Bridge Is Now Live on NVIDIA's DSX Air Demo Marketplace
Bridge from Armada is now available on the NVIDIA DSX Air Demo Marketplace. Anyone with a DSX Air account can launch a complete Bridge GPUaaS environment in simulation and run it end to end, from Infrastructure discovery to tenant workloads, without waiting on a single server. To simplify the experience, AI-as-a-Service features have been disabled in this environment.
Bridge is Armada's GPU cloud management software. Neoclouds, sovereign clouds, telcos, and enterprises use it to turn GPU clusters into a multi-tenant GPU resource with cloud services (IaaS, PaaS, and AIaaS) layered on top. Bridge also includes user management, RBAC, and billing.
From announcement to hands-on
In March, Armada was part of the inaugural DSX Air announcement at NVIDIA GTC 2026. We had used NVIDIA network simulation since late 2024 for Bridge development and customer PoCs. The plan was to extend that into full AI factory simulation so customers could "try before they rack."
The Marketplace listing delivers on that plan. The Armada Bridge GPUaaS demo is a self-serve, guided lab that any DSX Air user can launch on demand.
What's in the lab
The simulation models a GPUaaS fabric: dual-homed leaf-spine NVIDIA Cumulus switching with separate east-west and north-south networks, eight compute nodes, a storage node, and a controller with Bridge preinstalled. Users follow the same path a GPU cloud operator does:
- Discover: Bridge scans the out-of-band network, inventories compute, switches and storage, and configures the underlay automatically.
- Onboard: As Super Admin, create a tenant, set quotas and per-hour pricing, and add a Tenant Admin.
- Isolate: As Tenant Admin, build compute and converged VPCs with dedicated subnets, so no tenant can reach another's resources.
- Consume: Allocate BareMetal, mount storage, stand up a Kubernetes cluster, and deploy applications from the catalog, with node and cluster observability built in.
One note: the lab runs on CPU BareMetal, so AI-as-a-Service features such as JupyterHub, model deployment, inferencing, RAG, MLOps, model optimization, model benchmarking, and fine-tuning aren't exercised here. They build on the same workflow once GPU BareMetal is in place.
Why it matters
GPU capacity can be scarce and valuable, so proofs of concept often stall while teams wait for racks. In DSX Air, a prospective operator can evaluate Bridge's multi-tenancy, tenant isolation, pricing and IaaS + PaaS workflows in a 1.5–2-hour lab. They move to a hardware PoC already knowing how the platform behaves.
More from Armada and NVIDIA in 2026
The Marketplace listing caps months of close work with NVIDIA:
- NVIDIA DSX Air (March 16): How DSX Air reduces dev/test costs, accelerates PoCs, and lowers production risk for Armada and its customers.
- NVIDIA AI Grid (March 17): Armada and NVIDIA are operationalizing distributed AI by enabling AI Grid on the Armada Edge Platform for telcos, service providers, and enterprises.
- NVIDIA Infra Controller (July 31): Bridge integrates the open-source NVIDIA Infrastructure Controller (NICo) project, formerly Carbide, to take GPU servers from freshly racked to tenant-ready.
Try it
Log in to NVIDIA DSX Air, open the Demo Marketplace, and search for Armada. The Armada Bridge GPUaaS card will open the guided lab. Ready to take it to real GPUs? Schedule a demo with the Armada team.