Aolani & Rafay deploy NVIDIA DSX OS on GB200 NVL72

Aolani and Rafay are working together to deploy NVIDIA DSX OS on NVIDIA GB200 NVL72 infrastructure. They describe the project as among the first of its kind in the industry.

The collaboration focuses on turning AI hardware into a platform customers can use for development and operations, rather than offering GPU capacity alone. It combines Aolani's AI infrastructure with Rafay's software for orchestration, automation, multi-tenancy and lifecycle management.

The move reflects a shift in the AI infrastructure market, where suppliers are under pressure to show they can do more than install advanced systems. Buyers increasingly want environments that can be governed, shared across users and accessed through self-service tools without extensive manual setup.

Under the arrangement, users can provision Kubernetes clusters, virtual machines, AI workspaces and inference environments through a self-service model, according to the companies. The platform also provides central governance, policy enforcement and operational visibility.

Aolani, founded in Singapore, focuses on AI cloud infrastructure in Asia. Rafay sells software for managing modern infrastructure and AI workloads for cloud operators, telecoms providers, enterprises and sovereign AI operators.

Operational focus

The deployment uses NVIDIA DSX OS, a software layer designed to manage AI systems, on NVIDIA GB200 NVL72 infrastructure. The companies presented the project as an example of how operators can move from hardware installation to a working service environment for model development, training and inference.

That operational focus has become more prominent as spending on accelerated computing has increased. In this market, competitive advantage is increasingly tied to how quickly infrastructure owners can bring systems into service, add customers and manage shared use securely.

For operators, building a usable AI platform often requires more than connecting servers and GPUs. They must also assemble software for provisioning, access control, lifecycle management and developer tools, which can slow deployment and complicate governance.

Aolani and Rafay said their approach is intended to reduce that integration burden. Rather than requiring customers or operators to piece together separate operational software, the platform is designed to provide ready access to AI environments on top of the underlying infrastructure.

Nicholas Chia, Chief Executive Officer of Aolani, said speed of deployment and governance are central issues for customers using advanced AI systems.

"Aolani has always been committed to delivering faster time-to-value for our customers. Our customers are at the bleeding edge of AI development, and they need to provision, govern and scale from day one in an industry that moves at lightning speed. Building the next generation of AI cloud means solving for more than just compute capacity, but also production-grade platforms that enable operational readiness from the get-go. That's why we are so excited about this partnership with Rafay," said Chia.

Industry shift

The companies linked the deployment to broader changes in the AI infrastructure sector. As enterprises, cloud providers and state-backed AI projects install larger and more complex GPU systems, attention is shifting to the software layer that determines whether those systems can be run securely and offered commercially.

The change is especially relevant for operators supporting multiple tenants on shared infrastructure. They need tools that let developers access resources quickly while allowing operators to maintain oversight of usage, security policies and system health.

Rafay said its role in the project includes simplifying infrastructure bring-up and automating lifecycle management. It added that the platform gives developers immediate access to production-ready AI environments while preserving governance controls for operators.

Haseeb Budhani, Chief Executive Officer and Co-Founder of Rafay Systems, said the market debate is moving beyond simple hardware deployment.

"AI infrastructure has entered a new phase. The question is no longer how quickly organizations can deploy GPUs. It's how quickly they can transform that infrastructure into a governed, self-service platform that developers can use and operators can manage at scale. We're excited to collaborate with Aolani to help demonstrate what's possible with NVIDIA AI infrastructure and accelerate the path from hardware deployment to production AI services," said Budhani.

The platform is intended to support model development, training, inference and later AI service delivery. The companies positioned the deployment as an example of how infrastructure providers can improve AI system utilisation by making those systems easier to consume and manage.

Aolani was founded in 2023, while Rafay is part of NVIDIA Inception, a programme for start-ups. The companies said the deployment shows how operators can move beyond installing GPU systems and instead run AI platforms that customers can use from the outset.