2025 – The Evolution of AI Applications for Process Control at SEMICON West

This paper provides a comprehensive overview of these AI-driven semiconductor manufacturing applications, the system components required to implement them, and emerging capabilities shaping the future of Smart Manufacturing. It further examines how modern AI techniques—including large language models (LLMs), agent-based AI, and digital twins—are being integrated into enterprise-level manufacturing control systems to maximize return on investment (ROI) and enhance process control across fabrication, assembly, and test operations.

Semiconductor manufacturers and equipment OEMs have developed numerous AI and ML solutions that demonstrably reduce process variability, manufacturing costs, and defect rates while improving overall yield. However, many of these solutions remain siloed as point applications or are tightly coupled to specific equipment platforms. The key innovation presented in this work is the use of GenAI-powered LLMs and interactive agentic AI systems to securely link and orchestrate outputs from multiple AI models across these silos. This architecture addresses intellectual property (IP) protection and cybersecurity requirements while enabling controlled information sharing, reinforcement learning, and continuous improvement across heterogeneous AI systems.

Results demonstrate that this unified AI framework effectively connects manufacturing data, equipment data, and supply chain information through a comprehensive digital twin. By leveraging subject matter experts to train and guide the system, the platform operates with minimal human intervention, enabling faster decision-making and more adaptive control strategies. Measured outcomes include over 50% improvement in engineering productivity, more than 10% reduction in mean time to repair (MTTR), and significant gains in overall factory efficiency (OFE/OEE), validating the impact of modern AI on next-generation semiconductor manufacturing.