Skip to content

PlatformsAmazon

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

AWS Lambda at a glance

Amazon announced on November 18, 2026 that AWS Lambda now supports GPU instances for Event Functions on Lambda Managed Instances, so teams can run GPU inference on EC2 G-series instances with prebuilt vLLM, SGLang and PyTorch images. Pricing is EC2 On-Demand rates plus a 9.75% management fee.

Ask your AI about this story

Ask what changed in AWS Lambda and Lambda Managed Instances and what to check before using it in your business. Weigh pricing, what it connects to, setup effort and how hard it would be to leave later.

Sources checked Primary source (opens in a new tab)

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, generally available today. Teams can run GPU-accelerated workloads — AI inference, media processing, rendering and scientific computing — on EC2 G-series instance families G4dn, G5, G6, G6f, Gr6, Gr6f and G6e, with Lambda Managed Instances handling provisioning, scaling, patching and load balancing behind the familiar Lambda programming model.

Functions ship as OCI container images of up to 32 GB, and AWS provides prebuilt base images for CUDA, CuPy, PyTorch, vLLM and SGLang, so developers do not install GPU drivers or CUDA toolkits themselves. Each execution environment offers 3 to 48 GB of accelerator memory, up to 32 GB of ephemeral storage and up to 48 concurrent requests, with timeouts extended to 90 minutes on Lambda Managed Instances. Billing is the underlying EC2 On-Demand rate — Savings Plans and Reserved Instances apply — plus a 9.75% management fee for GPU instances and the standard $0.20 per million requests, and capacity scales automatically with demand.

Why AWS Lambda matters for businesses

Why it matters: until now, a business that wanted GPU inference — a custom model answering customers, document or image processing, product-media rendering — had to rent and babysit GPU servers that bill by the hour whether or not work arrives. Lambda's event-driven model lets a small team pay EC2 rates only when events run, use the same vLLM and SGLang stacks the big platforms use, and skip driver and capacity management, which lowers the cost of testing GPU features before committing to dedicated infrastructure.

Put it to work: AI services

Get AI News by email

Verified AI stories, each with its official source, on Monday and Thursday mornings. Add breaking news to hear about the biggest ones as soon as we publish them.

Every story links its official source

Free, in English or Spanish

Unsubscribe with one click, any time

What to send

Sources for this story

  1. Official

Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.

  • #aws
  • #lambda
  • #gpu
  • #serverless
  • #inference
  • #amazon

Want this working in your business?

We turn new AI capabilities into agents, voice receptionists, and automations you can actually use.