Cloud computing

Amazon had a problem that became an opportunity. Running the world's largest e-commerce operation required massive computing infrastructure—servers, storage, networking—that sat largely idle most of the year, then surged during holiday shopping peaks. Engineers realized they had built something valuable: a distributed computing platform that could scale on demand, with APIs that made adding or removing resources trivially easy. Why not sell this capability to others?

Amazon Web Services launched in 2006, initially offering storage (S3) and compute (EC2). The business model was revolutionary: instead of buying servers, companies would rent computing by the hour. No capital expenditure, no long-term commitments, no managing physical hardware. Startups could launch with the same infrastructure available to Fortune 500 companies. The 'cloud' metaphor—computing resources floating somewhere in the network, accessible from anywhere—captured the abstraction perfectly.

The adjacent possible had been assembling since the 1990s. Virtualization technology allowed multiple virtual machines to run on single physical servers. Data center design had evolved through web-scale companies. Broadband internet made remote computing practical. Web services and APIs standardized machine-to-machine communication. What Amazon contributed was packaging these technologies into a commercial service with pay-as-you-go pricing and aggressive automation.

The cascade reshaped the technology industry. Startups no longer needed to raise capital for servers before writing code. Enterprise IT departments could 'burst' to the cloud during peak demand. Netflix moved its streaming infrastructure entirely to AWS, proving that even demanding workloads could run on shared infrastructure. Microsoft launched Azure in 2010; Google Cloud Platform followed. By 2020, AWS alone generated over $40 billion in annual revenue.

Cloud computing created new patterns: 'serverless' architectures where developers wrote functions without managing servers at all, 'infrastructure as code' where entire systems were defined in version-controlled templates, 'multicloud' strategies hedging across providers. It also created dependencies—when AWS suffered outages, significant portions of the internet went dark. By 2025, the majority of enterprise computing ran on cloud platforms, transforming IT from a capital expenditure to an operating expense. The data center, once a strategic asset, had become a commodity—and Amazon's solution to its own scaling problem had become the foundation for a new computing paradigm.