Amazon Web Services has reached a $184 billion annualized revenue pace, and Amazon CEO Andy Jassy now sees a possible trillion-dollar business within five to 10 years. He projects that Amazon’s stores operation could reach the same threshold. Those forecasts arrived alongside a harder financial reality: heavy infrastructure spending has pushed the company’s trailing-year free cash flow to negative $42.4 billion.
Contracted Capacity Gives AWS a Clearer Near-Term Outlook
Jassy laid out his projections Thursday during Amazon’s third-quarter earnings discussion. Morgan Stanley analyst Brian Nowak pressed him on the unusually distant forecast, asking what supported that view.
For AWS, Jassy’s answer rested first on committed demand. Revenue growth reached 39% in the third quarter. He expects that growth to exceed 40% in both 2027 and 2028. A $578 billion backlog represents contracted revenue, giving the cloud operation a substantial base for that outlook.
Some of its largest customers reserve capacity in blocks reaching a gigawatt, Jassy said. Those commitments give Amazon visibility into demand through 2027 and 2028, extending into part of 2029. At that scale, planning physical infrastructure requires as much attention as selling software.
The forecast gets less certain further out. Jassy acknowledged that he can’t reliably predict growth several years ahead, especially at the rates AWS is experiencing. Contracted demand supports the near-term outlook, but moving beyond $1 trillion would require expansion to continue well past that window.
His expectations have moved sharply. Jassy told employees in March that AWS could achieve at least $600 billion in yearly revenue within roughly a decade, a figure already twice his previous expectation. His newer projection points to a much larger opportunity, though there’s no guarantee AWS will reach it.
AI Applications Must Turn Infrastructure Into Recurring Demand
Jassy’s broader argument is that business adoption of AI remains early. He pointed to declining model-running costs and emerging applications as reasons demand could keep expanding beyond the current capacity commitments.
Security work offers one concrete example. The models Jassy cited identify vulnerabilities and repair them, tying cloud consumption to operational work rather than just AI experiments.
For example, a security team could use such a model to inspect software, flag a flaw and suggest a correction for review. Used that way, the application would address a clear problem within an existing workflow. Its practical value would depend on checking the findings and proposed repairs.
Lower operating costs could make workflows like that more economical to deploy. They don’t, by themselves, establish that customers will use enough additional capacity to sustain Amazon’s forecast. The commercial question is how effectively new applications convert cheaper inference into broader, repeated use.
For AWS, the platform opportunity therefore extends beyond supplying compute for model development. Jassy’s thesis depends on businesses putting AI into routine operations, creating demand that persists after initial testing. Customers are committing to capacity, but those reservations don’t establish what returns they’ll earn.
The Stores Forecast Has a Different Growth Engine
Amazon’s retail projection rests on market share and a continuing shift toward online shopping. Jassy said the stores operation added more than $90 billion in gross sales during the trailing year, with quarterly additions running at roughly twice those of its next-largest U.S. rival.
Unit sales have increased nearly 15% this year, he said. He also noted that almost 80% of retail transactions still occur in physical locations, leaving room for more spending to move online.
The starting point for his trillion-dollar projection is less clear than it is for AWS.
| Business measure | Revenue base | What it establishes |
|---|---|---|
| AWS annualized revenue pace | $184 billion | A run-rate measure, rather than completed yearly revenue |
| Online and physical store sales | $309 billion | Revenue generated during the trailing year |
| Stores plus third-party seller services and advertising | About $575 billion | A broader potential interpretation of the retail operation |
Amazon has not clarified which retail scope Jassy intended. The scope makes a substantial difference to how far the business has to grow. For comparison, Walmart, the largest retailer globally, recorded $713 billion in revenue during its latest fiscal year, though that benchmark doesn’t resolve the definition.
The two forecasts should consequently be read separately. AWS relies heavily on expanding demand for cloud infrastructure and AI applications. Retail depends on competitive gains and changes in shopping behavior; Jassy’s case does not establish AI as the driver of that projection.
Spending Is Arriving Before the Projected Revenue
Amazon expects capital expenditure of about $220 billion this year, largely directed toward AWS. Jassy said spending would increase substantially in 2027. He argues that the company can see attractive returns ahead and is recovering its investment faster than previously.
The current financing burden is already visible:
- Free cash flow for the trailing year dropped from positive $14.8 billion a year earlier to negative $42.4 billion.
- This year, Amazon raised $115 billion in long-term debt.
- Third-quarter interest expense exceeded three times its earlier level, reaching $1.7 billion and reducing profit.
Shares dropped around 6% during after-hours trading after earnings missed Wall Street’s profit expectations. That reaction highlights the gap between a large future market and the cost of building for it now.
To meet Amazon’s forecast for AI adoption, capacity has to be funded before every application proves its value. Contracted demand gives AWS a credible foundation for the near term, while security workflows illustrate how usage could spread. The immediate test is how much cash that expansion generates: enough to cover rising investment and debt costs, or just higher revenue.
