Red Hat Will Now Ship CUDA Toolkit Directly in Its Products

Red Hat now bundles NVIDIA CUDA with its operating systems, so enterprises can install and update GPU software through the same tools they use for the rest of the stack.

Red Hat will begin shipping NVIDIA’s CUDA Toolkit inside Red Hat Enterprise Linux, OpenShift, and the new Red Hat AI product line. The move lets users install GPU drivers, libraries, and container toolkits directly from Red Hat repositories instead of downloading them separately from NVIDIA.

This brings several benefits that include:

• Less setup time: A single dnf or yum command pulls the entire CUDA stack and keeps it updated with the rest of the operating system.
• Fewer version conflicts: Red Hat and NVIDIA test each combined release, so driver, runtime, and container versions stay in sync across on-prem clusters and cloud instances.
• Single support path: Red Hat support covers both the OS layer and the CUDA packages, removing the need to open tickets with two vendors for driver-related problems.

As a developer you get the followings:

• CUDA runtime, compiler, math libraries, and the NVIDIA Container Toolkit.
• Matching kernel drivers packaged as kmod RPMs that rebuild automatically after kernel updates.
• Ansible automation roles and OpenShift operators so clusters can be scaled or updated without manual steps.

Availability / Download

The packages will appear soon in RHEL and OpenShift software channels; older releases shall follow in the next minor update cycle. Red Hat AI, the company’s new appliance for training and serving models, ships with the CUDA integration from day one.

Licensing and cost

CUDA remains under NVIDIA’s existing license; Red Hat is simply redistributing the binaries. Customers still need NVIDIA hardware and must accept NVIDIA’s license terms, but no extra subscription fee is charged for the CUDA packages themselves.

Via news story

Recent articles from DebugPoint.com