:
- Encryption at rest. S3 encrypts new objects by default with server-side encryption using S3-managed keys (SSE-S3, AES-256). This has been the default since January 2023. You can also use AWS KMS keys for more control and audit.
- Block Public Access. New buckets have Block Public Access turned on, and ACLs are disabled by default. Both have been the default for new buckets since April 2023.
- Access policies. Control access with IAM policies, bucket policies and access points. Grant only the access each user or application needs.
- Presigned URLs. Share a single object for a limited time. With AWS Signature Version 4, the maximum lifetime is 7 days.
Default encryption does not make a public bucket private. Encryption protects data on disk, but S3 decrypts objects transparently for any principal with permission to read them. For the wider picture, read Is cloud storage secure?
Data protection
- Versioning keeps earlier versions of each object, so an overwrite or delete does not erase the previous data.
- Object Lock applies write-once-read-many (WORM) protection for a retention period. It helps defend against ransomware and accidental deletion.
- Replication copies objects to another bucket, including one in another Region. Replication applies to new objects by default. Existing objects need a separate step.
- Lifecycle rules move objects to cheaper storage classes or expire them on a schedule.
How much does 1 TB of S3 cost?
S3 bills for several things, and storage is often not the largest part of the bill:
- Storage, per GB per month, which varies by class and Region. For S3 Standard in US East (N. Virginia), the list price has been about $0.023 per GB per month for the first 50 TB. At that rate, 1 TB is roughly $23 a month before other charges.
- Requests, such as PUT, GET and LIST calls.
- Retrieval fees for the infrequent and archive classes.
- Data transfer out to the internet or to other Regions.
Archive classes such as Glacier Deep Archive cost much less per GB to store, but retrieval is slower and billed separately. Check the S3 pricing page for current rates in your Region.
Common uses
- Backups and database exports
- Log storage, including AWS CloudTrail logs
- Static website hosting. S3 serves static sites over HTTP, and HTTPS is usually added through Amazon CloudFront.
- Data lakes and analytics
- Software build artifacts and deployment files
Common mistakes
- Treating S3 as a file system. For shared file access use Amazon EFS or FSx. For a server’s disk, use EBS.
- Forgetting retrieval and minimum-duration costs. Archive storage is inexpensive to keep, but reading it back costs more and takes longer.
- Assuming folders are real. Renaming a prefix copies every object under it.
- Leaving versioning off on buckets with important data, so overwrites and deletes are permanent.
Related AWS storage services
S3 is one part of AWS storage. It handles objects, such as backups, media and data lakes. For a server’s operating system and database disk, the usual choice is Amazon EBS. Shared Linux file systems fit EFS, and Windows or specialized file systems fit FSx. For a map of all of them, see What are different AWS storage services? To see how S3 fits into the broader landscape, start with types of cloud storage services, and see Is S3 a protocol or API? for how applications connect to it.