Manufacturers scaling physical AI across industrial operations must rethink their edge computing to handle massive data loads.
At Hannover Messe 2026 in Germany, NVIDIA and multiple industrial partners are demonstrating how accelerated computing, AI physics, autonomous agents, and humanoid robotics operate within live factory environments. These hardware and software deployments show that fully-automated production facilities are already under active construction.
As machine learning dictates how products and facilities are constructed and optimised, manufacturers require unified and secure foundations built for high-volume environments.
Deutsche Telekom has built the Industrial AI Cloud in Germany using NVIDIA hardware, establishing a sovereign foundation for European industrial applications. Keeping telemetry and design data within regional borders answers data residency compliance mandates for European manufacturers protecting proprietary blueprints.
Agile Robots, SAP, Siemens, PhysicsX, and Wandelbots use this platform to run workloads ranging from real-time simulation to software-defined robotics. EDAG operates its metys industrial metaverse platform on this cloud to deliver engineering services at scale. Dell Technologies, IBM, Lenovo, and PNY supply the physical hardware necessary to run these systems from local edge hardware deployments up to centralised data centres.
Bridging CAD systems with physics simulation
Industrial hardware systems are growing increasingly complex, requiring advanced software equipped with agentic AI and physics-based models for design and testing
Engineering teams from Cadence, Dassault Systèmes, Siemens, and Synopsys integrate CUDA-X, Omniverse libraries, and Nemotron open models to allow physics-grounded simulation and design exploration. Moving away from geometry-only computer-aided design files to active physical simulations prevents the requirement for expensive physical prototyping.
Full-scale factory digital twins facilitate process simulation, live operations monitoring, and the orchestration of complex robot fleets. Partners across the energy, automotive, and manufacturing sectors use digital twins built on OpenUSD to stress-test their operations.
OpenUSD serves as a common framework, allowing plant engineers to merge diverse proprietary CAD formats into a single spatial environment. This eliminates the data conversion errors that frequently plague multi-vendor operational technology stacks.
ABB integrates Omniverse libraries and Microsoft Azure services into its Genix Industrial IoT and AI Suite, allowing operations teams to monitor asset performance and deploy AI agents for root-cause analysis. Dassault Systèmes applies physical AI libraries to power virtual twin experiences for autonomous, software-defined production systems.
In the energy sector, Kongsberg Digital integrates Omniverse into its Kognitwin platform to map spatial intelligence across physical infrastructure. Combining live operational data with virtual models helps customers optimise performance well before implementing changes on the actual site.
Microsoft uses Fabric Real-Time Intelligence and the Azure Physical AI Toolchain to simulate physical systems accurately and accelerate the deployment of autonomous robots into production. Siemens implements Omniverse into its Digital Twin Composer, merging operational data into a simulation-ready twin that helps engineers identify production faults prior to physical alterations.
Processing video telemetry at the network edge
Standard AI resolves problems under pre-programmed conditions. AI agents introduce adaptive intelligence, analysing the context of a situation before executing physical actions.
Vision AI agents built on Metropolis libraries – along with Nemotron and Cosmos open models – combine multiple data streams using pre-existing camera infrastructure to monitor quality control, operational efficiency, and worker safety.
Invisible AI deploys its Vision Execution System to capture and analyse every production cycle directly on the factory floor in real-time. Built on the Metropolis VSS Blueprint alongside Cosmos Reason 2 and Nemotron models, these autonomous agents deliver actionable insights directly to local operators before assembly issues compound. This intelligence platform is already driving quantifiable gains at some of the largest automotive manufacturing factories worldwide, including Toyota.
Plant directors often struggle to merge new analytics tools with old machinery. Tulip Interfaces developed Factory Playback using the VSS blueprint and Cosmos Reason 2 to synchronise machine telemetry, video feeds, quality events, and operator workflows into a searchable timeline. Terex, an industrial equipment manufacturer operating over 40 plants, utilises this platform to achieve an estimated 3 percent increase in yield and a 10 percent reduction in rework.
Fogsphere expands vision AI compatibility into harsh manufacturing environments by supporting ARM-based edge deployment. The Vision Agent platform allows customers to build and finetune visual AI agents locally. Saipem implements this system to detect and respond to high-risk environmental and safety events. Processing high-definition video data directly on ARM-based edge nodes prevents the network saturation that typically occurs when routing uncompressed continuous video feeds back to a central server.
Breaking hardware constraints for mobile physical AI
AI reasoning allows industrial robots to break free from single-task constraints and time-consuming reprogramming cycles. By navigating unstructured environments and learning new tasks, these machines operate autonomously in live production settings.
At a Siemens blueprint autonomous electronics factory located in Erlangen, Germany, Humanoid’s HMND 01 wheeled robot completed autonomous logistics operations as a first proof-of-concept. This robot runs the Jetson Thor edge AI module for local computation, relying on Isaac Sim and Isaac Lab open frameworks for simulation and reinforcement learning. The simulation-first approach compressed typical hardware development timelines from up to two years down to just seven months.
Connecting autonomous hardware to enterprise networks demands strict adherence to functional safety protocols and isolation from standard IT network traffic. SCHUNK’s GROW automation cell integrates physical AI into standard production lines, using Omniverse and Isaac frameworks to simulate, train, and validate robot behaviour before the cell goes live. Wandelbots and EY connect these simulations to the shop floor to scale operating models across small and medium-sized enterprises throughout Europe.
Hexagon Robotics accelerates robot training and deployment using the Physical AI Data Factory Blueprint and IGX Thor hardware, which provides industrial-grade edge compute combined with functional safety. This implementation enables the AEON humanoid to perform assembly operations at a BMW Plant in Leipzig, marking one of the first humanoid deployments within a German production environment.
Ensuring exact determinism in edge processing prevents hardware collisions, production delays, and heavy machinery accidents. QNX has expanded its collaboration with NVIDIA to power life-safety edge systems across medical and industrial applications. The integration of QNX OS for Safety 8.0 on the IGX Thor module and Halos safety stack locks in the necessary operational boundaries. Processing commands directly on the local hardware guarantees that mechanical reaction speeds remain within exact required safety parameters, avoiding the lag spikes associated with cloud computing.
Bridging the gap between digital simulation and the physical shop floor requires this level of rigorous infrastructure. Advancing humanoid robots and vision agents from isolated pilot phases into active production environments depends entirely on securing high-volume data streams at the edge, ensuring industrial hardware safely matches the pace of AI.
See also: Boston Dynamics Spot uses DeepMind for machinery inspections
Want to learn more about the IoT from industry leaders? Check out IoT Tech Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including AI & Big Data Expo and the Cyber Security Expo. Click here for more information.
IoT News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
