NVIDIA DSX Air Launches to Simulate AI Factories and Speed Deployment

NVIDIA has unveiled NVIDIA DSX Air, a revolutionary software-as-a-service (SaaS) platform designed to simulate entire AI factories before any physical hardware is deployed. Announced at GTC 2026 and highlighted in an article by Scott Martin, DSX Air is set to dramatically accelerate the "time to first token" for enterprises building sophisticated AI infrastructure.

Part of the NVIDIA DSX Sim suite within NVIDIA's overarching blueprint for AI factories, the NVIDIA DSX platform, this platform allows for the logical simulation of complex deployments. This includes everything from compute, networking, storage, orchestration, and security, all validated through reference architectures. It provides high-fidelity digital simulations of NVIDIA hardware infrastructure, such as GPUs, SuperNICs, DPUs, and switches, and seamlessly integrates with leading partner solutions via open, API-based connectivity.

Why Simulation is a Game-Changer for AI Deployment

The traditional process of setting up AI factories can be lengthy and fraught with unforeseen issues, often taking months to fully operationalize. NVIDIA DSX Air transforms this paradigm by enabling organizations to simulate and troubleshoot their entire AI factory ecosystem before a single server is unboxed. This shift significantly reduces deployment time from months to mere days or even hours, leading to substantial savings in both time and cost.

This capability is akin to creating a complete digital twin of a hyperscale AI factory, allowing for early identification and resolution of integration challenges. Companies like CoreWeave are already leveraging DSX Air to simulate and validate their intricate AI infrastructure environments, underscoring the platform's immediate impact on accelerating AI deployment at scale.

Real-World Impact and Ecosystem Integration

NVIDIA showcased the power of DSX Air at GTC 2026 with several compelling demonstrations. One highlighted a multi-tenant RTX PRO Server environment, running entirely in simulation, featuring Netris for network orchestration, Rafay for host orchestration, and NVIDIA Run:ai for optimizing GPU allocation. Another demonstrated a video retrieval-augmented generation (RAG) workload on the VAST AI Operating System, illustrating how complex AI applications can be designed and tested virtually.

The platform also proves invaluable for validating security measures. A GTC security demo featured Check Point's distributed firewall running on simulated BlueField DPUs, with TrendAI Vision One detecting threats and Keysight Cyperf generating realistic traffic. This allows partners across the ecosystem – from server manufacturers to security vendors – to validate their offerings alongside NVIDIA infrastructure in a scalable and cost-effective manner.

A New Operational Model for the AI Factory Lifecycle

DSX Air extends its utility beyond initial deployment, introducing a transformative operational model for AI factories. Once the production environment is built and validated in simulation, customers can deploy with high confidence. This pre-validation dramatically increases the probability of a smooth bring-up, allowing teams to focus on running workloads rather than troubleshooting infrastructure.

Furthermore, DSX Air serves as a safe environment for ongoing change management. Long-lived simulations enable customers to test upgrades, rehearse maintenance windows, validate patches, and predict the operational impact of any change before it touches the live production environment. This lifecycle approach ensures maximum uptime and infrastructure availability as AI factories scale.

Read more: NVIDIA DSX Air Announcement. Explore how DSX Air can transform your AI factory deployment today.