The AI revolution is built on a physical world of servers and data centers most of us will never see. To grasp its true scale, you have to look past the hype and dig into the numbers. Amazon’s financial statements are a perfect place to start.
Forget the Headline Revenue Growth
The first
number everyone looks at for Amazon Web Services (AWS) is revenue growth. In the third quarter of 2026, AWS posted impressive 39.3% growth, its fastest in nearly five years. While impressive, that figure only tells part of the story. It reflects what customers are paying for now, not what Amazon is building for the future. The real story of AI demand is in the expenses, specifically the massive, almost unprecedented, capital expenditures—or capex. This is the money Amazon is spending on property and equipment, and it serves as a direct proxy for the company's confidence in future AI-driven growth.
The Capital Expenditure Tsunami
Amazon's spending is staggering. The company is on track to spend approximately $220 billion on capital expenditures in 2026 alone, with a significant portion dedicated to building out AWS's AI capabilities. Third-quarter capex hit nearly $76 billion. This spending spree has even pushed the tech giant’s free cash flow into negative territory, a move that initially spooked some investors. However, CEO Andy Jassy has compared this investment cycle to the early days of AWS, arguing that building capacity ahead of demand is necessary to capture what he calls a 'once in a lifetime opportunity'. When a company of Amazon's scale is willing to sacrifice short-term cash flow for long-term infrastructure, it’s the strongest possible signal that the demand it sees on the horizon is monumental.
The Language of Investment
The clues aren't just in the numbers, but in the words executives use to describe them. During recent earnings calls, Amazon's leadership has been explicit about the drivers behind the spending. Jassy noted that 'almost as fast as we install capacity, customers are gobbling it up', adding that the company was installing a new AI accelerator roughly every 14 seconds. The AWS order backlog, which represents future contracts, has swelled to $578 billion. This combination of a rapidly growing backlog and candid commentary about struggling to keep up with demand paints a vivid picture. They aren't just building for a theoretical future; they are racing to fulfill existing and anticipated orders from a customer base that is rapidly embracing AI.
Decoding Depreciation and Custom Chips
A more subtle clue lies in the depreciation of assets. As AI models and the chips that run them evolve at a blistering pace, the useful life of a server is getting shorter. An increase in the rate of depreciation can signal that the company anticipates a faster, more expensive upgrade cycle to keep its infrastructure at the cutting edge. Furthermore, Amazon is increasingly focused on its own custom silicon, like its Trainium chips. Jassy has stated that the cost of running AI models is still 'an order of magnitude too expensive' and that developing its own chips is key to bringing that cost down. This focus on custom hardware is a long term play to control its destiny and improve margins in a world where AI workloads become a dominant and permanent part of the cloud computing landscape. According to AWS's CFO, the payback period for these AI server investments is already under three years, suggesting the economics are proving successful.
