Foxconn tests AI robots using virtual factory simulations

In short: Foxconn is testing AI-powered robots and using virtual copies of its factories to train them before putting them on real production lines.

What's going on

Foxconn, a major electronics manufacturer, is developing and trialing AI-enabled industrial robots and humanoid robots. A humanoid robot is shaped more like a person, with arms and hands that can handle a wider mix of tasks.

A key part of Foxconn’s approach is training robots inside “digital twins” of factories. A digital twin is a detailed virtual copy of a real place, like a flight simulator but for a factory floor. Foxconn uses NVIDIA Omniverse to simulate factory layouts, robot movements, and production steps, so teams can test changes without stopping real work.

Foxconn has also discussed deploying humanoid robots in a next-generation plant that builds AI servers, including a facility in Texas. The company has pointed to NVIDIA Isaac GR00T models, which are AI models (software that helps robots learn patterns) designed to help robots understand what they see and decide what to do next. Trials are focused on practical details like where robots fit, how fast they can work, and which jobs they can do safely.

What to watch

The big question is how quickly Foxconn can move from trials to routine use across its factories. Watch for updates on whether humanoid robots actually take on regular assembly tasks, and whether simulation training reduces the need for slow, expensive testing in real factories.

Source: Financial Times