DISRUPTIVE / BUSINESS / RETROSPECTIVE

FILE Nº 022

Business · Retrospective

AWS at Twenty: The Accidental Cloud That Ate the Industry

On March 14, 2006, Amazon launched a storage service. Nobody called it a revolution. Twenty years later, it is the infrastructure the internet runs on.

AWS data centers in the us-west-2 availability zone, Oregon
AWS data centers in Oregon’s us-west-2 availability zone, with a fourth under construction. Twenty years ago, Amazon was a bookstore that had built, almost by accident, one of the world’s great utilities.Photo: Tedder, Wikimedia Commons (CC BY-SA 4.0) Source

Twenty years ago, Amazon was a bookstore that had survived the dot-com crash and built, almost by accident, one of the world’s great computing infrastructures. On March 14, 2006, it opened that infrastructure to the world: Amazon S3, the Simple Storage Service, launched in the United States — the first AWS infrastructure service generally available, letting any developer store and retrieve any amount of data, at any time, from anywhere on the web.

Three AWS data centers in Oregon
Three AWS data centers in a single availability zone. What Amazon had built, without quite intending to, was the utility model for computing — metered, instant, and rented by the hour.Photo: Tedder, Wikimedia Commons (CC BY-SA 4.0) Source

The idea had been circulating inside Amazon since 2003, when engineer Benjamin Black and his manager Chris Pinkham wrote a memo proposing the company sell access to virtual servers as a service — turning sunk infrastructure investment into revenue. S3 was the first product of that thinking; EC2, rented virtual computers at ten cents an hour, followed in August 2006. Early adopters proved the thesis fast: the photo startup SmugMug attributed around $400,000 in storage-cost savings to S3 within the year. As one history of the launch put it: “To many at Amazon, that day marked the birth of AWS.”

The Amazon Spheres at Amazon's Seattle headquarters campus
The Spheres at Amazon’s Seattle headquarters. The cloud idea had been circulating inside Amazon since 2003, when engineer Benjamin Black and his manager Chris Pinkham wrote a memo proposing to sell the company’s infrastructure as a service.Photo: Andy Li, Wikimedia Commons (CC0) Source

What Amazon had built, without quite intending to, was the utility model for computing. Before AWS, starting a technology company meant buying servers — capital expenditure, raised upfront, on hardware that would be obsolete in three years. After AWS, it meant renting capacity by the hour — an operating expense that scaled with the business. The cloud turned fixed costs variable, and in doing so, it lowered the cost of starting a company by orders of magnitude.

The disruption

The deepest disruption wasn’t technical; it was financial. AWS didn’t just sell storage and compute — it sold the removal of permission. A founder no longer needed to convince anyone to buy hardware before writing code. That single change is downstream of half the startup history of the last twenty years: the companies that scaled on AWS, the venture model that funded them, and the incumbents — IBM, HP, Oracle — who watched their enterprise businesses get rented out from under them by the hour.

The Amazon data center in Boardman, Oregon, in 2013
Amazon’s data center in Boardman, Oregon, in 2013. The deepest disruption wasn’t technical; it was financial — AWS sold the removal of permission: no capital committee, no procurement cycle, just a credit card and an API.Photo: Visitor7, Wikimedia Commons (CC BY-SA 3.0) Source

Twenty years on

AWS is now the default answer to “where should this run?” — which makes it easy to forget how strange the idea once sounded. A retailer selling spare server capacity? In 2006, it read as a curiosity. In retrospect, it was the purest Schumpeterian move of the century: new infrastructure, new economics, old industry destroyed from within. The cloud didn’t just eat the server room. It ate the idea that computing was something you owned.