In a landmark announcement published on March 9, 2026, the AlmaLinux project confirmed that NVIDIA has added official support for enterprise Linux–compatible distributions — AlmaLinux among them — beginning with the release of CUDA version 13.2. The move, made possible through a formal collaboration between the two organizations, represents a significant improvement in the GPU software experience for AlmaLinux users worldwide and is expected to eliminate the recurring version-synchronization problems that have periodically disrupted workloads on the platform.

CUDA — NVIDIA’s parallel computing platform and programming model — is the foundational layer through which GPU hardware participates in general-purpose computation. By offloading computationally intensive tasks to the GPU, applications can achieve execution speeds that far exceed what is possible with CPUs alone. This capability has become indispensable in the era of large language models, deep learning, and scientific simulation, making reliable CUDA support on enterprise Linux distributions a high-priority concern for data center operators and researchers alike.

“With the release of version 13.2, NVIDIA has added official support for enterprise Linux compatible distributions, including AlmaLinux, ensuring that our users can get support from NVIDIA when using NVIDIA infrastructure.”
— Jonathan Wright, Infrastructure SIG Lead & ALESCo Member, AlmaLinux Project

What Changed — and Why It Matters

Before this agreement, AlmaLinux had to independently package and distribute NVIDIA’s userspace and CUDA components. That process introduced a small but consequential lag: whenever NVIDIA issued a new driver release, AlmaLinux users would face a brief window in which the kernel-space driver and the userspace CUDA toolkit existed at mismatched versions. In practice, those mismatches could prevent NVIDIA hardware from functioning correctly, requiring users to pin package versions or apply manual workarounds.

The new arrangement closes that gap entirely. NVIDIA-built RPM packages for the userspace stack and CUDA toolkit are now hosted directly at nvidia.repo.almalinux.org and will be distributed through AlmaLinux’s own infrastructure. Because AlmaLinux continues to build and sign the open-source NVIDIA kernel modules itself, both halves of the driver stack will be updated in tandem — the kernel module and the userspace components will always share the same release cadence.

Key Facts at a Glance
  • Official NVIDIA CUDA support begins with CUDA version 13.2
  • Supported distributions include AlmaLinux 9 and AlmaLinux 10
  • CUDA and userspace RPMs are signed by NVIDIA; kernel modules remain signed by AlmaLinux
  • Packages are served from nvidia.repo.almalinux.org, not the AlmaLinux Mirror System
  • Existing users will be migrated automatically on the next system update — no manual steps required
  • Documentation is available at wiki.almalinux.org/documentation/nvidia.html

What Users Need to Do

For the vast majority of existing AlmaLinux users who already have the open-source NVIDIA driver stack installed, the transition is fully automated. The next routine dnf update will silently reconfigure the relevant repository pointers and update all affected packages to the new distribution path. No manual intervention is required.

Users who wish to confirm that they are correctly receiving packages from the new AlmaLinux-hosted NVIDIA repository can run the following command in their terminal:

Terminal — Verify NVIDIA package source
[user@host ~]$ rpm -q almalinux-release-nvidia-driver

The expected output differs by major release version. On a correctly configured system, users should see one of the following:

AlmaLinux OS 9 — Expected output
# AlmaLinux 9 [root@alma9-nvidia ~]# rpm -q almalinux-release-nvidia-driver almalinux-release-nvidia-driver-9-4.el9.x86_64
AlmaLinux OS 10 — Expected output
# AlmaLinux 10 [root@alma10-nvidia ~]# rpm -q almalinux-release-nvidia-driver almalinux-release-nvidia-driver-10-4.el10.x86_64

If DNF errors appear when running the command, it may indicate that NVIDIA packages were installed directly from NVIDIA’s own repositories rather than through AlmaLinux’s managed stack — a configuration that can produce conflicts. The AlmaLinux community provides support for these scenarios at chat.almalinux.org.

Background: An Earlier Milestone

This latest announcement builds on progress made the previous year. Last August 2025, AlmaLinux began shipping NVIDIA’s open GPU kernel modules — a move that was itself greeted with significant community enthusiasm and adoption. The March 2026 announcement extends that foundation by bringing the userspace CUDA stack into the same integrated delivery pipeline, completing the picture of a fully NVIDIA-supported, AlmaLinux-distributed GPU software environment.

Together, the two milestones mark a decisive shift in how enterprise Linux distributions and hardware vendors can collaborate on GPU software delivery — a model likely to be watched closely by other distributions navigating the same challenges.