Physical AI for Autonomous Manufacturing: Real Time Sensing, Data, Models, Control, and Advanced Packaging

Recent advances in artificial intelligence (AI) have accelerated AI’s transition from purely digital applications to cyber-physical systems that interact with the real world in real time. This evolution, often described as Physical AI, has already been demonstrated in domains such as autonomous driving and is now drawing significant attention in manufacturing. Physical AI-based manufacturing requires continuous acquisition of process data, real-time inference, and closedloop control of equipment and robotic systems. Although such capabilities are challenging even at the level of a single machine, their deployment across a factory introduces substantial system-level bottlenecks. In particular, the semiconductor platforms that support these workloads must sustain massive computational throughput while meeting stringent requirements for bandwidth, latency, power efficiency, thermal management, and long-term reliability. This talk introduces practical examples of Physical AI-enabled autonomous manufacturing in the automotive industry and argues that advanced packaging technologies are a key enabler for practical and scalable implementation.