Zadara has introduced that its multi-tenant AI cloud platform has aligned with NVIDIA’s Software program Reference Information. This alignment consists of design ideas reminiscent of full-stack workload isolation, community partitioning, per-tenant knowledge quantity separation, VM-based GPU tenancy, and tenant-controlled orchestration.
At NVIDIA GTC 2026 in San Jose, Zadara is presenting this structure, which addresses challenges confronted by NVIDIA Cloud Companions and different organisations when scaling AI. These challenges embody tenant isolation, GPU allocation governance, networking efficiency, and operational overhead. The platform combines a cloud enablement framework with NVIDIA GPU cluster architectures to help multi-tenant environments, automated provisioning, and policy-based operations.
The platform is a part of Zadara’s Impartial AI Manufacturing unit mannequin, a multi-tenant AI infrastructure method that separates GPU funding from service supply. This permits regional cloud suppliers and knowledge centres to enhance GPU utilisation whereas enabling managed service suppliers and enterprises to entry AI providers. The mannequin additionally shifts AI consumption from capital expenditure to operational expenditure.
Zadara additionally outlined high-performance AI networking architectures aligned with NVIDIA Enterprise Reference Architectures, together with help for NVIDIA Spectrum-X Ethernet networking and NVIDIA BlueField DPUs. The platform integrates parts reminiscent of knowledge providers, workload scheduling, and AI safety.
Zadara delivers its multi-tenant AI providers by means of a worldwide associate ecosystem, offering infrastructure capabilities for organisations searching for safe and cost-effective AI cloud environments tailor-made to their necessities.
