Red Hat Enterprise Linux for NVIDIA Life Cycle
Overview
Red Hat Enterprise Linux (RHEL) for NVIDIA is a specialized, partner-optimized “fast lane” edition designed to deliver “Day 0” support for cutting-edge NVIDIA hardware, such as the Grace Blackwell and Vera Rubin platforms. To match the velocity of NVIDIA’s hardware and software release cadence, this edition ingests NVIDIA’s drivers, kernel updates, and virtualization patches immediately through CentOS Stream, often before they are fully accepted into the upstream Linux kernel.
RHEL for NVIDIA is an edition built for speed rather than long-term stability, therefore its life cycle and support policies differ from standard Red Hat Enterprise Linux.
Note: RHEL for NVIDIA follows the Red Hat Production Support model except as highlighted in this document.
Impacted Packages
Packages delivered in the Red Hat Enterprise Linux 10 for ARM 64 - Nvidia (RPMs) repository will follow the policies and schedules laid out in this document. At the time of writing, this consists of the following three packages:
- kernel
- qemu-kvm
- libvirt
Note: Additional packages will be added as necessary, and the repository contents should be used as the primary reference.
Non-Impacted Packages
Any package from RHEL 10’s BaseOS, AppStream, Code-Ready Builder, Extensions, and Supplementary repositories will follow the standard RHEL lifecycle policy for CVE Fixes and release updates. These updates are seamlessly integrated into RHEL for NVIDIA, and as they are resolved in the mainline RHEL release, they will be immediately available to RHEL for NVIDIA customers.
Life Cycle Phases and Support Model
“Forward-Only” Release
RHEL for NVIDIA explicitly does not carry the standard 10-year lifecycle commitment associated with mainline RHEL. Instead, it utilizes a “forward-only” release model with a quarterly release schedule.
Supported Versions: In this fast lane, the only supported version is the newest one.
Bug Fixes and CVEs: RHEL for NVIDIA will only address Critical CVEs (CVSS Score of 8.9 or higher), Major Incidents, or Known Exploited Vulnerabilities (KEVs) for delivery into the currently shipping version (e.g., 26.01, 26.02).
Red Hat will not backport bug fixes or CVE patches to previous releases of RHEL for NVIDIA. To receive new functionality or non-critical security fixes, customers must upgrade to the latest available version (e.g., upgrading from version 26.01 to 26.02).
Partner Validated Support: This edition is offered under a Partner Validated support designation. While Red Hat will take support calls and provide L1/L2 assistance, the support model is a joint effort that relies heavily on NVIDIA for validation and issue resolution. Support is generally defined as “the current version works on the current hardware”.
The “Off-Ramp” to Mainline RHEL
RHEL for NVIDIA is not a permanent fork of RHEL; it serves as a temporary bridge. The packages modified for this edition (specifically the kernel, QEMU, and libvirt) overwrite their standard RHEL counterpart packages via a dedicated repository: Red Hat Enterprise Linux 10 for ARM 64 - Nvidia (RPMs).
As NVIDIA’s hardware-enablement code is reviewed, accepted upstream, and eventually merged into a stable standard RHEL release, customers are expected to transition back to the standard RHEL platform. This “off-ramp” is achieved by deleting the custom NVIDIA repository and performing a standard system update, returning the system to mainline RHEL for long-term support and stability.
Key Differences: RHEL for NVIDIA 26 vs. RHEL 10
While RHEL for NVIDIA is built utilizing RHEL 10 components (such as the RHEL 10.2 kernel for the initial 26.01 release), its lifecycle policies diverge significantly from the standard RHEL 10 lifecycle.
| Feature / Policy | Standard RHEL 10 | RHEL for NVIDIA 26 |
|---|---|---|
| Lifecycle Duration | 10-year standard lifecycle commitment. | Short-term; only the newest version is actively supported. |
| Versioning Scheme | Standard major/minor release format (e.g., 10.1, 10.2, 10.3) | Year/Iteration format (e.g., 26.01, 26.02) to explicitly differentiate it from standard RHEL. |
| Release Cadence | Predictable 3-year major, 6-month minor release cadence. | Highly asynchronous, releasing iteratively (e.g., Q1, Q2, Q3) to align with NVIDIA’s hardware and driver milestones. |
| Updates & Backports | Continuous backporting of bug fixes and CVEs to active and Extended Update Support (EUS) streams. | “Forward-only” model. No backports of CVEs or bugfixes to prior versions; customers must update to the latest release for fixes. Critical CVE fixes will be released only for the currently shipping version of RHEL for NVIDIA. Once the next version of RHEL for NVIDIA is released, no fixes will be applied to the prior version(s). |
| ABI/API Stability | Guaranteed stable ABI/API across the major release. | No backward compatibility guarantees. Upgrades between versions, or migrating back to mainline RHEL, may cause breakage. |
| Code Maturity | Code must be accepted upstream into the Linux kernel before integration. | Leverages CentOS Stream to ingest out-of-tree NVIDIA patches prior to upstream acceptance. |
| Target Use Case | Broad, stable enterprise deployments. | Day 0 hardware enablement, fast-paced Red Hat OpenShift, OpenShift Virtualization, and AI Factory POCs, and cutting-edge exploratory environments. |
