AWS Lambda Extends Function Timeout to 90 Minutes for Managed Instances

AWS has announced a significant expansion to AWS Lambda's execution limits: a 90-minute function timeout for asynchronous invocations and event source mappings (ESM) when running on AWS Lambda Managed Instances (LMI). This 6x increase from the traditional 15-minute barrier marks one of the most substantial operational changes to the platform's compute boundaries since Lambda was originally launched. For platform engineers and backend architects, the 15-minute cap was long the primary justification for migrating batch jobs, media transcoding, machine learning inference, and complex ETL pipelines out of serverless environments and into container runtimes like AWS Fargate or Amazon ECS. By extending continuous execution to an hour and a half on managed capacity, engineering teams can retain Lambda's operational simplicity—including built-in routing, granular scaling triggers, and native integrations—without having to engineer synthetic chunking mechanisms or step-function workarounds solely to avoid timeouts. This shift fits into the broader enterprise trend of blurring the lines between pure Function-as-a-Service (FaaS) and managed container/instance capacity. As compute models mature, cloud providers are meeting enterprise requirements by bringing serverless developer experiences and event-driven semantics to longer-running and steady-state computational workloads. Furthermore, because the extended timeout also integrates with Lambda durable functions (which support checkpointing and step replays), durable multi-step executions can now handle longer compute-heavy intermediate stages seamlessly. Practitioners should note that this 90-minute capability applies specifically to asynchronous and ESM invocations on Lambda Managed Instances, rather than default on-demand synchronous invocations. Teams managing heavy data processing pipelines should evaluate their SAM or CloudFormation templates to test increasing timeout configurations up to 5,400 seconds, while monitoring CloudWatch and X-Ray to track execution durability and optimize memory sizing for long-running processes.