NVIDIA Cloud Accelerator (NCX)

Type: Platform Tags: NVIDIA, NCX, cloud accelerator, AI cloud, cloud partners, infrastructure, AI factory Related: NVIDIA-DGX-Cloud, NVIDIA-Cloud-Functions, NVIDIA-Dynamo, KAI-Scheduler, NVIDIA-Grove, NVIDIA-Fleet-Intelligence, NVIDIA-NCX-Infra-Controller, NVIDIA-AI-Cluster-Runtime, NVIDIA-NVSentinel, NVIDIA-DOCA-Platform-Framework, NVIDIA-Project-GPUd, NVIDIA-Mission-Control, NVIDIA-BlueField-DPU, NVIDIA-Spectrum-X, NVIDIA-Cloud-Native-Technologies Sources: https://docs.nvidia.com/ncx/index.html, https://docs.nvidia.com/ncx/ncp-software-reference-guide/latest/index.html Last Updated: 2026-04-29

Summary

NVIDIA Cloud Accelerator (NCX) is a portfolio of open, modular infrastructure software components for cloud partners building NVIDIA-powered AI clouds. It packages NVIDIA AI factory learnings into composable infrastructure and platform components for cloud operators.

Detail

NCX spans hardware lifecycle management, health monitoring, operational automation, infrastructure-as-a-service, container-as-a-service, AI platform services, telemetry, observability, workload isolation, and cloud partner reference architectures.

Representative components include NVIDIA-Cloud-Functions, KAI-Scheduler, NVIDIA-Grove, NVIDIA-Dynamo, NVIDIA-Fleet-Intelligence, NVIDIA-NCX-Infra-Controller, NVIDIA-AI-Cluster-Runtime, NVIDIA-NVSentinel, NVIDIA-DOCA-Platform-Framework, and NVIDIA-Project-GPUd. The docs target NVIDIA Cloud Partners, cloud service providers, and ISVs building AI cloud services on NVIDIA hardware.

Connections

• NVIDIA-DGX-Cloud - NCX supports the partner/operator side of NVIDIA AI cloud infrastructure.
• NVIDIA-Cloud-Functions - unified API layer for scaling inference and simulation workloads.
• NVIDIA-Dynamo - distributed inference-serving framework listed among NCX components.
• KAI-Scheduler - scalable Kubernetes scheduler for GPU resource allocation.
• NVIDIA-Grove - Kubernetes API for defining and scaling multi-component inference workloads.
• NVIDIA-Fleet-Intelligence - managed GPU fleet health monitoring and predictive failure signals.
• NVIDIA-NCX-Infra-Controller - bare-metal lifecycle management and secure infrastructure provisioning.
• NVIDIA-AI-Cluster-Runtime - validated NVIDIA-accelerated Kubernetes runtime recipes.
• NVIDIA-NVSentinel - Kubernetes-native GPU monitoring and fault remediation.
• NVIDIA-DOCA-Platform-Framework - DPU service orchestration for BlueField-accelerated infrastructure.
• NVIDIA-Project-GPUd - lightweight GPU telemetry agent listed in NCX.
• NVIDIA-Mission-Control - AI factory operations and management adjacency.
• NVIDIA-BlueField-DPU - DPU-accelerated infrastructure and DOCA Platform Framework relevance.
• NVIDIA-Spectrum-X - AI networking layer for cloud AI data centers.

Source Excerpts

• NVIDIA describes NCX as open, modular infrastructure software components for building and operating NVIDIA-powered AI clouds.
• The NCX reference guide covers infrastructure, Kubernetes/container services, AI platform services, observability, isolation, and cloud partner operations.