Amazon reported on 29 October 2026 that AWS generated $46.0 billion in sales during the quarter ended 30 September, up 39.3% from a year earlier. The result measures spending across the cloud business. It does not tell a team whether a particular instance type or AI accelerator will be available for its workload in a chosen location.
What the quarter measured
AWS operating income reached $17.8 billion, compared with $11.4 billion a year earlier, according to the same Amazon results. Amazon also described the quarter’s sales as a $184 billion annualized run rate. That figure multiplies one quarter by four; it is neither a completed year of sales nor a forecast that every later quarter will match it.
Amazon reported a $42.4 billion free cash flow outflow for the trailing 12 months, primarily reflecting higher purchases of property and equipment tied to AI investment. That is a company-wide measure. It cannot be assigned to AWS alone or used as a measure of an individual cloud workload’s economics.
What Amazon says about demand
In its summary of the earnings call, Amazon quoted CEO Andy Jassy saying AWS backlog stood at $578 billion and that the company was installing an AI accelerator roughly every 14 seconds. He also said core compute and storage revenue grew at their fastest rates in several years. These are management descriptions of aggregate demand and construction, not a published inventory of available capacity by Region, Availability Zone or instance family.
The figures support a clear conclusion: customers are spending more on AWS while Amazon is building more infrastructure. They do not establish when a particular buyer can obtain a chosen configuration, what performance it will deliver, or what that deployment will cost.
Checks for a deployment plan
AWS’s instance-type availability guide shows that support differs by Region and can differ among Availability Zones within one Region. Its Capacity Reservations documentation describes a way to secure matching EC2 capacity in a specified zone, subject to account quotas, configuration and billing terms.
A team planning an AI or conventional cloud deployment should therefore confirm the exact service and instance type in its intended Region and zone, check its account quota, review any reservation terms, and test cost and performance with its own workload. Amazon’s Q3 growth provides context for demand. Those local checks determine whether a proposed deployment is feasible.
