Zadara has announced that its multi-tenant AI cloud platform has aligned with NVIDIA’s Software Reference Guide. This alignment includes design principles such as full-stack workload isolation, network partitioning, per-tenant data volume separation, VM-based GPU tenancy, and tenant-controlled orchestration.
At NVIDIA GTC 2026 in San Jose, Zadara is presenting this architecture, which addresses challenges faced by NVIDIA Cloud Partners and other organisations when scaling AI. These challenges include tenant isolation, GPU allocation governance, networking performance, and operational overhead. The platform combines a cloud enablement framework with NVIDIA GPU cluster architectures to support multi-tenant environments, automated provisioning, and policy-based operations.
The platform is part of Zadara’s Neutral AI Factory model, a multi-tenant AI infrastructure approach that separates GPU investment from service delivery. This allows regional cloud providers and data centres to improve GPU utilisation while enabling managed service providers and enterprises to access AI services. The model also shifts AI consumption from capital expenditure to operational expenditure.
Zadara also outlined high-performance AI networking architectures aligned with NVIDIA Enterprise Reference Architectures, including support for NVIDIA Spectrum-X Ethernet networking and NVIDIA BlueField DPUs. The platform integrates components such as data services, workload scheduling, and AI security.
Zadara delivers its multi-tenant AI services through a global partner ecosystem, providing infrastructure capabilities for organisations seeking secure and cost-effective AI cloud environments tailored to their requirements.
