Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
At Computex 2024, Nvidia showcased how Delta Electronics, Foxconn, Pegatron and Wistron were using or adopting parts of its industrial software stack to simulate factories, train robots, inspect products and connect virtual designs with production data. The effort is bigger than a 3D factory model: Nvidia wants its Omniverse, Isaac and Metropolis technologies to help manufacturers design and improve real-world operations. The examples are company-specific and do not establish that all four manufacturers—or all their factories—use the same system.
What Nvidia announced at Computex 2024
The announcement, made at Computex in Taiwan, was a set of factory and robotics workflows rather than a single turnkey factory product or an exclusive agreement among four manufacturers. The technologies have distinct roles:
| Technology | Intended role |
|---|---|
| Omniverse | Connects 3D data and supports rendering, physics, simulation and digital-twin workflows. Nvidia currently describes it as a collection of libraries, APIs, services, blueprints and tools—not a complete factory operating system. |
| Isaac | Supports robotics development, simulation, training, perception and validation. |
| Metropolis | Supports computer vision, camera-based analytics, inspection and related factory intelligence. |
Together, these tools can link factory design and simulation with robot testing and production monitoring. The companies’ announced uses differ; “adopting,” “demonstrating” and deploying a system at production scale are not interchangeable. GamesBeat’s coverage of the Computex announcement provides the reported company examples.
What counts as a factory digital twin?
A 3D model shows what a factory or machine looks like. A simulation adds rules or physics so engineers can test possible layouts, movements or processes. A connected operational twin also incorporates data from the physical facility—such as equipment telemetry, sensors, cameras and production systems—so the virtual representation can inform decisions about the real one. A robot-training environment is a related use: engineers can test robot behavior in simulation before trying it on equipment.
A useful industrial twin may combine CAD and other 3D design data, factory layouts, equipment models, production-process information, IoT and machine telemetry, camera feeds, physics and operational or enterprise-system data. A polished visualization alone does not establish that a twin is accurate, current or connected to live operations. Nvidia’s digital-twin material describes the combination of 3D, operational, IoT and enterprise data involved.
Foxconn: a virtual factory for Guadalajara
Foxconn was the most concrete example in the announcement. The company demonstrated a virtual factory for a new facility in Guadalajara, Mexico, intended to support production of Nvidia Blackwell HGX systems. Nvidia described a workflow that brought Siemens Teamcenter data into Omniverse; Isaac Sim was used to train and validate robots.
In a virtual environment, engineers can work through process definitions, robot positions, sensor locations and tasks such as handling servers or performing inspection movements before making corresponding changes on the physical line. That can expose layout or workflow problems earlier, but the demonstration does not show that Nvidia twins operate Foxconn’s entire global factory network.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesNvidia’s current digital-twin overview also discusses later Foxconn factory work in Houston involving Siemens digital-twin technology built on Omniverse libraries. That is subsequent context, not part of the 2024 Guadalajara example.
How the other manufacturers are using the stack
Delta Electronics: synthetic data for inspection
Delta’s reported workflow uses Isaac Sim and Omniverse/OpenUSD to integrate virtual production lines and generate realistic synthetic visual data. That data can help train computer-vision models for automatic optical inspection and defect detection. The appeal is practical: collecting and labeling images of rare defects on a real line can be costly. But synthetic images do not automatically represent every real-world condition. Models still need validation against actual products, lighting, wear, camera angles and production variation.
Pegatron: camera analytics and operator assistance
Pegatron was described as deploying a Metropolis multi-camera workflow and a factory-twin approach involving Omniverse and Metropolis. The announcement coverage reported more than 21 million square feet of factory space and more than 15 million assemblies per month; those are figures attributed to the announcement, not independently audited current totals.
Pegatron also described using Nvidia NeMo and NIM technologies to let operators interact conversationally with production information. A chat interface that retrieves information or offers recommendations is not the same as autonomous factory control. Responsible use calls for access controls, reliable underlying data, audit logs and human approval for safety-critical actions, with recommendations kept distinct from machine commands.
Wistron: server factories and data-center simulation
Wistron’s reported work includes digital twins of factories making Nvidia DGX and HGX servers, as well as extending Omniverse to simulate data centers used to test assembled HGX systems. The stated uses include testing layouts and processes and incorporating live IoT data from machines.
Wistron’s reported results included bringing a factory online in two and a half months rather than five, improving worker efficiency by more than 50%, and cutting end-to-end cycle time by 50%. These are company-reported figures in coverage of the announcement, not independently verified benchmarks. The available account does not establish the measurement method, baseline, scope across facilities or how much of each change came from simulation rather than other process or equipment changes.
