EN
Back to the archive

The encyclopedia · Product & Design · Product decision · 2006–2010

Amazon S3 sold storage by the gigabyte over HTTP and ended server buying

Launched March 2006, S3 offered unlimited object storage through a REST/SOAP API at 15 cents per GB-month, so startups stored data without provisioning servers.

Amazon Web Services

The solution

In March 2006 Amazon quietly launched S3 — Simple Storage Service — a data store accessed over the web. Jeff Barr's announcement post described 'reliable, highly scalable, low-latency data storage' where developers store any number of blocks, each up to 5 GB, associated with a user-defined key and metadata.

The radical part was pricing: 15 cents per GB per month stored and 20 cents per GB transferred, with no minimums. Storage became a utility like electricity — pay for what you use — while Amazon ran the machines. That inverted the old model of buying servers and capacity you might never need.

S3's simple key-value interface and REST API meant any program could store and fetch objects from anywhere, which made it the default backend for a generation of startups; combined with EC2, it became the foundation of cloud computing as an industry.

Why it worked

  • Pay-as-you-go pricing removed capacity planning
  • REST API made storage callable from any program
  • Infinite-scale illusion for a team of one
  • Per-object ACLs kept security simple
What it achievedSell storage as a utility: pay per byte, scale to infinityclever

What can be applied

Customers don't want capacity, they want the capability: metered, API-first delivery converts an expensive asset (idle servers) into a service with no minimum commitment.

Aftermath

S3 became the canonical object store, hosting the web's largest data lakes and backing countless startups; by the 2020s it stored hundreds of trillions of objects and helped make AWS a hundred-billion-dollar-a-year business. Its key-value design remains the template for object storage everywhere.

Sources

spotted an error? The archive wants to know.

Related cases