For years, Amazon Web Services (AWS) was the undisputed cash cow of the Amazon empire, a profit machine with enviable margins. But the AI gold rush is changing the equation, forcing a high-stakes bet that pits massive upfront costs against future dominance.
The Profit Engine: What Is an Operating Margin?
Before diving into the numbers, let's clarify the key metric: operating margin. Think of it as a measure of profitability. For every dollar of sales a company makes, the operating margin shows what percentage is left over after paying for all the day-to-day costs of doing business—things like wages, materials, and marketing, but before interest and taxes. For a long time, AWS posted incredibly high margins. In its latest quarter, the division reported a stunning 38.7% operating margin. This means that for every dollar of cloud services it sold, it kept nearly 39 cents as operating profit. This incredible efficiency is why AWS, which accounted for just over 22% of Amazon's total sales, generated a whopping 63% of the company's entire operating income in the third quarter of 2026. It's the engine that has funded much of Amazon's broader expansion.
The Old Math: How Cloud Computing Was a Money-Printing Machine
AWS's traditional business model was a masterclass in economies of scale. By building colossal data centers and renting out slices of computing power, storage, and database services to millions of customers, it created a flywheel. More customers led to more infrastructure, which lowered costs for everyone and attracted even more customers. The services were standardized, the contracts were often long-term, and the business was highly predictable. This model allowed AWS to generate consistent, high-margin revenue, making it the envy of the tech world and the financial bedrock of Amazon. That reliable formula is now facing its biggest-ever disruption: the artificial intelligence boom.
The New Math: The Staggering Cost of AI Infrastructure
Generative AI requires a completely different and vastly more expensive kind of computing power. The workhorse of AI is the Graphics Processing Unit, or GPU, a specialized chip that costs a fortune. A single high-end NVIDIA H100 GPU can rent for anywhere from around $2 to nearly $7 per hour on a major cloud platform like AWS. Now imagine building data centers packed with tens of thousands of them. This is the reality of the AI arms race. Amazon has committed to capital expenditures of approximately $220 billion for 2026, with plans for that figure to be 'meaningfully more' in 2027. This massive spending spree is necessary to build out the infrastructure to power AI services and satisfy a backlog of customer commitments that has ballooned to $578 billion. These investments are so large that they are turning Amazon's free cash flow—the cash left after paying for operating expenses and capital expenditures—deeply negative, to the tune of negative $42.4 billion over the last year.
The Bottom Line: Margin Pressure vs. Future Dominance
This is the central tension in AWS's new mathematical reality. The company is spending cash faster than ever before to secure its place as a leader in the next generation of computing. While AI-related business is growing at triple-digit rates, the margins on these new services are initially lower than the company's core offerings. Investors are watching closely, weighing the shortterm pain of negative cash flow and massive spending against the promise of longterm gain. CEO Andy Jassy frames it as a 'once in a lifetime opportunity,' projecting AWS can sustain growth above 40% in 2027 and 2028 and eventually become a trilliondollar business. The math has changed from a simple formula of efficiency and scale to a complex calculation of strategic investment. AWS is betting that by burning cash today to build the essential infrastructure for AI, it will own the platform on which the future economy is built, ensuring its profit engine roars for another decade.
