Zadara aligns AI cloud platform with NVIDIA security guide

Zadara has aligned its multi-tenant AI cloud platform with NVIDIA's Software Reference Guide, which sets out design patterns for secure, sovereign multi-tenant AI cloud deployments.

The alignment covers architectural principles for sharing AI infrastructure across multiple organisations while keeping workloads and data separated. Zadara plans to present its approach at NVIDIA GTC in San Jose, with its team at booth 7034.

Reference guide

NVIDIA's Software Reference Guide outlines design expectations for multi-tenant AI environments. Zadara says its platform implements elements of the guide, including full-stack workload isolation, network partitioning, per-tenant data volume separation, VM-based GPU tenancy, and tenant-controlled orchestration patterns.

Multi-tenancy has become a key requirement for AI infrastructure operators as enterprises seek access to accelerated compute without buying and operating dedicated systems. The model also appeals to service providers aiming to increase utilisation of costly GPU infrastructure.

Zadara positions its platform for cloud providers, telecoms operators, and managed service providers deploying AI services across distributed locations. It describes the offering as "sovereign-by-design", meaning customers and operators can keep data and workloads under specific jurisdictional and operational controls.

Isolation and control

Zadara says the alignment reflects a focus on isolation and governance across several layers of the stack. Full-stack workload isolation and network partitioning are intended to let tenants share compute and network resources while keeping performance predictable and boundaries clear.

Per-tenant data volume separation addresses storage segmentation. VM-based GPU tenancy allocates GPU resources through virtual machines rather than exposing bare-metal accelerators directly. Tenant-controlled orchestration patterns describe how each tenant manages scheduling and workload lifecycle.

As AI clusters scale, operators often face challenges around GPU allocation governance, networking performance, and operational overhead. Zadara argues these issues become more complex in shared environments where multiple customers run training and inference workloads side by side.

Zadara combines its cloud platform with NVIDIA GPU cluster architectures to support secure multi-tenancy, automated provisioning, and policy-driven operations across distributed environments. The platform is aimed at providers operating across multiple sites and seeking consistent operations between them.

Networking focus

At GTC, Zadara plans to highlight networking architectures aligned with NVIDIA Enterprise Reference Architectures, including support for NVIDIA Spectrum-X Ethernet networking and NVIDIA BlueField DPUs.

It also says the platform integrates with "mission-critical elements" of AI operations, including data services, workload scheduling, and AI security. Zadara did not provide performance benchmarks or customer deployment figures tied specifically to the reference-guide alignment.

Neutral AI factory

Zadara presents the alignment as part of its "Neutral AI Factory" model, which it describes as separating investment in GPU infrastructure from the delivery of AI services by service providers and enterprises.

In Zadara's framing, regional cloud providers, telecoms operators, and data centre operators can run shared AI infrastructure and increase utilisation, while managed service providers and enterprises consume AI services that meet sovereignty requirements.

"NVIDIA's Software Reference Guide sets a high bar for performance, isolation, and operational rigor in multi-tenant AI environments," said Yoram Novick, chief executive officer at Zadara.

"Our platform has been tested against these requirements, enabling NCPs and cloud providers to deploy multi-tenant AI infrastructure with confidence, predictable performance, and sovereign control. At the same time, our Neutral AI Factory approach ensures these deployments are economically sustainable and operationally scalable."

Partner views

Zadara says it brings its multi-tenant AI services to market through partners. US north-east operator Micro Support Group says the partnership supports customer requirements for resilience and sovereignty.

"Our partnership with Zadara empowers us to deliver new levels of performance, resilience, and sovereignty for our customers across the US northeast to meet the requirements of their AI workloads," said Dayna Morton, chief executive officer at Micro Support Group.

"Enterprises today need more than traditional IT. They need secure, cost-effective multi-tenant AI clouds tailored to their businesses," Morton said.

Africa-focused provider BCN says it has deployed what it describes as the region's first Neutral AI Factory using Zadara's approach.

"Our partnership with Zadara has enabled us to deliver Africa's first Neutral AI Factory," said Ibrahim Dikko, chief executive officer at BCN.

"Cloud providers across Africa need a trusted partner to supply the AI cloud infrastructure that powers their services while maintaining full control over the infrastructure, strong tenant isolation and unwavering data sovereignty," Dikko said.

Zadara says it operates more than 500 edge clouds globally and can deploy its platform on premises, at the edge, and in hybrid and multi-cloud environments.