Nvidia and Linux, a Question

If most AI servers run on Linux and AI models require high-performance GPUs such as NVIDIA, yet NVIDIA GPUs are known to have issues on Linux, how do organizations handle this?