NVIDIA Opens FOX Early Access for European Manufacturers; Siemens and Mercedes-Benz Put It to Use

On Oct. 21, 2026, NVIDIA announced examples of European manufacturing and design companies using CUDA-X, Omniverse, Cosmos and Nemotron across their operations. The centerpiece is the Factory Operations Blueprint (FOX), now available for early access on GitHub. It enables the development of always-on management agents that coordinate work across a factory, as well as specialized agents for visual inspection, root-cause analysis, report generation and fleet management. Siemens says it has integrated FOX-based agents into Intelligence Center X and expects an estimated 20% cost reduction.

NVIDIA’s announcement is significant because it shows, through examples from Siemens and Mercedes-Benz, a move toward handling factory operations and design validation on the same computing and simulation platform. Siemens in particular says it expects a 20% cost reduction and has cut calibration time for inspection processes by 40% at its Erlangen plant. This shifts the question from whether to use AI to how much existing process data can be connected.

The focus this time is NVIDIA’s emphasis on a reference design for a set of agents targeting factory operations, rather than simply providing computing infrastructure. FOX envisions management agents that work across operational tasks such as inspection, root-cause analysis, report generation and fleet management. It connects existing platforms such as CUDA-X and Omniverse to factories’ decision-making layers. This is novel, while also extending existing efforts to bring simulation, PLM and IT/OT systems together. Our previous coverage, AWS unveils open-source tool suite for robot development and NEURA Robotics introduces 4NE1 Gen 3.5, pointed to a trend in which software platforms and robot hardware are taking center stage in robot development. This announcement, by contrast, clearly shows a framework that brings factory operations, design simulation and digital twins together on the same computing platform, rather than focusing on development tools. In other words, the target is not an individual robot but the implementation environment spanning design, validation, manufacturing and maintenance. The implications for the industry include whether companies can handle factory data, CAD/CAE, PLM, OT logs and 3D models across systems. Siemens’ reported 20% cost reduction and 40% time reduction at its Erlangen plant show that data integration can directly affect operational metrics, but those results may not transfer directly to other sites or processes. The examples from Mercedes-Benz, Audi, Schaeffler and Volkswagen Sachsen cover distinct areas — aerodynamics, battery production, actuators and assembly processes — pointing to use-case-specific optimization on a common platform. Open questions include how reproducible FOX’s early-access release will be across different factory environments, whether each company’s deployment is at the trial or operational stage, and what data connections and safety requirements will be needed.

Siemens’ Process Preparation X uses NVIDIA Cosmos Data Factory tools and skills to generate defect images for circuit-board inspection. The company says this approach cut calibration time for inspection processes by 40% at its electronics factory in Erlangen, Germany. On the design side, Mercedes-Benz runs Siemens’ Simcenter STAR-CCM+ on NVIDIA GPUs to accelerate aerodynamic simulations. Audi is using a digital twin of battery production built with NVIDIA Omniverse to validate production concepts before physical equipment is available. Schaeffler has integrated digital twins of humanoid actuators into the NVIDIA Isaac simulation framework, while Wandelbots is using NOVA and Isaac Sim to digitally validate assembly processes at Volkswagen Sachsen’s Transparent Factory.