Foxconn (Hon Hai Technology Group) · Electronics manufacturing · Taiwan, Mexico and United States
Foxconn: digital twins for new server plants with Omniverse, Isaac and Metropolis
Foxconn uses NVIDIA Omniverse digital twins to plan production lines, Isaac to simulate robots and Metropolis for camera-based monitoring, from Hsinchu to new server plants in Mexico and the US. Most published results are expectations, such as a forecast energy cut of over 30 percent in Mexico.
- Status
- In production
- Status as of
- 26 Aug 2025
- NVIDIA technologies
- NVIDIA Omniverse, NVIDIA Isaac, NVIDIA Metropolis, NVIDIA Jetson, NVIDIA cuOpt, NVIDIA NeMo
The challenge
Foxconn had to bring up new plants for NVIDIA Blackwell server systems quickly in several countries while keeping lines consistent between sites. Each layout change on a physical line costs time, and robot cells, sensors and cameras need to be placed and tested before production starts.
Foxconn also wanted to copy proven lines between sites (NVIDIA's customer story uses Taiwan to Mexico as the example) instead of designing each one from scratch.12
What was implemented
Digital twin of production lines
Foxconn says its work with NVIDIA began with a 3D digital twin used to plan and simulate automated production lines at its Hsinchu factory in Taiwan, shown at Hon Hai Tech Day 2023, and that Omniverse use will be scaled out to its factories worldwide. Its November 2024 release names NVIDIA Omniverse, NVIDIA Isaac and NVIDIA Modulus (now called PhysicsNeMo), plus OpenUSD for data exchange.
New Blackwell server plants
NVIDIA reported in 2024 that Foxconn was building a digital twin of a new plant in Guadalajara, Mexico, alongside Siemens Xcelerator software including Teamcenter, to define processes and train robots before the physical plant ran. NVIDIA names Isaac Sim and Isaac Manipulator for robot simulation, Metropolis for camera placement and vision AI, cuOpt for route planning and Jetson as the computer on autonomous mobile robots.
Fii Omniverse Digital Twin platform
NVIDIA's customer story describes Foxconn's Fii Omniverse Digital Twin platform for simulation and real-time monitoring of factories, PhysicsNeMo models for thermal simulation, video analytics agents built on Metropolis, and a Foxconn language model (FoxBrain) trained with NeMo. It also describes a Houston, Texas facility that Foxconn is building, with its systems validated virtually before construction; the page does not state whether that site is operating.
2025: RTX PRO Servers
In August 2025 Foxconn said it is adopting NVIDIA RTX PRO Servers and deploying several uses on them, including Omniverse digital twins of production lines and Isaac for autonomous mobile robot development.123456
Reported outcomes
Mexico plant expected to cut electricity use by more than 30 percent a year14
More than 30% lower kWh per year (expected)
Projected, not measuredFactory planning and time to revenue could be cut by about half2
About 50% shorter (belief stated by a Foxconn manager)
Projected, not measuredNVIDIA states PhysicsNeMo models run thermal CFD simulations up to 150 times faster than traditional methods2
Up to 150x faster, minutes instead of hours (vendor figure)
Projected, not measuredFoxBrain language model trained with NeMo in four weeks2
4 weeks training time
Measured
Each outcome links to its source; the label there says whether the company, NVIDIA or a third party reported it.
What others can learn, and the limits
What others can learn: Foxconn treats the digital twin as the place where a proven line is copied to a new site, which pays off most when the same line is built many times. Robot cells, camera placement and traffic routes can be tested in the same scene before hardware arrives. Limits: Foxconn is a very large contract manufacturer with NVIDIA as a customer and partner, so its access to engineering support is unusual. Almost every published number is a forecast or an NVIDIA figure without a stated baseline; no measured energy result for the Mexico plant was found. Plants that do not repeat lines across sites will see less of the copy-and-replicate benefit.
Explore a similar project
Use the Digital Twin Opportunity Lab with your own assumptions. Results are independent of this case.
Sources
- Foxconn Expands Blackwell Testing and Production With New Factories in U.S., Mexico and Taiwan (opens in a new tab)
- Foxconn develops physical AI-enabled smart factories with digital twins (customer story) (opens in a new tab)
- Foxconn to Build AI Factories with NVIDIA Omniverse Platform (opens in a new tab)
- Foxconn Trains Robots, Streamlines Assembly With NVIDIA AI and Omniverse (opens in a new tab)
- Foxconn to Transform its Manufacturing and Business Operations with NVIDIA RTX PRO Servers (opens in a new tab)
- nvidia-physicsnemo (PyPI project page, maintained by the NVIDIA PhysicsNeMo team) (opens in a new tab)
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.