Aviz ONES Is Positioned for Operational Use in NVIDIA DSX Air

At GTC 2026, NVIDIA introduced DSX Air as part of its DSX platform and paired it with a Vera Rubin DSX Artificial Intelligence (AI) Factory reference design and an Omniverse DSX Blueprint. The vendor brief frames Aviz ONES as an operational layer for repeatable, multi-tenant AI factory networking aligned to those architectures, with validation and lifecycle workflows.

Research Overview

The post connects NVIDIA’s AI factory messaging at GTC 2026 with Aviz’s product positioning around DSX Air. It argues that AI factories require full-stack coverage spanning compute, networking, storage, orchestration, security, and operations.

Within that framing, the update centers on how Aviz ONES integrates simulation and operational readiness for AI factory networking rather than treating networking as a separate component. The post also points to field sessions and a joint fireside chat on the Aviz–NVIDIA partnership and Day 0/1/2 operations.

Key Findings

NVIDIA’s presentation is described as an ecosystem approach to AI factories, with DSX Air intended to support ecosystem partner software images for integration and day-one interoperability. The blog positions Aviz as part of the deployable layer that helps teams validate, deploy, share, and run AI factory networking at scale.

Aviz ONES is presented as enabling a shift-left workflow that moves integration and troubleshooting into simulation. The post states that instead of discovering integration problems late, teams can design, simulate end to end in DSX Air, and then move into production with repeatable configuration and validated workflows.

Technical Breakdown

The technical snapshot describes Aviz ONES in NVIDIA DSX Air as a deployment model for designing, simulating, and deploying multi-tenant AI factory networking aligned to NVIDIA reference architectures. It lists coverage that includes Spectrum-X, BlueField, InfiniBand, and NVLink, and extends validation across compute, storage, orchestration, servers, NICs, and networking.

For lifecycle coverage, the post maps Day 0 to design and validation, Day 1 to tenant onboarding and fabric provisioning, and Day 2 to telemetry, health, troubleshooting, upgrades, and controlled change management. It also describes observability that tracks 250+ network, server, and Graphics Processing Unit (GPU) metrics and provides end-to-end monitoring across switches, NICs, and GPUs, including RDMA over Converged Ethernet (RoCE) telemetry for Power Factor Correction (PFC), ECN, and queue counters.

The architecture is described as agentless and built on containerized microservices using YAML templates plus validate/apply/verify workflows. The brief further states that the solution includes alerting and integrates with customer tools.

Operational Impact

The post frames the operational outcome around shifting integration and troubleshooting into simulation. It states that NVIDIA describes DSX Air as a way to reduce time to first token by cutting timelines from weeks or months to days or hours.

It quotes Amit Katz, VP of Networking at NVIDIA: “Aviz ONES adds the operational layer that helps customers turn AI factory networking into a repeatable, multi-tenant platform.” It also ties the approach to NVIDIA’s Enterprise AI Factory validated design, which the blog describes as full-stack architecture built with Blackwell accelerated computing, BlueField DPUs, Spectrum-X Ethernet, NVIDIA AI Enterprise software, and a partner ecosystem.

The conclusion emphasizes that the post’s central message is about closing the gap between AI factory architecture and operations using operational playbooks and simulation-driven validation. It states this “Blog Signals brief” is a fact-based summary of the vendor blog.