Cloud Computing
AWS Lambda Extends Function Timeout to 90 Minutes, Reshaping Serverless Architecture
AWS has lifted execution limits for Lambda Managed Instances to 90 minutes—six times the old cap—enabling longer-running workloads while raising questions about serverless economics.
InfoQ
AWS Lambda Managed Instances can now execute for up to 90 minutes, up from the 15-minute ceiling that has constrained the platform since 2018. This milestone marks a significant shift in how the serverless offering positions itself, as the extension addresses a longstanding friction point in the community.
What Changed#
The new timeout applies specifically to Lambda Managed Instances, which allow multiple requests to run on the same underlying compute resource. Traditional synchronous Lambda invocations remain capped at 15 minutes. AWS has also begun supporting Graviton5-powered instances in the same offering.
The company's 15-minute limit, set in 2018, had become a design bottleneck. Workloads spanning media encoding, complex financial modeling, large-scale data pipelines, machine learning inference, and bulk file operations routinely exceeded that boundary, forcing developers to break jobs into smaller chunks or migrate to other services.
Engineering Considerations for Long-Running Functions#
Running operations for up to 90 minutes introduces new complexity. AWS warns developers to carefully manage three critical areas:
- Network stability: Connections must remain open and functional across the entire invocation.
- Credential freshness: Temporary security credentials need to stay valid without expiration mid-execution.
- Idempotency: Since Lambda does not guarantee exactly-once delivery, failed or retried steps may run multiple times. Operations like payments or database writes must produce identical results regardless of how many times they execute. AWS recommends using Powertools for AWS Lambda to bake idempotency into application logic.
Industry Reception#
The change has drawn mixed reactions. Serverless advocates and AWS community leaders have largely welcomed the move. Yan Cui, an AWS-recognized serverless expert, noted that while the change "blurs the line between running a server and a Lambda invocation," it "makes sense, especially for the Lambda for agentic workflows use case."
