Long-running serverless work has often forced IT teams to split jobs into shorter stages or move them to managed infrastructure. Amazon.com, Inc. (NASDAQ: AMZN), through Amazon Web Services, has raised the maximum timeout for Lambda Event Functions to 30 minutes, doubling the prior 15-minute limit and giving teams more room to run AI inference, data-processing and media workloads without redesigning an application solely around execution duration.
The change applies to synchronous, asynchronous and event-source-mapping invocations of Event Functions that run on demand. AWS also extended synchronous invocations on Lambda Managed Instances to 30 minutes. The company said asynchronous and event-source-mapping invocations on Lambda Managed Instances retain their previously announced 90-minute maximum.
The practical benefit is most apparent for jobs that are too variable to justify always-on capacity but can exceed a quarter hour at peak demand. AWS cited chained model calls and tool invocations, media transcoding, file conversion, financial calculations, report generation and extract-transform-load steps triggered by Amazon SQS or Amazon S3 events. A longer supported duration can let teams keep those workloads in a serverless operating model rather than add an architectural handoff just to avoid a timeout.
The expanded limit does not remove the operating constraints of a long-running function. AWS said teams can set an Event Function timeout as high as 1,800 seconds, while Lambda rejects higher values. Existing runtime, handler, IAM execution-role and virtual-private-cloud configurations remain in place. For an event-source mapping, an organization must also allow time for the setting to propagate after it is changed.
Lambda durable functions can extend the usefulness of the new limit for workloads that cannot cheaply repeat work after a failure. The company said durable functions checkpoint application state through step and wait operations, allowing an execution to resume from its last checkpoint rather than rerun completed steps. Individual invocations still follow the applicable function timeout, but a durable execution can run for up to one year under the documented conditions.
That distinction matters for enterprise design. A 25-minute model-inference task that fails late in its run can restart from the beginning without checkpoints, AWS said, while a durable-function workflow can resume completed work. Teams need to weigh that resilience against the added design and testing requirements of checkpointed workflows; the longer timeout is a capacity option, not an assurance that a job is idempotent or safe to retry.
Longer invocation times also change capacity planning. AWS noted that concurrency equals request rate multiplied by average duration, so a workload running for 30 minutes instead of 15 minutes needs twice the concurrency at the same arrival rate. The company advises customers to review account and reserved-concurrency quotas, and to ensure that downstream connection, credential and client timeout settings accommodate the longer execution window.
There are workload-specific limits to account for. For Amazon SQS event sources, AWS recommends setting queue visibility timeout to at least six times the function timeout, which means at least three hours for a 30-minute function. Certain Amazon MQ, Amazon DocumentDB and older Kafka event-source mappings remain capped at 14 minutes. Organizations processing streams also need to monitor for backlogs because a longer invocation can slow shard processing and increase the amount of work lost when a batch fails.
Legacy serverless architectures often treat a 15-minute boundary as a hard design constraint, pushing longer tasks into separate queues, containers or staged workflows. AWS Lambda’s new 30-minute ceiling will not make every extended process suitable for a single invocation, but it gives engineering teams a larger supported execution window. That raises the competitive expectation for serverless platforms: they must pair longer-running capacity with the controls, observability and failure-handling guidance that enterprise workloads require.
