According to the article, Physical AI—AI that interacts with the physical world—is moving to the forefront of industry as large model capabilities and computing infrastructure mature. Over the past two years, AI in capital market transactions has primarily been the 'brain' of AI, with themes revolving around enabling larger model scales, faster training speeds, and lower inference costs. However, most AIs still operate within screens and the digital world. NVIDIA defines physical AI as enabling AI to move beyond the screen, allowing autonomous systems like robots, cameras, and self-driving vehicles to perceive environments and perform reasoning, decision-making, and complex actions. Based on Jensen Huang’s recent public remarks, NVIDIA is building an end-to-end foundational platform spanning training, simulation, inference, and deployment for machines entering the physical world. Physical AI requires understanding spatial relationships and physical laws, leveraging world models, training data, simulation environments, edge computing, machine vision, sensors, and motion control. In the market context, Physical AI largely overlaps with 'embodied intelligence' but has a broader scope, encompassing humanoid robots, autonomous vehicles, industrial robots, drones, smart factories, warehouse systems, and intelligent spaces.