Dev.to AI 🤖 Ai 👁 0 📖 2 min read

AWS Lambda adds GPU instances: serverless GPU inference on Lambda Managed Instances with prebuilt vLLM and SGLang images

Amazon announced on November 18, 2026 that AWS Lambda now supports GPU instances for Event Functions on Lambda Managed Instances, allowing serverless GPU inference without creating or managing dedicated GPU clusters. Fun

Amazon announced on November 18, 2026 that AWS Lambda now supports GPU instances for Event Functions on Lambda Managed Instances, allowing serverless GPU inference without creating or managing dedicated GPU clusters. Functions ship as OCI container images of up to 32 GB, and AWS provides prebuilt base images for CUDA, PyTorch, vLLM and SGLang. It is available in preview in US East (N. Virginia), US East (Ohio), US West (Oregon), EU (Ireland), Europe (Frankfurt) and Asia Pacific (Tokyo).

What changed with AWS Lambda

Amazon announced on November 18, 2026 that AWS Lambda now supports GPU instances for Event Functions on Lambda Managed Instances, allowing serverless GPU inference without creating or managing dedicated GPU clusters. Functions ship as OCI container images of up to 32 GB, and AWS provides prebuilt base images for CUDA, PyTorch, vLLM and SGLang. It is available in preview in US East (N. Virginia), US East (Ohio), US West (Oregon), EU (Ireland), Europe (Frankfurt) and Asia Pacific (Tokyo).

AWS Lambda now supports GPU instances for Event Functions, on Lambda Managed Instances. Lambda Managed Instances are the EC2 instances that run your event functions, with provisioned capacity per concurrency and per instance type.

With this new capability, you can run serverless GPU inference workloads by choosing from GPU instances powered by NVIDIA L4S, L40S or H200 GPUs, with 24 GB, 48 GB or 141 GB of GPU memory and up to 8 GPUs per instance. You can also select the CPU and memory configuration.

You can use the GPU instances to run AI agents, LLM serving, and data processing workloads. You can build functions using OCI container images of up to 32 GB, and AWS provides prebuilt base images for CUDA, PyTorch, vLLM and SGLang.

The GPU instances offer a pay-per-use model with per-second billing, with a 60-second minimum, and scale to zero when not in use. Capacity is provisioned for an instance type, either by reserving a fixed concurrency or by autoscaling within minimum and maximum capacity limits, on Lambda Managed Instances.

Why AWS Lambda matters for businesses

Why it matters: until now, a business that wanted GPU inference — a custom model answering customers, or an AI agent running a background task — had to stand up a dedicated GPU cluster, or use a third-party hosting service.

GPU instances on Lambda Managed Instances add GPU inference to Lambda's existing serverless model: per-second billing, scale to zero, and no infrastructure to manage.

Put it to work:

Sources

Originally published at IA Compa.

📰 Read the original article on Dev.to AI

Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.