Physical AI—AI systems that perceive, reason and act in the real world—is emerging as the next wave of automation, driven by breakthroughs in generative AI, simulation and hardware.

Physical AI combines AI models with sensors, actuators and control systems to allow AI to act upon real-world environments. Unlike rule-based robots of the past, AI-powered agents equipped with large language models (LLMs—AI that understands and generates text) can perceive environments, reason about them and learn from outcomes.

Jensen Huang, CEO of Nvidia, predicted during a January 2026 podcast interview a future with 'a billion robots' and described it as 'the ChatGPT moment for robotics.' That same month, Nvidia released open models, frameworks and infrastructure for physical AI to speed up workflows across the robot development lifecycle.

Training uses simulation (computer-generated environments with physics modeling and photorealistic rendering) combined with reinforcement learning (a machine-learning process in which agents learn through trial and error). Robots are first trained in simulation across varying conditions—different weather, crowds, terrain—then fine-tuned in limited real-world environments before deployment to more complex settings.