Foxconn is building an AI factory that turns industrial data into intelligence to power a self-learning manufacturing ecosystem. Using NVIDIA infrastructure and software, the company trains AI models and deploys specialized agents that support real-time decisions across manufacturing, electric vehicles, and smart cities.

At the center is FoxBrain, Foxconn’s proprietary large language model, developed by adapting open foundation models with industrial data. FoxBrain supports specialized platforms, including Genesis for manufacturing and CityGPT for urban services.

NVIDIA DGX SuperPOD serves as the AI factory’s training foundation. Foxconn reports 2.48x faster training for 120-billion-parameter models compared with its previous H100-based systems, processing one trillion tokens in 12 days. NVIDIA software helps manage infrastructure and deploy AI agents across more than 200 facilities.

The company reports practical manufacturing improvements:

• 25% fewer labor hours spent on production scheduling.
• 50% less engineering workload for injection-molding process changeovers.
• More than 20 major defect root causes addressed, with a correction rate exceeding 90%, on surface-mount production lines.

Foxconn also uses NVIDIA RTX PRO servers, Omniverse, and Isaac to simulate factories and robots before deployment, helping replicate production lines globally. Through its subsidiary Visionbay.ai, an NVIDIA Cloud Partner, Foxconn is extending its AI factory leveraging NVIDIA-Certified Systems to deliver AI-as-a-Service to enterprises and governments across Asia.

Discover how Foxconn and NVIDIA are architecting the future of industrial intelligence. Read the full Foxconn DGX SuperPOD case study.