sourced story
14 March 2006 (S3 launch)Primary source · 2 sourcesWell documented

Amazon launches S3 and EC2, starting the cloud

Amazon rents out the spare infrastructure it built to run its own store

On the timeline · around 14 March 2006 (S3 launch) · Social, Mobile, and the CloudThe Web and the Dot-Com BoomSocial, Mobile, and the CloudAmazon launches S3 and EC2, starting the cloud20022004200620082010

Quick facts

S3 launched
14 March 2006
EC2 launched
Limited beta, 25 August 2006
Parent company
Amazon.com
Core idea
Renting storage and compute as a metered service

What happened

Amazon had learned, running its own retail business, how expensive and difficult it was to provision and manage computing infrastructure, which distracted engineering teams from building the actual product. On 14 March 2006, Amazon launched Simple Storage Service (S3), letting any customer store data on Amazon's own servers instead of buying and maintaining storage hardware; 12,000 developers signed up within a single day. Amazon followed a few months later, in a limited public beta announced 25 August 2006, with Elastic Compute Cloud (EC2), which gave customers on-demand access to computing power itself, letting them run applications on Amazon's servers rather than their own, paid for by usage rather than owned outright.

Why it matters

S3 and EC2 turned computing infrastructure into a utility a startup could rent by the hour instead of capital equipment it had to buy upfront, removing one of the biggest barriers to launching a new internet company and setting the pattern every other major cloud provider, including Microsoft Azure and Google Cloud, later followed.

How we know

Amazon Web Services documents its own founding motivation and the S3 and EC2 launch dates directly, describing the shift from owning hardware to renting infrastructure as a service.

Sources

See something wrong? . Corrections with a source get fixed fastest.

Related timelines

  • Artificial Intelligence · The rentable, on-demand compute AWS pioneered later became the infrastructure that trained and served large-scale machine learning models.
Part of a timelineThe Internet and Computing36 events · From a mechanical engine that never ran to a network that never sleepsView all →