NVIDIA Infra Controller
docs.nvidia.com
Spec sheet
- What it does
- NICo is an open-source suite of microservices for site-local, zero-trust bare-metal lifecycle management.Read 4 Oct 2026 · docs.nvidia.com
- Purpose
- NICo is an open-source suite of microservices for site-local, zero-trust bare-metal lifecycle management aimed at NVIDIA Cloud Partners and infrastructure operators.Read 3 Oct 2026 · docs.nvidia.com
- Intended users
- The documentation identifies NVIDIA Cloud Partner infrastructure operators and infrastructure operators managing bare-metal fleets as users, and also addresses ISV developers and platform engineers integrating through NICo APIs.Read 4 Oct 2026 · docs.nvidia.com
- Integrations
- Documented integrations include Keycloak and compatible IAM solutions, Prometheus metrics, Grafana, Loki, OpenTelemetry-compatible stacks, and Kubernetes.Read 3 Oct 2026 · docs.nvidia.com
- Security
- The site controller's services communicate over mTLS/gRPC, and its operator debug web UI uses HTTPS with OIDC authentication.Read 4 Oct 2026 · docs.nvidia.com
- License
- NICo is open source under the Apache 2.0 license.Read 4 Oct 2026 · docs.nvidia.com
- Headquarters
- Santa Clara, California, USARead 28 Sep 2026 · docs.nvidia.com
- Founded
- 1993Read 28 Sep 2026 · docs.nvidia.com
- APIs and deployment
- NICo exposes REST and gRPC APIs to higher-level systems, and its REST service can run alongside the site controller or centrally in the cloud.Read 4 Oct 2026 · docs.nvidia.com
- APIs and interfaces
- NICo provides REST and gRPC APIs, an administrator CLI, and a debugging UI; its REST API can be deployed alongside the site controller or centrally in the cloud.Read 3 Oct 2026 · docs.nvidia.com
- Attestation limit
- Measured Boot attestation occurs during managed-host ingestion and after a host is released by a tenant; NICo says it does not attest hosts continuously during tenant use.Read 4 Oct 2026 · docs.nvidia.com
- Deployment and support boundary
- NICo commonly runs as microservices on a Kubernetes cluster and can be deployed with Helm charts or Kubernetes Kustomize manifests; operating system patching, upgrades, and runtime configuration are the platform's responsibility.Read 4 Oct 2026 · docs.nvidia.com
- Deployment requirements
- The documented architecture deploys NICo microservices on a Kubernetes cluster co-located in the managed datacenter and specifies at least three nodes for high availability.Read 3 Oct 2026 · docs.nvidia.com
- DPU management
- NICo installs the DPU operating system, provisions HBN, manages DPU firmware, and runs an agent that retrieves desired configuration over gRPC.Read 4 Oct 2026 · docs.nvidia.com
- External dependencies
- NICo requires externally deployed components including PostgreSQL, Vault, Temporal, cert-manager, External Secrets Operator, MetalLB, and the NVIDIA NGC Registry; telemetry and logging components are described as optional but strongly recommended.Read 4 Oct 2026 · docs.nvidia.com
- Hardware compatibility
- The hardware compatibility list says it is not an exhaustive platform certification or qualification, and directs users to hardware manufacturers for that information.Read 3 Oct 2026 · docs.nvidia.com
- Hardware limits
- NVIDIA says its Hardware Compatibility List identifies hardware currently supported from NICo's perspective and is not an exhaustive platform certification or qualification.Read 4 Oct 2026 · docs.nvidia.com
- Hardware onboarding
- NICo discovers hardware through Redfish, validates machine SKUs, supports hardware burn-in and testing, and enforces firmware baselines during ingestion.Read 4 Oct 2026 · docs.nvidia.com
- Hardware validation
- NICo supports Redfish discovery, SKU validation, hardware testing, and enforcement of UEFI and BMC firmware baselines before hosts become available to tenants.Read 3 Oct 2026 · docs.nvidia.com
- Lifecycle management
- It automates hardware discovery, firmware validation, DPU provisioning, network isolation, tenant provisioning, and sanitization.Read 3 Oct 2026 · docs.nvidia.com
- Network isolation
- NICo describes tenant isolation across Ethernet using BlueField HBN, InfiniBand using UFM P_Key assignments, NVLink using NMX-C, and Spectrum-X partitioning.Read 4 Oct 2026 · docs.nvidia.com
- Scope limits
- NICo does not manage host OS patching or runtime configuration, workload scheduling, physical network underlay design, or deployment and management of UFM.Read 3 Oct 2026 · docs.nvidia.com
- Support model
- NICo documentation says operating-system management, including patching, configuration, and image generation, is the user's responsibility.Read 3 Oct 2026 · docs.nvidia.com
Where Specifiction places it
- Bare Metal as a Service (BMaaS) Software#8 of 127.3
- Bare Metal Provisioning Software#24 of 287.2