NVIDIA has introduced a new platform aimed at the era of the 'Agentic Factory,' where artificial intelligence (AI) autonomously manages factory operations.
On June 4, NVIDIA unveiled the 'Factory Operation Blueprint' (FOX), a reference design for building autonomous factory management agents, at the GTC Taipei 2026 event.
Manufacturing facilities are evolving beyond individual equipment automation to connect the entire factory into a single intelligent system. There is a rapidly growing demand for AI platforms that can analyze vast amounts of data generated from production equipment, sensors, quality control systems, work instructions, and operational alerts to make informed decisions.
FOX is designed to meet this demand by monitoring and inferring various data in real-time while coordinating specialized AI agents responsible for quality control, material transport, and worker safety, ultimately enhancing productivity and operational efficiency.
Built on NVIDIA's NemoClaw, AI-Q Blueprint, and Nemotron open models, FOX supports the entire process from factory system integration to AI model development and operational automation. Developers can use it to create a central agent that manages the entire factory.
FOX is optimized to run on NVIDIA's AI supercomputer, the DGX Station, which features the GB300 Grace Blackwell Ultra Superchip. This chip offers 20 petaflops (FP4) performance and 748GB of unified memory, capable of running large-scale AI models with up to 1 trillion parameters in a local environment.
Through FOX, NVIDIA provides three core functionalities: connecting factory systems and robots, automating AI model training, and managing intelligent factory workflows. The AI can autonomously identify data gaps, generate training data, and retrain models, as well as visualize factory operations in real-time using video analysis and digital twin technology.
Major manufacturing companies in Taiwan have already adopted FOX. Foxconn, the world's largest electronics contract manufacturer, is using FOX to build a multi-agent manufacturing operation system called 'MoMClaw.' This system connects hundreds of AI agents with production equipment, sensors, and various digital systems to provide factory managers with natural language-based decision support.
Foxconn expects that implementing this system will reduce root cause analysis time by 80%, increase labor productivity by 15%, and decrease equipment failure rates by 10%.
Pegatron is also developing a factory operation platform that integrates specialized agents for material transport, AI inspection, and work procedure management. This is anticipated to enhance robot utilization efficiency and reduce redundant equipment investments, potentially lowering asset costs by about 15%.
Advantech, an industrial computer company, has established an 'AI Factory Brain' to autonomously manage energy usage in heating, ventilation, and air conditioning (HVAC) and lighting systems. The company estimates this will reduce factory energy consumption by approximately 10%.
Wistron is developing AI agents for surface mount technology (SMT) processes to automate production line analysis and quality control. Additionally, various software companies, including DeepHow, OverviewAI, Roboflow, and Spingence, are developing manufacturing-specific AI agents based on the NVIDIA platform.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.