Buying NVIDIA H100 or Blackwell is now the easy part. The real nightmare begins after delivery: assembling the orchestration layers, configuring the network, sequencing operator dependencies and validating the infrastructure. This process, which generally takes weeks for experienced engineers, has just found its antidote. Mirantis announces the integration of NVIDIA Run:ai into its k0rdent AI platform, promising to deploy a production-ready “AI Factory” in just minutes.
This announcement is a major step towards the solution k0rdent, this open source environment that we follow closely. The idea is to remove the “technical plumbing” that separates the bare-metal server from the first training or inference job.
Put an end to manual assembly of AI bricks
The challenge of this collaboration is to bridge the gap between hardware provisioning and operationalization. As a certified member of the “NVIDIA AI Cloud Ready” initiative, Mirantis now automates all critical sequencing: from managing external DNS, to installing NVIDIA GPU and network operators, to deploying the Run:ai planning layer. For businesses, this means standardized, repeatable deployment, eliminating dependence on the “hidden knowledge” of a few in-house experts.
For data scientists, the change is radical. They can now submit their workloads or launch notebooks interactive without having to worry about Kubernetes clusters or the underlying GPU configuration. The k0rdent AI platform manages updates and configuration drift in the background, ensuring that infrastructure remains aligned with production requirements.
A formidable weapon for the sovereign cloud
One of the strengths of this integration lies in its ability to operate in “air-gapped” environments, that is to say completely isolated from the Internet. This is a strong argument for regulated sectors and digital sovereignty projects that cannot tolerate any external network dependency for their private AI Factories.
Additionally, the solution natively supports the latest rack-wide architectures, such as NVIDIA Grace Blackwell NVL72 systems, via zero-touch server lifecycle automation. By radically simplifying access to computing resources, Mirantis and NVIDIA enable specialized cloud providers (neoclouds) to provision AI environments on demand, thus maximizing the profitability of hardware investments which amount to millions of euros.