Why Taiwan’s manufacturers matter to Nvidia
The manufacturers highlighted are major electronics producers, and some make AI servers or other systems that incorporate Nvidia technology. That gives Nvidia a useful showcase: companies involved in producing computing hardware can also demonstrate software for designing and operating factories.
The commercial logic is a potential flywheel, not a separately verified financial forecast. Nvidia hardware can run simulation and AI workloads; manufacturers can use its software to improve factory planning, robotics and inspection; more camera, robot and edge-inference deployments may create additional demand for computing and support. The broader ambition is for Nvidia to participate in industrial infrastructure, not just sell components into it.
Recommended Free Tools
The announcement should not be read as proof of a single exclusive industrial program or uniform production deployment across all four companies. It described a mix of public demonstrations, company use cases, integrations and adoption claims.
Rank #4
Where Kenmec and other partners fit
Taiwanese systems integrator Kenmec was identified as an early implementer of Omniverse and Metropolis workflows and a provider of services to manufacturers such as Giant Group. That points to a central reality of industrial twins: software alone does not connect a digital model to a working factory.
Integrators may need to connect CAD and product-lifecycle systems, manufacturing-execution software, enterprise systems, programmable controllers, robot controllers, cameras and sensors; build and maintain the simulation; and support commissioning. Nvidia supplies part of the platform, while manufacturers, automation vendors and integrators supply other essential parts.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What the strategy does—and does not—replace
Nvidia’s stack sits alongside established industrial systems rather than automatically replacing them. Siemens Teamcenter can manage product and engineering data; Rockwell’s Emulate3D and comparable tools support factory simulation and virtual commissioning; robot makers provide robot hardware and controllers; manufacturers retain responsibility for plant operations. Nvidia’s Omniverse and Isaac can contribute interoperability, simulation and robotics workflows, while Metropolis addresses vision and analytics. Operational-data providers and systems integrators can fill still other roles.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →The practical buying question is therefore which layer a factory needs, and how that layer will connect to its existing systems—not whether one vendor supplies every part. Nvidia’s current Omniverse positioning is based on components and workflows that can be integrated into applications, rather than an out-of-the-box factory-management suite.
Best Value
What manufacturers should assess before investing
- Define the problem. Decide whether the priority is layout planning, commissioning, inspection, maintenance, robot training, energy use or operator access to production information.
- Check data readiness. Establish whether CAD, PLM, MES, ERP, PLC, SCADA, camera, sensor and maintenance data are accessible, accurate and compatible.
- Set the required fidelity. A visualization may be enough for communication; testing collisions, robot reach, cycle time or material flow requires a more accurate model.
- Validate sim-to-real performance. Test whether simulated robot and vision results transfer to the physical line, including unusual lighting, wear, vibration and product variation.
- Plan for integration and upkeep. Equipment changes can make a twin stale. Budget for sensor installation, model updates, integration, compute infrastructure, support and retraining—not just software.
- Protect safety and security. Simulation does not replace physical risk assessments or safety-rated controls. Separate AI recommendations from safety-critical machine commands and assess cybersecurity, data location and availability.
- Measure business results. Set baselines for commissioning time, downtime, throughput, defects, scrap, maintenance effort or energy use so a project can be judged on operational outcomes.
- Consider dependence on vendors. Review how much a workflow relies on particular GPUs, APIs, libraries and support contracts, and how data can move between systems.
A greenfield facility may be easier to model coherently because equipment, layout and data systems can be designed together. A brownfield plant can be harder: legacy equipment, undocumented modifications and inconsistent data can undermine the model. A stable production line may not warrant a full twin unless there is a clear quality, maintenance, energy or planning problem to solve.
From factory twins to physical AI
Nvidia’s broader framing now emphasizes physical AI: simulation-ready environments, robotics, synthetic data and validation. The idea is to let robots and perception systems learn or be tested virtually, then evaluate their behavior before deployment. Nvidia said more than 100 companies were adopting Isaac Sim for robotic-application simulation, including Hexagon, Husqvarna Group and MathWorks. That is an ecosystem-adoption claim, not evidence that every named company runs autonomous production at scale.
The 2024 Computex announcement is best understood as an early industrial showcase within that larger direction. Its success depends less on how lifelike a virtual factory looks than on whether its data is reliable, its models reflect real conditions and its use produces measurable operational gains.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
