All the numbers: Amazon Prime Day 2026 powered by AWS

During the four‑day sales event, Amazon.com leaned heavily on its own cloud. Almost half of the EC2 workload ran on ARM‑based Graviton processors, while Elastic Block Store logged more than 24.8 trillion I/O operations and moved over an exabyte of data each day. Serverless compute surged to 2.3 trillion Lambda invocations per day, and the container platform launched an average of 158 million ECS tasks daily, a 48 % jump from the previous year. Content delivery via CloudFront handled 2.1 trillion HTTP requests, and DynamoDB processed 59 trillion requests, peaking at 192 million per second. Aurora stored 5,491 TB and moved 1,194 TB, while ElastiCache delivered 2.3 quadrillion requests in a day. Messaging pipelines—SNS, SQS, Kinesis—recorded multi‑trillion‑scale traffic, and security‑monitoring services like GuardDuty and CloudTrail saw 59 % and 44 % year‑over‑year increases respectively. The scale required over 44,000 fault‑injection experiments to validate resilience, underscoring the engineering effort behind the consumer experience.

These figures illustrate how Amazon is using its own cloud to push the envelope of what a retail platform can handle. The dominance of Graviton indicates a strategic shift toward custom silicon for cost and performance gains, echoing similar moves by Google (TPU) and Microsoft (Azure Arm). The massive serverless and container usage signals that Amazon is consolidating its micro‑services architecture on fully managed services, reducing operational overhead while maximizing elasticity. Compared with prior Prime Days, the 47 % rise in ECS/Fargate tasks and the 5 % lift in CloudFront requests suggest that traffic patterns are becoming more distributed and latency‑sensitive, likely driven by global expansion and richer interactive experiences such as AR product previews.

Looking ahead, the sheer volume of data and request rates raises questions about cost efficiency and environmental impact; while AWS can absorb spikes, the underlying expense may pressure margins if not offset by higher conversion rates. The reliance on automated fault‑injection and continuous monitoring highlights a growing security surface that could attract sophisticated attacks, making the robustness of GuardDuty and CloudTrail critical. Competitors will watch whether Amazon can translate this internal capability into a commercial advantage for external customers via services like AWS Countdown Premium, potentially reshaping the market for event‑driven cloud scaling.