Ubuntu Pro is now available on AWS Deep Learning Amazon Machine Images (DLAMI)

Key points

AWS has published its first Deep Learning AMIs (DLAMIs) built on Ubuntu Pro, in the Base GPU family, starting in September 2026. The images include Ubuntu Pro's Expanded Security Maintenance (ESM) from first launch, and Canonical describes the availability on its blog.

About DLAMIs

• DLAMIs are machine images that AWS maintains for GPU and accelerator instances. They come with the NVIDIA driver, CUDA, the Elastic Fabric Adapter (EFA) networking stack and related tooling installed and tested together.
• AWS builds 31 DLAMI lines: Base AMIs with drivers only, framework images with PyTorch preinstalled, and Neuron images for Trainium and Inferentia.

Images and versions

• Four Ubuntu Pro Base GPU images are available, covering Ubuntu Pro 24.04 LTS and 22.04 LTS on x86_64 and ARM64.
• The 24.04 LTS images use kernel 7.0.0-1013-aws, and the 22.04 LTS images use 6.8.0-1064-aws.
• All four ship NVIDIA driver 595.91.07 and CUDA 12.8, 12.9, 13.0 and 13.2, with 13.2 as the default.
• The x86_64 images support 14 GPU instance families, from G4dn and G5 through P4d, P5 and P5en up to the Blackwell P6-B200 and P6-B300.
• The ARM64 images support G5g and the Grace Blackwell P6e-GB200 and P6e-GB300 families.
• Grace Blackwell support works out of the box, which is a first on AWS. Ubuntu Pro 24.04 LTS covers both GB200 and GB300, while 22.04 LTS covers only P6e-GB200.

Kernel and security updates

• The kernel is pinned, and the NVIDIA and EFA drivers are compiled against it. AWS holds the kernel packages and delivers security fixes as a new image every week or two.
• ESM extends Canonical security updates from the Main repository to Universe and lengthens coverage from 5 to 10 years.
• Kernel Livepatch applies in-memory security fixes to the running kernel to avoid unplanned reboots.
• On ARM64, Kernel Livepatch is available only from Ubuntu 26.04 LTS onwards.

Universe packages covered by ESM at launch

• The base image includes 38 Universe packages, all covered by Ubuntu Pro from first boot.
• MPI and RDMA: libopenmpi3t64, libfabric1, libpmix2t64, libucx0, libpsm2-2, libpsm-infinipath1, libhwloc15, libhwloc-plugins, libibdm1, libopensm9, libosmcomp5, libosmvendor5, ibutils, libmunge2.
• GPU and accelerator runtime: libamdhip64-5, libamd-comgr2, libhsa-runtime64-1, libhsakmt1, ocl-icd-libopencl1.
• Python packaging: python3-pip, python3-pip-whl, python3-wheel, python3-venv, python3.12-venv, python3-setuptools-whl.
• Other packages: apt-transport-https, check, environment-modules, fuse, libfuse2t64, libsnmp-dev, libsnmp-perl, libnetsnmptrapd40t64, libsubunit0, libsubunit-dev, pahole, pigz.
• Deployments built on the image usually pull in more dependencies, mostly from Universe. Those packages receive security support only through a Pro subscription via ESM.

How to get the images

• Each image line has a public AWS Systems Manager (SSM) parameter that points to its latest build, which you can query with the AWS CLI.
• The x86_64 parameter for 24.04 is:

aws ssm get-parameters --names /aws/service/deeplearning/ami/x86_64/base-oss-nvidia-driver-gpu-ubuntu-pro-24.04/latest/ami-id

• The ARM64 parameter for 24.04, covering Graviton and Grace Blackwell, is:

aws ssm get-parameters --names /aws/service/deeplearning/ami/arm64/base-oss-nvidia-driver-gpu-ubuntu-pro-24.04/latest/ami-id

Checking coverage and subscription

• To see which installed packages ESM covers, run this on the Ubuntu Pro 24.04 LTS Base GPU image:

pro security-status --esm-apps

• To confirm the subscription is active and see entitlements, run this on a running instance:

pro status

• The AWS Deep Learning AMI documentation has setup guides, the support policy table and the per-image lists of supported instance types and installed packages.

Summary by DistroNews, based on the original post.