AICR v1.0: Open, stable, and verifiable GPU cluster configuration

NVIDIA announced the v1.0 release of its AI Cluster Runtime (AICR), a tool that provides version-locked, validated recipes for configuring GPU-accelerated Kubernetes clusters. The release establishes what NVIDIA describes as a stable compatibility contract across the CLI, REST API, Go SDK, bundle layout, and artifact schemas, allowing operators and integrators to build on public interfaces.
AICR addresses the challenge of version compatibility across host kernels, GPU drivers, container runtimes, Kubernetes, networking, storage, device plugins, operators, schedulers, and workload frameworks, each with its own release cycle. According to NVIDIA, a configuration validated for one service, GPU generation, fabric type, machine shape, and Kubernetes release may fail for another.
AICR provides four independent capabilities. Snapshot records observed cluster state, including Kubernetes, operating system, kernel, GPU, and topology information. Recipe describes the desired version-locked component configuration with constraints and validation phases. Bundle renders the recipe into artifacts for Helm, Argo CD, Flux, or Helmfile. Validation compares the recipe with observed state and, where declared, runs deployment, conformance, and performance checks, recording signed evidence of the result.
A live validation dashboard allows operators to find recipes by service, GPU, operating system, workload intent, and optional platform, then inspect each recipe's status and published evidence. Ecosystem integrations include Pulumi Labs, which exposes AICR through an infrastructure-as-code provider, and Mirantis's k0rdent, which packages it for multi-cluster management. NVIDIA states that AICR has over 100 distinct contributors, with almost half from outside NVIDIA.
AICR v1.0 defines compatibility rules for the aicr CLI's public commands, flags, exit semantics, and structured output; the aicrd REST API and OpenAPI contract; the exported API of the github.com/NVIDIA/aicr/pkg/client/v1 package; and generated bundle layout and artifact schemas. Each public interface has a committed baseline checked before changes merge. After v1.0, removing or incompatibly changing a stable public interface requires a new major release. The pkg/client/v1 package exposes the supported workflow without requiring imports from internal packages, and the CLI and REST server use the same facade.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
GPU-accelerated Kubernetes clusters depend on compatible versions across dozens of components, each on its own release cycle: host kernels, GPU drivers,...