Optimizing semiconductor manufacturing with AI: from Tool-level and WIP predictions to fab scheduling

An overview is given of how AI is transforming operations in semiconductor manufacturing and Digital twins. In particular, we demonstrate our deployments of Deep Reinforcement Learning for fab scheduling at a 300mm manufacturer, and tool digital twin to improve real-time dispatching at a foundry. We explain how the modularity of deep learning, for which the sum is greater than its parts, allows integrating information from the equipment to the full fab level, enabling a comprehensive Digital Twin formulation incorporating learned Neural Network based models.