The AWS Origin Myth Everyone Believes — And Why It's Completely Wrong
The AWS Origin Myth Everyone Believes — And Why It's Completely Wrong
Why This Matters Right Now
AWS generated $128.7 billion in revenue in 2025. It accounts for less than 20% of Amazon's total sales but produces more than half of its operating profit. This is not a business born from an offhand idea about idle hardware. Yet that myth persists — in business school case studies, YouTube explainers, and investor presentations alike.
Understanding how AWS actually came to be isn't historical trivia. In 2026, Amazon is planning to spend $200 billion in capital expenditures, the majority directed at AWS infrastructure and AI. The strategic logic driving that investment traces directly back to the original insight that created AWS. And that insight has nothing to do with spare servers.
If you're tracking AMZN, competing in tech, or building a startup, the real story of AWS is one of the most instructive case studies in modern business history. [LINK: Amazon's AI strategy deep dive]
Deep Dive: The Numbers Behind the Headlines
Before getting to the origin story, it's worth anchoring in just how large AWS has become — because the scale makes the "spare server" explanation even more implausible in retrospect.
In Q4 2025, AWS posted $35.6 billion in quarterly revenue, its fastest growth rate in 13 quarters at 24% year-over-year. The annual run rate exceeded $142 billion. AWS holds 32% of the global cloud infrastructure market, more than Microsoft Azure (22%) and Google Cloud (12%) combined.
| Metric | AWS (2025) | Azure (est. 2025) | Google Cloud (2025) |
|---|---|---|---|
| Annual Revenue | $128.7B | ~$75B | $43.2B |
| YoY Growth Rate | +19–24% | +39% | +36–48% |
| Market Share (Q4 2025) | 32% | 22% | 12% |
| Operating Income | $45.6B | Not disclosed | $13.9B |
| % of Parent's Op. Income | ~50%+ | ~35% | ~25% |
Sources: Amazon FY2025 Annual Report, Omdia Q4 2025, Microsoft FY2025 Earnings, Alphabet FY2025 Earnings
Impact on Asian and Global Markets: Why Korea Should Care
AWS isn't just an American story. South Korea's tech ecosystem — from Samsung Cloud to Kakao, Naver, and thousands of startups — runs substantially on cloud infrastructure. The question of who controls that infrastructure, on what terms, and at what price, is a direct economic policy question.
Korean tech companies face a structural dependency on US cloud providers. AWS, Azure, and Google Cloud collectively dominate enterprise cloud spending in Korea. Understanding the origins and business model of these platforms helps forecast how pricing power, service availability, and geopolitical risk might evolve. [LINK: Cloud sovereignty in Asia]
There's also the investment angle. AMZN is one of the most widely held global equities by Korean institutional and retail investors. The direction of AWS growth — and the $200 billion capex bet Amazon is making in 2026 — will materially affect returns. AWS's AI revenue run rate already exceeded $15 billion annualized in Q1 2026, growing at triple-digit rates.
The Debate: What Experts Are Getting Wrong
Here's something I think most analysts understate: AWS doesn't have one origin story. According to detailed research by the Acquired podcast and GeekWire's extensive 20th-anniversary investigation, AWS has at least four competing origin narratives — all partially true, none complete.
"Part of the challenge: Amazon, despite operating the storehouse of the internet, isn't great at preserving its own history." — GeekWire, March 2026
The four origin threads: the 2002 e-commerce API launch that showed external developers could build on Amazon's infrastructure; Andy Jassy's 2003 six-pager proposing AWS as a business; the South Africa-based team that independently developed EC2; and the internal SOA (Service-Oriented Architecture) mandate that restructured Amazon's entire engineering approach after 2001.
What all four share — and what distinguishes them from the spare server myth — is that each is rooted in solving a real internal problem at Amazon, not capitalizing on leftover capacity.
The Real Origin: Pain, Not Surplus
By the late 1990s, Amazon's rapid growth was exposing a fundamental architectural problem. The company ran on a tightly coupled monolithic system. Every new feature required touching shared infrastructure. One team's changes could break another team's service. Allocating server resources across projects was chaotic and opaque.
The fix, rolled out after 2001, was Service-Oriented Architecture. Every internal team was required to expose its capabilities via standardized APIs. If team A needed something from team B, it called the API — it didn't get direct access to B's database or servers. This forced discipline: every team had to think of itself as a service provider, not just an internal function.
