Physical AI marks a shift from machines that follow fixed instructions to embodied systems that perceive, reason, and act autonomously. It can be deployed in robots, autonomous vehicles, drones, and industrial systems. Commercialization began in 2026 across industries ranging from discrete manufacturing and healthcare to mining, energy, and transportation. It will accelerate from 2027 as the cost of dexterous actuation and edge computing hardware falls and large language models (LLMs) mature, redefining the future of robotics, says GlobalData, a leading intelligence and productivity platform.
GlobalData’s latest Strategic Intelligence report, “Physical AI”, states that physical AI is not a single technology but rather a fusion of multiple technologies including the Internet of Things (IoT), generative AI, agentic AI, machine learning, world models, vision-language models (VLMs), and vision-language action (VLA) models. Physical AI operates through a four-step continuous loop: perception, reasoning and context, learning, and action. This perception–action loop uses localized feedback loops and iterative adaptation, but it does not use recursive learning, which has been in the news recently.
William Rojas, Research Director, Strategic Intelligence at GlobalData, comments: “Intelligence is no longer confined to the digital realm. We are shifting from machines that blindly follow instructions in a deterministic fashion to systems that perceive, reason, and act independently. Physical AI learns through structural adaptation, not just statistical training. It gains skills via real-world experience, modulating force and movement, which reshape its control models.”
A few dominant countries shape the competitive landscape for physical AI. Japan, China, South Korea, Taiwan, Singapore, and the US lead in R&D investment and in actual deployment. Japan alone has committed JPY10 trillion ($63 billion) to advanced robotics and AI. Moreover, Nvidia has formed a coalition with 10 companies and organizations that have considerable expertise in robotics. Additionally, China is coordinating funding and support for physical AI between central and local governments. The 15th Five-Year Plan outlines a commitment to nurturing this technology and establishing mechanisms for funding and risk-sharing.
Rojas continues: “This is a global race, and it is being led out of Asia-Pacific. Japan and South Korea’s heritage in precision robotics, combined with enormous state backing, gives the region an enviable head start. Physical AI is being treated as critical national infrastructure, not a science project.”
AI regulation is not well developed and struggles to keep pace with technological breakthroughs. Physical AI systems are currently regulated not by any dedicated law, but through overlapping pre-existing frameworks that are extended to fit embodied systems. In practice, technical standards hold these together, while unresolved liability questions loom.
Nilesh Raghoo, Associate Analyst, Strategic Intelligence at GlobalData, concludes: “Regulation in physical AI varies widely by region. Globally, no specific physical AI law exists. Instead, machinery law, product-safety law, product-liability law, and horizontal AI law across applicable jurisdictions all apply simultaneously. Much of the current work involves harmonizing these regulations.”
