Object Storage

Last reviewed: August 2026

Overview

To store files in your own data center, you must purchase NAS or SAN and add disks as capacity runs short. Poor capacity planning leads to either running out of space or overinvesting.

Object storage is cloud storage that lets you store files without a capacity limit. You don’t need to decide capacity in advance — you pay only for what you store. It’s used to store nearly all types of unstructured data, including images, videos, backups, logs, and data lakes.

Why Object Storage

Very inexpensive — At around $0.02–0.03 per GB per month, it’s far cheaper than block storage ($0.08–0.10 per GB) or file storage. Using an archive class can bring this down to below $0.001 per GB, though archive classes incur separate retrieval costs and wait times.
Unlimited capacity — There’s no upper bound on storage capacity. From 1KB to several PB, storage works the same way.
Durability — Every vendor offers 99.999999999% (11 nines) durability. This means, if you store 10 million objects, you would expect to lose one roughly every 10,000 years. Data is automatically replicated across multiple AZs, making data loss practically nonexistent.
HTTP API access — Instead of a file system (folders/paths), it uses a key-value structure, accessible from anywhere via REST API.
S3-compatible API — AWS S3 has become the de facto standard API, and most vendors and tools support an S3-compatible API.

Product Comparison

Vendor	Product	Notes
AWS	S3 (Simple Storage Service)	The de facto industry standard. A wide range of storage classes (Standard, IA, Glacier, etc.)
Azure	Blob Storage	Hot/Cool/Cold/Archive tiers. Integrated with Data Lake Storage Gen2
Google Cloud	Cloud Storage	Standard/Nearline/Coldline/Archive. Automatic replication with Multi-region/Dual-region
OCI	OCI Object Storage	Standard/Infrequent Access/Archive tiers. Supports an S3-compatible API

Storage Classes (Cost Optimization by Access Frequency)

Access frequency for data tends to decrease over time. Every vendor offers storage classes that lower storage costs based on access frequency. Storage cost decreases but retrieval cost increases, so choose a class that matches your access pattern.

Access frequency	AWS	Azure	Google Cloud	OCI
Frequent access	S3 Standard	Hot	Standard	Standard
Occasional access	S3 Standard-IA	Cool	Nearline (30 days)	Infrequent Access
Rare access	S3 Glacier Instant	Cold	Coldline (90 days)	—
Archive	S3 Glacier Deep Archive	Archive	Archive (365 days)	Archive
Automatic tiering	S3 Intelligent-Tiering	—	Autoclass	Auto-Tiering

AWS S3 Intelligent-Tiering, Google Cloud Autoclass, and OCI Auto-Tiering automatically analyze access patterns and move data to the optimal class, reducing operational burden since manual lifecycle policies aren’t required.

Key Differences

AWS S3 — Launched in 2006 as one of the first object storage services in the public cloud, the S3 API has been widely adopted as an industry-compatible standard. Most third-party tools, other cloud vendors, and on-premises storage support an S3-compatible API. AWS is evolving storage itself to embed analytics and AI capabilities, with services like S3 Tables, S3 Metadata, S3 Vectors, and S3 Annotations. S3 Annotations lets you attach up to 1GB of queryable context (metadata) directly to an object, allowing AI agents to discover, understand, and process data without a separate metadata system.

Azure Blob Storage — Blob Storage and Data Lake Storage Gen2 are unified within the same storage account. Hierarchical namespace (folder structure) support makes file management convenient for big data workloads. Integration with Microsoft Fabric simplifies building analytics pipelines.

Google Cloud Cloud Storage — Multi-region and Dual-region options automatically replicate data across multiple regions without separate replication configuration. Autoclass supports automatic storage class transitions, and BigLake enables direct querying from BigQuery.

OCI Object Storage — Supports an S3-compatible API, with Auto-Tiering automatically switching between Standard and Infrequent Access based on access patterns. A free egress policy of 10TB/month provides a significant cost advantage for large data transfers.

When to Choose What

When	Choose this
You want to make full use of the S3-compatible API ecosystem	AWS S3
You need big data + hierarchical namespace (folder structure)	Azure Data Lake Storage Gen2
You want automatic multi-region replication with no extra setup	Google Cloud Cloud Storage (Multi-region)
You want automatic storage class transitions	AWS S3 Intelligent-Tiering, Google Cloud Autoclass, OCI Auto-Tiering
You want to reduce large-scale egress costs	OCI Object Storage (10TB/month free)
You want to run SQL analytics directly on object storage	AWS Athena + S3, or Google Cloud BigQuery External Tables

Usage Patterns

Use case	Description	Why object storage
Static web hosting	Serving SPAs and static sites	CDN integration, serverless, unlimited scaling
Data lake	Repository for raw data	No schema required, low-cost at large scale, supports diverse formats
Log/event archive	Storing audit logs and event streams	Append-only, automatic archiving via lifecycle policy
ML training data	Images, text, feature stores	Large-scale unstructured data, parallel reads
Backup/DR	DB snapshots, system images	99.999999999% durability, cross-region replication
Media storage/streaming	Original video and image files	Large scale, CDN origin, input for transcoding pipelines

The Evolution of Object Storage

Object storage has grown beyond simple file storage to become the foundational storage layer for data lakes. In the past, data had to be copied into a separate data warehouse for analysis, but now the lakehouse architecture — analyzing data directly where it sits in object storage — has become standard.

The Lakehouse Pattern

Raw data stays in object storage and is structured using table formats (Iceberg, Delta Lake, Hudi), enabling direct SQL queries without a separate data warehouse.

Medallion Architecture (Bronze/Silver/Gold):

Bronze — Raw data as-is (a mix of JSON, CSV, Parquet). Automatically archived via lifecycle policy
Silver — Cleansed/transformed data (Parquet, schema applied)
Gold — Business-level aggregates/marts (ready for immediate analysis)

Event-Driven Pipelines

Object uploads trigger events that automatically run transformation/analysis.

Vendor	Trigger	Processing
AWS	S3 Event Notification	Lambda, Step Functions, EventBridge
Azure	Blob Trigger	Functions, Data Factory
Google Cloud	Cloud Storage Trigger	Cloud Functions, Dataflow
OCI	OCI Events	OCI Functions, Data Flow

Vendor Data Platform Services

Feature	AWS	Azure	Google Cloud	OCI
Data lake	S3 + Lake Formation	Data Lake Storage Gen2 + Fabric	Cloud Storage + BigLake	OCI Object Storage + Data Lake
Table storage (Iceberg)	S3 Tables	Data Lake Storage + Synapse	BigLake (native Iceberg)	—
Automatic metadata management	S3 Metadata	Blob Index Tags	—	—
Vector storage/search	S3 Vectors (Preview)	AI Search + Blob integration	BigQuery Vector Index	AI Vector Search (Autonomous DB)
Direct SQL query	S3 Select, Athena	Query Acceleration, Synapse	BigQuery External Tables	OCI Data Flow (Spark)

Common Mistakes

No lifecycle policy configured — Without a lifecycle policy, old data remains in a high-cost storage class indefinitely, accumulating unnecessary cost.
Leaving public access unblocked — Failing to block public access to a bucket/container can expose sensitive data to the internet.
Versioning not enabled — Without versioning enabled, accidentally overwritten or deleted data cannot be recovered.