What the NVIDIA DSX Platform Is and Why It Matters
The NVIDIA DSX platform is a full-stack operating system for AI factory operations that unifies chips, data center infrastructure, software, simulation, and partner technologies so organizations can design, test, and run large-scale AI infrastructure more efficiently and at lower risk. Positioned as a common framework for infrastructure builders, NVIDIA DSX brings together open source, modular software libraries, APIs, reference designs, NVIDIA accelerated computing platforms, and third-party systems into one coordinated architecture. Rather than treating energy, servers, networking, storage, and facilities as separate projects, DSX aligns the five-layer stack—energy, chips, infrastructure, models, and applications—around AI factory workloads. NVIDIA describes this as giving builders a “complete playbook to build AI factories,” with the goal of reducing token cost, shortening time to first production, and improving reliability as AI factories scale and serve multi-tenant, mission-critical AI services.
Inside DSX OS: Modular Software for AI Factory Operations
At the heart of the NVIDIA DSX platform is DSX OS, a collection of open, modular software components designed to operate AI factories at scale. DSX OS packages the software NVIDIA uses to run its own DGX Cloud infrastructure and releases it as open source so ecosystem partners do not need to rebuild core capabilities from scratch. According to NVIDIA, this can save months of custom development and speed time to revenue for providers of AI services. The software targets multi-tenant environments, linking data center controls, AI platforms, and facility systems into a coordinated control plane that aims to improve tokens per watt and lower token cost. DSX OS also supports continuous operations by addressing hardware faults and workload scheduling in a unified way, strengthening reliability and resiliency as clusters grow in size and complexity.
Designing for Efficiency: MaxLPS, Power, and Infrastructure Simulation
NVIDIA DSX also tackles the physical and energy side of AI factory operations. DSX MaxLPS is a suite of technologies designed to maximize token performance per megawatt within a fixed power budget, combining 45-degrees-Celsius liquid cooling with in-rack optimizations so operators can run up to 40% more GPUs at their most energy-efficient operating point with minimal impact on workload performance. Beyond hardware tuning, the DSX platform treats power and grid behavior as part of the core software stack instead of a separate facility concern. This approach supports infrastructure simulation before build-out: chips, racks, cooling, and controls can be modeled as a single system, helping reduce late design changes and improving the match between theoretical capacity and real-world AI demand. For organizations, it means higher AI output from the same power envelope and fewer surprises at deployment.
Digital Twins and Vertiv SmartRun: Validating AI Factories Before Build
Infrastructure simulation is central to the NVIDIA DSX platform, and the integration of Vertiv SmartRun highlights how digital twins change AI factory design. Vertiv SmartRun, an overhead converged physical infrastructure system, is integrated as a configurable digital twin within NVIDIA Omniverse DSX Blueprint workflows. Data center teams can design, simulate, and validate power, cooling, and controls as one system before any physical build, replacing document-heavy processes and siloed handoffs. By capturing configurations and dependencies in a virtual environment, the SmartRun digital twin helps reduce integration risk, accelerate time from planning to operational readiness, and preserve engineering intent across the lifecycle. Vertiv describes this as the first phase of its AI factory digital twin roadmap, aimed at closing the gap between rapid compute innovation and the slower pace of physical infrastructure readiness.
From AI Factory Operations to Autonomous Factory Systems
Beyond data centers, NVIDIA is linking the NVIDIA DSX platform to autonomous factory systems on the manufacturing floor. The Factory Operations Blueprint, codenamed FOX, is a reference design that unifies data from PLCs, SCADA, MES, and ERP into a single decision-making layer. Today, these systems rarely integrate cleanly, which blocks plant-wide intelligence and slows root cause analysis and quality control. FOX outlines how to ingest live machine signals and quality data into central AI models, creating a feedback loop between digital simulation and physical operations. Powered by NVIDIA’s vision AI framework Metropolis and other DSX-aligned components, the blueprint moves factories from isolated automation to connected, AI-managed workflows. When combined with infrastructure simulation and digital twins, FOX shows how organizations can first test autonomous factory systems virtually, then deploy them with higher confidence and clearer operational economics.
