What an AI Factory Operating System Is and Why NVIDIA DSX Matters
An AI factory operating system is a full-stack software and infrastructure framework that coordinates chips, power, cooling, networking, storage, and AI services so organisations can design, simulate, deploy, and operate large-scale AI workloads like a continuous production line for intelligence tokens. NVIDIA DSX is NVIDIA’s answer to this need, turning scattered AI pilots into production-grade AI factories. The NVIDIA DSX platform combines accelerated computing hardware with open, modular software, reference designs, and simulation tools that align energy, infrastructure, models, and applications. It treats AI as essential infrastructure, not a side project, and focuses on improving tokens per watt, lowering token cost, and shortening time to first production. As Jensen Huang notes, “With the DSX platform, you can simulate the entire factory before you spend a dollar, validate performance before a single rack is installed and operate with the kind of reliability that production AI demands.”
Inside the NVIDIA DSX Platform: From MaxLPS to DSX OS
The NVIDIA DSX platform frames AI factories as a five-layer stack—energy, chips, infrastructure, models, and applications—co-designed instead of treated as separate silos. DSX MaxLPS targets power as the primary constraint in industrial AI infrastructure, pairing 45-degrees-Celsius liquid cooling with in-rack performance tuning. NVIDIA states that this combination lets operators run up to 40% more GPUs at their most energy-efficient point within a fixed power budget with minimal impact on inference workloads. On top of this, DSX OS delivers the AI factory operating system layer: open source, modular software components drawn from NVIDIA’s own DGX Cloud operations. By exposing multi-tenant, production-tested services as building blocks, the NVIDIA DSX platform helps infrastructure builders move from months of custom coding to reuse of proven components, improving efficiency, resiliency, and overall AI infrastructure scaling across data centers.
Open, Modular Architecture for Industrial AI and Physical AI Agents
DSX OS is built as open, extensible software that fits into existing industrial AI infrastructure instead of requiring a clean slate. Its modular design lets operators plug capabilities into current platforms, coordinate with building management systems, and connect power-grid behavior to AI scheduling. This approach suits both data center AI factories and physical AI agents such as robots and inspection systems on the factory floor. By providing agentic AI infrastructure software aligned to NVIDIA accelerated computing, DSX OS turns physical AI into a first-class workload alongside language and vision models. That means enterprises can standardize toolchains across simulation, control, and deployment, then tailor them to industry-specific needs. The result is a path from experimental robotics pilots to scaled, reusable skills and toolkits shared across plants, giving industrial teams a common AI factory operating system for both digital and physical intelligence.
Digital Twins with Vertiv SmartRun and Omniverse DSX Blueprint
Simulation is central to making AI factories repeatable and safe to scale. Vertiv’s SmartRun digital twin, integrated into the NVIDIA Omniverse DSX Blueprint, enables planners to design, simulate, and validate converged physical infrastructure as a single system before build-out. Vertiv SmartRun captures dependencies across power, cooling, and controls in a virtual environment, shifting away from document-based processes and siloed handoffs. This model-based approach reduces late-stage design changes and integration risks while improving coordination between teams responsible for high-density AI deployments. As AI factory capacities grow, such digital twins help close the gap between accelerated compute innovation and facility readiness, preserving engineering intent from early configuration through deployment and lifecycle optimization. For enterprises adopting the NVIDIA DSX platform, SmartRun acts as a practical bridge from abstract reference designs to verified, site-specific infrastructure configured for autonomous factory operations.
Factory Operations Blueprint: Toward Autonomous Factory Operations
NVIDIA’s Factory Operations Blueprint, codenamed FOX, extends the AI factory concept from data centers into manufacturing plants. Rather than a single product, FOX is an architectural guide for building autonomous factory operations on top of existing PLC, SCADA, MES, and ERP systems. The blueprint proposes a unified decision layer that ingests live machine signals, vision-based quality data, and operational alerts into central AI models. NVIDIA Metropolis supplies vision AI for automated inspection, while NVIDIA Omniverse maintains a continuous feedback loop between digital simulations and physical processes. This reference design helps move factories from isolated task automation to plant-wide, AI-driven optimisation of workflows in real time. Combined with DSX OS and physical AI agent tools, FOX offers systems integrators and internal teams a reusable pattern for deploying industrial AI infrastructure that can scale across sites and support increasingly autonomous factory operations.