Jeff Bezos reportedly sent an internal memo requiring all teams to expose data and functionality only through service interfaces, to design those interfaces for external use from the start, and to enforce this rule without exception. This "API mandate" is widely credited as one of the key structural decisions that made AWS possible.
By 2003, someone — the histories disagree on exactly who — recognized that the infrastructure Amazon had built to solve its own problems was itself a product. Other companies faced identical scaling challenges. Amazon could sell the solution. This insight, not a surplus of idle servers, is what became AWS.
What Smart Investors Are Doing Now
The AWS origin story has direct implications for how to think about AMZN as an investment in 2026. A few key considerations:
The moat is structural, not just scale-based. AWS isn't dominant only because it got there first. It's dominant because it established the vocabulary, the pricing model, and the toolchain that hundreds of thousands of developers learned on. Switching costs are enormous — not because AWS locks you in contractually, but because your engineers' brains are AWS-shaped.
The AI capex bet is a double-edged sword. Amazon's $200 billion capex plan for 2026 signals enormous confidence in AI-driven cloud demand. AWS CEO Matt Garman noted the company added nearly 4 gigawatts of computing capacity in 2025 — double what it had in 2022. But capital spending at this scale compresses near-term free cash flow. Investors with shorter horizons should monitor this carefully.
The growth rate gap matters. Azure (+39%) and Google Cloud (+48%) are growing faster than AWS (+24%). On a percentage basis. But on an absolute dollar basis, AWS added $21.2 billion in 2025 versus Google Cloud's $15.5 billion. Jassy made this point explicitly on the Q4 earnings call. The gap is narrowing, but AWS remains the structural winner — for now.
AWS Key Milestones: From Internal Fix to Global Monopoly
| Year | Event | Significance |
|---|---|---|
| 2001 | SOA Mandate Issued | All Amazon teams required to expose capabilities via API — the foundational architectural shift |
| 2002 | Amazon e-Commerce APIs Public | Tens of thousands of external developers signed up within months, validating the "platform" thesis |
| 2003 | Andy Jassy's AWS Six-Pager | Formal internal proposal to productize Amazon's infrastructure for external customers |
| 2006 (March) | S3 Launched | First generally available AWS service: object storage at 15 cents/GB per month |
| 2006 (August) | EC2 Launched | Compute power for 10 cents/hour — no servers to buy, no data centers to build |
| 2015 | AWS Becomes Amazon's #1 Profit Source | Cloud unit overtakes retail as the primary driver of Amazon operating income |
| 2025 | $128.7B Annual Revenue | 32% global market share; AWS operating income of $45.6B — more than 3x Google Cloud |
| 2026 | $200B Capex Plan | Largest single-year infrastructure investment in AWS history, driven by AI demand |
My Take: What Comes Next
AWS will not lose its cloud infrastructure dominance in the next two to three years. The switching costs are too high, the customer backlog ($244 billion at year-end 2025, up 40% YoY) too large, and the operational scale too embedded. I'd be surprised if AWS's share fell below 28% by 2028.
The real battleground is the AI application layer — who controls the platforms on which AI applications are built and run. Microsoft has OpenAI. Google has Gemini. Amazon has Bedrock and a growing suite of foundation models, plus Anthropic as a key partner. This layer is less mature, less locked-in, and therefore genuinely contested.
What I find most fascinating about the AWS story, though, is this: the company that accidentally invented cloud computing didn't do so by accident at all. It did so by taking its own operational pain seriously enough to systematize a solution — and then being bold enough to ask whether anyone else might need the same thing. That's not a business strategy anyone taught in a classroom. It came from the field. And that, more than any technology advantage, is what makes Amazon's history worth understanding.
Sources & Further Reading
- Amazon FY2025 Annual Report / Shareholder Letter (SEC EDGAR, 2026)
- Omdia Cloud Infrastructure Market Report, Q4 2025
- GeekWire: "AWS at 20: Inside the rise of Amazon's cloud empire" (March 22, 2026)
- Acquired Podcast: "Amazon Web Services" episode
- TechCrunch: "AWS revenue continues to soar" (February 5, 2026)
- CNBC: Amazon Q4 2025 Earnings (February 5, 2026)
- Brad Stone, "The Everything Store" (2013)
- Andy Jassy, Amazon Shareholder Letter (2025)
Comments
Post a Comment