Amazon Web Services (2006-2024): the $107B revenue cloud-infrastructure business that defined the category

The 2006 launch and category creation

Amazon Web Services launched its first major services in 2006: Amazon S3 (Simple Storage Service) in March and Amazon EC2 (Elastic Compute Cloud) in August. The strategic thesis was that the same infrastructure Amazon had built to run Amazon.com (storage, compute, networking, monitoring, scaling) could be rented to third-party developers as cloud services. Before AWS, developers who needed servers and storage either bought hardware and operated data centers themselves or rented dedicated servers from hosting companies on multi-year contracts.

AWS's structural innovations were specific. Pay-as-you-go pricing (you pay only for what you use, by the hour or by the byte). API-driven provisioning (developers could spin up new servers in minutes via API call, not days via a sales-and-procurement cycle). Geographic redundancy (servers in multiple AWS regions for failover and latency). The combination changed what was possible for software development: startups could build at scale without raising the capital previously required to set up infrastructure. The category-creation effect compounded across 2008-2015 as AWS added more services (RDS for databases, EBS for block storage, ELB for load balancing, S3 for object storage, etc.) and as enterprise customers began migrating workloads from on-premises data centers to AWS.

The 2015-2024 scaling and competitive response

AWS revenue scaled rapidly through 2010-2018 with consistently high growth rates (often 40-50% year-over-year). The economics were structurally distinctive: high gross margins, high operating margins (frequently 30%+), and continued revenue growth as customer workloads moved to the cloud. AWS effectively funded Amazon's broader investment in everything else during this period — Amazon.com retail, Alexa, Prime Video, etc. By 2018 AWS revenue had crossed $25 billion annually.

Microsoft Azure's competitive response became more serious from approximately 2014 onward under Satya Nadella's leadership. Azure's integration with Microsoft enterprise customers (Office 365, Windows Server, Active Directory) gave it a stronger enterprise-incumbent position than AWS had organically built. Google Cloud Platform grew more slowly through this period due to a less-mature enterprise-sales motion and a smaller initial product set. By 2024 the three-way competition had stabilised: AWS ~30% market share, Azure ~21-25%, Google Cloud ~11-12%, with various smaller competitors (Oracle Cloud, Alibaba Cloud, IBM Cloud) fragmenting the remaining share.

The AI inflection (2023-2025)

The 2022-2023 generative-AI wave produced a substantial inflection for cloud infrastructure demand. Training foundation models requires extraordinary compute infrastructure (thousands of GPUs running for weeks); inference at scale requires substantial ongoing GPU capacity. The biggest AI labs (OpenAI, Anthropic, Google, Meta) all became major cloud customers. The cloud providers responded with massive capital expenditure on GPU infrastructure (Nvidia H100, H200, and successor chips; AWS Trainium and Inferentia; Microsoft Maia; Google TPU).

AWS Q2 2024 revenue grew 19 percent year-over-year, the highest growth rate since 2022. The growth was substantially driven by AI-related infrastructure demand. Matt Garman became AWS CEO in June 2024 (succeeding Adam Selipsky), with the explicit mandate to position AWS for the AI-workload era. AWS has expanded its Anthropic partnership (Anthropic is Amazon's primary frontier-model partner; Amazon has invested billions in Anthropic), its in-house Nova models, and its Trainium and Inferentia chip lineup. The AI-workload competitive question is still being adjudicated through 2025-2026 — whether AWS, Azure, or Google Cloud will win the largest share of frontier-AI infrastructure spend.

How RGM thinks about category-defining infrastructure businesses

When clients ask about infrastructure-as-a-service category creation, AWS is the defining 20-year reference. Three structural lessons. First, the category creation depended on a credible decoupling of infrastructure from the use cases it supports. AWS could only become a cloud-infrastructure leader because Amazon was willing to sell the same infrastructure that powered Amazon.com to its own retail competitors (Walmart and others have notably avoided AWS for that reason; Target, Kroger, and other retailers have used cloud competitors). The willingness to commercialize internal capabilities was structurally hard for many large companies. Second, the scaling economics compounded over years. Pay-as-you-go pricing produced relatively small per-customer revenue at small scale but compounded across millions of customers and billions of API calls. The cumulative effect was a $100B+ annual revenue business with high margins. Third, the competitive response from incumbents (Microsoft Azure, Google Cloud) was meaningful but did not displace AWS's category-leadership position — the first-mover advantage in infrastructure-as-a-service is durable as long as the leader continues to invest at scale.

The pattern is hard to copy in adjacent infrastructure categories without comparable internal-capability commercialization opportunity. Most companies do not have the scale of internal infrastructure to spin off as an external service. The cloud-infrastructure pattern itself has been substantially saturated by AWS, Azure, and Google Cloud — new entrants face structural challenges competing against the established three-way oligopoly. We tell clients in infrastructure categories to expect the AWS pattern to be the exception rather than the template; most infrastructure-services businesses operate at substantially smaller scale than AWS without violating any business-model laws.

Frequently asked questions

Related