Is Aurora storage S3-compatible?

Yes. Aurora Storage is fully S3-compatible, supporting standard S3 API calls, SDKs, and tools. It works as a drop-in replacement for AWS S3 with no code changes required.

What does Aurora S3 storage cost?

Aurora S3 storage is priced at $5.99/TiB/month with no egress fees. A free 1 TiB trial is available with no credit card required.

Does Aurora charge for data egress?

No. Aurora charges zero egress fees on all storage tiers — active and archive. This is a key differentiator from AWS S3 and GCP Cloud Storage.

S3-Compatible Object Storage | Aurora Infrastructure

No egress fees. No retrieval penalties. No API surcharges.

On Aurora, pulling your data to your compute is free, every time. You are billed for what you store, and you can read all of it back as often as your workloads demand at no extra cost.

That matters most for AI work, where the same datasets and checkpoints get read over and over. A training run that pulls 200 TiB of data back to your GPU nodes hundreds of times a month costs nothing to move on Aurora. On a hyperscaler, those reads are metered, and for heavy AI workloads the transfer line can run larger than the storage itself.

The result is simple: your data stays active on Aurora without a meter running every time you touch it.

One platform covers object, block, and file, with lifecycle and immutability built in.

Object storage (S3-compatible)

Buckets, access keys, bucket policies, and ACLs through the S3 API. Drop-in for most S3 workflows with no code changes.

Block volumes (iSCSI)

Attach raw block devices to your instances for databases and latency-sensitive workloads.

Filesystem mounts (NFS/SMB)

Shared POSIX-style file access for applications that expect a mounted path.

Point-in-time snapshots.

Capture volume state for fast rollback and recovery.

Hot and cold in one namespace.

Active (hot) and archive (cold) tiers live under a single namespace, with automatic lifecycle policies that age objects down to cold on the schedule you set.

Immutable WORM backup

Write-once, read-many storage with configurable retention and legal hold, so a backup written today cannot be altered or deleted until its retention clears.

What is S3-compatible object storage? S3-compatible object storage exposes the same API surface as Amazon S3. Because Aurora implements full AWS S3 API compatibility, tools built for S3 (the AWS CLI, boto3 and other SDKs, s3cmd, rclone, Terraform providers, and most backup and data tools that target S3) talk to Aurora by changing the endpoint and credentials. For most S3 workflows there are no code changes. You point at Aurora, keep your buckets-and-keys mental model, and drop the egress bill.

What is WORM storage? WORM stands for write-once, read-many. Once an object is written to a WORM-protected bucket, it cannot be modified or deleted until its retention period expires. Aurora supports configurable retention windows and legal hold, which pins objects in place regardless of the retention clock until the hold is lifted. This is the control ransomware and accidental-deletion recovery plans depend on, and it is the control auditors ask about for records retention.

Specifications

API	Full AWS S3 API compatibility; drop-in for most S3 workflows, no code changes
Protocols	S3 (object), iSCSI (block), NFS/SMB (file)
Tiers	Active (hot) and archive (cold) in a single namespace
Lifecycle	Automatic hot-to-archive policies, configurable
Durability	11 nines (99.999999999%)
Throughput	40+ GB/s
Scale	TiB to PiB, no repricing, no architectural ceiling
Snapshots	Point-in-time volume snapshots
Immutability	WORM, configurable retention, legal hold
Access control	Access keys, bucket policies, ACLs, IAM integration
Encryption	In-transit and at-rest; bring-your-own-KMS (BYO-KMS)
Price	$5.99 per TiB per month
Free tier	1 TiB free for one month, no credit card
Egress / API / retrieval	No egress fees, no API surcharges, no retrieval penalties

Security & Compliance

Encryption in-transit and at-rest.
Bring-your-own-KMS (BYO-KMS) for customer-managed keys.
WORM immutability with configurable retention and legal hold.
Access keys, bucket policies, ACLs, and IAM integration for fine-grained access control.
Runs on Aurora's private, sovereign infrastructure. [CONFIRM specific compliance attestations before listing any, e.g. SOC 2, ISO 27001, HIPAA.]

What is Aurora Storage used for?

Feeding AI training and inference

Datasets, checkpoints, and model weights sit in object storage and stream to GPU nodes at 40+ GB/s. Because reads are free, re-running training and evaluation over the same corpus adds no cost each time you touch the data.

Backup, Archive, & DR

WORM buckets with configurable retention and legal hold hold immutable backups. Point-in-time snapshots give fast rollback. Cold data ages into the archive tier under lifecycle policy without a separate migration.

Migrating Off a Hyperscaler

Because Aurora offers full AWS S3 API compatibility, moving is an endpoint-and-credentials change for most S3 workflows rather than a rewrite. Teams point their existing S3 tooling at Aurora, copy the data across, and stop paying egress to read it back.

The cluster-integrated performance tier... For GPU-cluster feeds that need more than the self-serve object product delivers, Aurora also offers a high-performance managed storage tier that integrates directly with your cluster. This tier is built on established parallel and high-throughput storage engines such as Weka, DDN, GPFS, and VAST. [CONFIRM which of these are live and offered.] It is distinct from the self-serve S3 object storage product above and is provisioned with your compute. To scope it, Talk to Sales.

S3-Compatible storage built for AI workloads.