Cloud Computing and Data Centers

Amazon Launches GPU Support in AWS Lambda

AWS announced the provision of GPU support (G-series) within Lambda Managed Instances functions, enabling the execution of AI workloads and media processing in a serverless computing style.

2026-11-18
2 min read
18 views
certi.news Editorial Team
Amazon Launches GPU Support in AWS Lambda

Amazon Web Services (AWS) announced a fundamental update to the AWS Lambda platform, where developers can now run functions that rely on EC2 G-series GPUs within the Lambda Managed Instances (LMI) environment. This move aims to integrate the power of graphics computing with the flexibility and ease of the serverless computing model.

What changes in practice?

This update allows developers to run intensive workloads such as AI model inference, video processing, 3D rendering, and scientific computing, without the need to manage complex infrastructure. Previously, developers had to deal with GPU drivers, CUDA versions, and capacity planning; now, Lambda handles instance lifecycle management, auto-scaling, and event integration.

Key Technical Capabilities

  • Container Deployment: GPU-accelerated functions are deployed as container images (OCI), facilitating the development and deployment process.
  • Event Integration: GPU functions can be triggered via more than 200 AWS event sources such as S3, SQS, and API Gateway.
  • Utilization Optimization: The platform allows for the integration of multiple functions on a single instance, increasing GPU utilization efficiency and reducing costs compared to allocating a full instance to each task.
  • Long-running Task Support: The execution timeout has been extended to reach 90 minutes, supporting large batch processing or complex inference operations.

Why does this news matter?

This announcement solves two main problems previously faced by developers: the complexity of managing GPU infrastructure and high costs due to low resource utilization rates. Through this model, users pay the base EC2 instance fees plus a management fee of 9.75%, while benefiting from auto-scaling features that ensure resources are only running when needed.

The update also supports container images up to 32 GB in size and up to 32 GB of RAM, with the ability to process up to 48 concurrent requests in a single execution environment, making it an ideal choice for custom model inference applications that do not require massive infrastructure like that used in Amazon Bedrock.

News source
AWS News Blog
Open original source ↗
c
Author

certi.news Editorial Team

In the same category

You may also like

View all news