Rafay wins NVIDIA AI Cloud-Ready validation for AI clouds

Rafay Systems has achieved NVIDIA AI Cloud-Ready validation, placing it among an early group of software vendors validated under NVIDIA's standard for AI cloud infrastructure.

The designation means Rafay's platform has met NVIDIA's software requirements for operating production AI cloud environments for neocloud and sovereign AI operators. It covers the software layer needed to deliver API-driven services in multi-tenant settings.

Rafay is framing the milestone against a market in which cloud operators are under pressure to prove they can offer infrastructure as a service for AI workloads, rather than simply rent out GPU capacity. In that environment, software management, governance and tenant isolation have become part of the buying criteria for enterprise and sovereign customers.

Its platform works with NVIDIA Infra Controller, which handles rack-scale provisioning of NVIDIA Grace Blackwell systems, while Rafay provides orchestration, governance and service delivery above that layer. Together, the two provide operators with a validated software and hardware stack from the start of deployment.

Market standard

NVIDIA's AI Cloud-Ready programme sits alongside its Cloud Partner programme, which sets reference designs for hardware infrastructure. The newer validation addresses the software side of AI factory deployments, covering how cloud operators expose and manage services for customers.

The platform supports Kubernetes, virtual machines, SLURM and bare metal as services through published APIs. Rafay also cited hard and soft multi-tenancy, quota enforcement, policy governance, lifecycle management, security controls, compliance functions, self-service workflows and monitoring as part of the validated environment.

Rafay added that the platform supports token-metered access to models hosted as NVIDIA NIM microservices and can be used with NVIDIA NeMo libraries and AI Blueprints. It said this gives operators a way to monetise AI infrastructure from the point it comes online.

"NVIDIA AI Cloud-Ready and its Cloud Partner program, together, offer a standard for how to build and deploy an AI factory from the hardware through the software stack," said Haseeb Budhani, Chief Executive Officer and Co-Founder of Rafay Systems.

"Every GPU provider that has worked hard to meet NVIDIA Cloud Partner requirements will now need to ensure that their offering is operating an NVIDIA AI Cloud-Ready software stack. By partnering with Rafay, neoclouds and sovereign AI cloud providers can deliver validated hardware and software on day 1," Budhani said.

Technical alignment

Rafay described the validation as the result of a multi-year technical relationship with NVIDIA. The platform complies with NVIDIA's Enterprise AI Factory validated design for Blackwell-based enterprise deployments and supports NVIDIA BlueField-3 DPUs and RTX PRO 6000 Blackwell Server Edition.

It is also available through NVIDIA DSX Air for customers seeking to validate deployments from infrastructure to AI services. In addition, Rafay has published a reference architecture for GPU platform services using NVIDIA accelerated computing and NVIDIA AI Enterprise software.

For cloud operators, the practical question is whether a provider can give customers controlled access to AI infrastructure across teams, applications and geographies. Rafay said its software is designed to address that through workload isolation at the hardware and platform levels, along with the operational controls large buyers require before committing capacity.

Warren Barkley of NVIDIA linked the validation to the rapid build-out of AI factories by service providers and specialist cloud operators.

"Neoclouds and service providers are racing to build AI factories capable of supporting the most demanding production-grade workloads," said Warren Barkley, Vice President of Product Management at NVIDIA.

"By becoming NVIDIA AI Cloud-Ready validated, the Rafay Platform - using NVIDIA Infra Controller - provides a turnkey, full-stack recipe for deploying an API-driven AI cloud. This validated software stack helps neoclouds and sovereign AI cloud operators deliver the secure user isolation and built-in guardrails required to scale enterprise AI globally," Barkley said.

Global footprint

Rafay said its platform is already being used in sovereign and neocloud AI deployments across six continents. Named users include Yotta in India, Cassava Technologies in Africa, Firmus Technologies in Australia and TELUS in Canada, with additional deployments in the Middle East, Latin America and Southeast Asia.

Those examples point to a broader shift in the AI infrastructure market, where regional operators are trying to build sovereign or local alternatives to larger global cloud platforms. In these projects, software control layers that manage tenancy, governance and billing are becoming as important as access to advanced chips.

Rafay's announcement also suggests that software validation may become a competitive checkpoint for operators pursuing large enterprise or public-sector AI contracts. As Barkley put it, the validated stack is intended to help providers deliver "the secure user isolation and built-in guardrails required to scale enterprise AI globally".