Demonstrations at NVIDIA GTC will showcase end-to-end automation from bare metal to production-ready AI cloud services.
San Jose / Madrid, March 17, 2026 — OpenNebula Systems has expanded its integration with NVIDIA technologies, enabling organisations to deploy secure, production-ready AI Factories at scale. By combining OpenNebula's cloud management and virtualisation platform with NVIDIA accelerated computing, networking, and bare-metal lifecycle management solutions, enterprises, HPC centres, and neocloud providers can deliver high-performance AI infrastructure with full tenant isolation and automated lifecycle control.
The integrations cover GPU virtualisation, network offload, and automated bare-metal provisioning. OpenNebula supports NVIDIA GB200 NVL4 GPUs via PCI passthrough, providing virtual machines with direct hardware access and preserving native performance. Multi-Instance GPU (MIG) support enables hardware-level partitioning, allowing multiple tenants to share accelerators securely while maintaining predictable resource allocation.
NVIDIA BlueField DPUs are fully managed within OpenNebula's control plane, offloading network and security tasks from host CPUs. Per-tenant traffic enforcement, containerised network functions, and programmable switching policies reduce CPU load, strengthen workload isolation, and improve infrastructure efficiency across multi-tenant environments.
Bare-metal provisioning is automated through OpenNebula's integration with NVIDIA Base Command Manager, allowing operators to manage GPU-accelerated infrastructure through a single interface. Automated workflows handle everything from hardware discovery to OS deployment and driver configuration, reducing manual intervention and accelerating time to production.
From Bare Metal to Production-Ready AI Clouds
At NVIDIA GTC, OpenNebula Systems will demonstrate the full deployment path — from bare-metal provisioning to running AI workloads inside isolated virtual AI Factories. The demo will show how operators can start with raw hardware, automatically provision servers with the correct firmware, drivers, and OS, then deploy GPU-enabled virtual machines for AI training and inference — all managed through a single pane of glass.
At OpenNebula Systems, we believe the future of AI infrastructure is open, multi-tenant, and fully automated. Our expanded integrations with NVIDIA technologies allow organisations to build AI Factories that are not only powerful but also sovereign, secure, and operationally efficient. Whether you are an enterprise building private AI clouds, an HPC centre serving research communities, or a neocloud provider offering GPU-as-a-service, our platform delivers the automation and isolation needed to scale with confidence.
Constantino Vázquez Blanco, CTO at OpenNebula Systems
Key Integration Capabilities
- NVIDIA GB200 NVL4 GPU Support via PCI Passthrough: Full hardware access for AI workloads inside virtual machines, preserving native GPU performance.
- Multi-Instance GPU (MIG): Hardware-level GPU partitioning for secure, multi-tenant sharing with predictable resource allocation.
- NVIDIA BlueField DPU Management: Network and security offload managed within OpenNebula's control plane, enabling per-tenant traffic enforcement and containerised network functions.
- Automated Bare-Metal Provisioning with NVIDIA Base Command Manager: End-to-end server lifecycle management from hardware discovery through OS deployment and driver configuration.
- Unified Management Interface: Single pane of glass for managing virtualised and bare-metal GPU infrastructure across multi-tenant environments.
For more information, visit the OpenNebula booth at NVIDIA GTC or contact the OpenNebula Systems team.
