Red Hat has revealed plans to distribute NVIDIA's CUDA toolkit through its repositories for Red Hat Enterprise Linux, Red Hat AI, and OpenShift. The announcement, made at NVIDIA's GTC conference, aims to simplify GPU-accelerated computing for AI workloads. This move includes integration of NVIDIA BlueField DPUs into OpenShift for enhanced security and performance.
On October 28, 2025, during NVIDIA's GTC conference, Red Hat announced a deepened collaboration with NVIDIA to distribute the CUDA toolkit directly via its official repositories. This applies to Red Hat Enterprise Linux (RHEL), Red Hat AI, and OpenShift, allowing users to install CUDA and related NVIDIA software using standard package managers like DNF or YUM, rather than manual downloads from NVIDIA's site.
The initiative addresses customer demands for easier integration of NVIDIA's parallel computing platform, particularly in AI and machine learning applications. It builds on prior efforts, such as a 2019 collaboration to simplify GPU driver deployments, and extends to the full CUDA stack, including libraries for frameworks like TensorFlow and PyTorch. The distribution will support CUDA 13.0 and later versions on RHEL 9 and the recently released RHEL 10.
Red Hat is also integrating support for NVIDIA's BlueField data processing units (DPUs) into OpenShift. This enables offloading of networking, security, and storage tasks from CPUs, optimizing performance in virtualized and containerized environments. Red Hat's chief technology officer, Chris Wright, stated, “This collaboration empowers enterprises to build and deploy AI solutions more efficiently on a trusted platform.”
The announcement aligns with Red Hat's AI push, following the September 2024 launch of RHEL AI, a platform for hybrid cloud generative AI. Additional features include STIG-hardened Universal Base Images (UBI) for NVIDIA GPUs on OpenShift and support for confidential containers with NVIDIA GPUs to secure AI assets. These updates also cover optimizations like vLLM V1 upgrades for hybrid large language models.
This positions Red Hat amid competition from distributions like Ubuntu and SUSE, which have enhanced CUDA support, emphasizing enterprise-grade testing and compatibility with RHEL's security features.