AWS Growth Hits 39.3%, Its Fastest Pace in Nearly Five Years, Says Jassy

Growth, backlog and chip installations

Jassy said AWS posted the 39.3% growth despite lapping a comparable year-over-year period in which it added a lot of capacity. According to Amazon, backlog is $578 billion, growing at triple-digit rates year over year, and is well diversified.

Jassy said the company is installing an AI accelerator about every 14 seconds, and that customers are taking up capacity almost as fast as it comes online. Core compute and storage revenue grew at their fastest rates in several years, he said.

On profitability, Jassy said operating margins on the Core offering have long been strong and are improving rapidly on the AI side, which he described as well ahead of where Core was at a similar stage. He said this holds even with heightened component costs such as memory and significant depreciation from capital spending. Amazon says it continues to accelerate its capex payback.

The company's chips business is also growing triple-digit percentages year over year, according to Jassy, and revenue commitments for Graviton and Trainium have each increased more than 20x year over year.

The flywheel argument

Jassy framed the results with the flywheel idea from Jim Collins' Good to Great. In his telling, leading functionality, security and operational performance draw customers to run apps and store data on AWS. That drives AI demand, because customers prefer production inference to sit near the rest of their data and apps. AI in turn drives Core demand, since it uses Core for work such as reinforcement learning, post-training and tool usage, which feeds reinvestment into broader AWS functionality.

He contrasted this with a model of primarily hosting GPUs or renting short-term datacenter capacity, calling AWS's approach a long-term platform strategy with attractive long-term economics regardless of supply scarcity.

On the Core side, Jassy said about 85% of global IT spend is still on-premises and that a 10 to 20 year cycle will flip that toward the cloud. He said workloads once thought unlikely to move, such as mainframes and storage array networks, are now migrating thanks to AWS cloud substitutes and AI-driven migration capabilities.

He also announced that, with its acquisition of DuckLabs, DuckDB will become the universal SQL dialect of AWS storage and analytics. Customers will be able to write SQL queries once and run them on any AWS analytics service, with zero migration or portability costs between query engines, Jassy said.

Bedrock, security and agents

Jassy said Amazon Bedrock, the company's inference service, is seeing remarkable growth. Fortune 500 companies spent nearly 7x more on it year over year, and total token volume grew nearly 17x. He called it the fastest growing multi-billion dollar service in AWS's history. OpenAI models were recently added and usage is growing quickly, and Bedrock-managed versions of Kimi, GLM and SpaceX models were also added.

On data protection, Amazon says inputs and outputs are never shared with model providers or used to train base models. When customers tune a foundation model, a private copy is created for their use, and data is encrypted in transit and at rest.

AWS Continuum, built to discover, prioritize, validate and remediate code vulnerabilities at machine speed, is already used by Capital One, IBM, Infor, Luminate and Robinhood. Amazon announced in Q3 that it will integrate the service into Anthropic's Claude Code, OpenAI's Codex and AWS's Kiro. Jassy said customer interest exceeds anything the company has seen for a new security service.

Tens of thousands of customers use Bedrock AgentCore, deploying 2x more agents in Q3 than in the prior three quarters combined, according to Jassy. Turnkey services include Kiro for coding, Amazon Connect for contact center agents and Amazon Transform for migrations. Amazon Quick, its AI work companion, is being chosen by enterprises such as Mondelez, EchoStar, 3M, Southwest Airlines and DXC.

What it means

The figures are Amazon's own claims from its earnings call. They indicate that the company sees AI demand lifting its traditional cloud business rather than replacing it, and that it is betting on custom chips, managed model access and agent tooling to keep that demand on its platform. Jassy said he has more conviction than before that the AWS investment will yield a much larger revenue and free-cash-flow generating business, with strong return on invested capital.