Physical AI to Redefine Robotics

Physical AI is a new class of embodied intelligence that moves beyond fixed instructions to perceive, reason, and act autonomously. By integrating IoT, generative AI, and large language models, it enables robots, vehicles, and drones to learn through real‑world interaction. GlobalData reports that…

By Felo News Desk · Published Sat, 19 Sep 2026 05:04:41 GMT

Physical AI represents a fundamental shift in how machines operate. Instead of following a rigid set of commands, these systems perceive their surroundings, reason about what they see, learn from each interaction, and then act in a way that adapts to new conditions. The result is a robot or vehicle that can navigate a factory floor, a hospital ward, or a mining site without human intervention, continually improving its performance as it works.

From Fixed Instructions to Autonomous Learning

The idea of a machine that can learn from its environment is not new, but the combination of technologies that make it possible has only just come together. Physical AI is not a single invention; it is a fusion of the Internet of Things (IoT), generative artificial intelligence, agentic AI, machine learning, world models, vision‑language models (VLMs), and vision‑language‑action (VLA) models. Together, these components form a four‑step loop: perception, reasoning and context, learning, and action. Each step feeds back into the next, creating a continuous cycle that allows the system to adapt on the fly.

Unlike traditional robots that rely on pre‑programmed routines, Physical AI systems use localized feedback loops. They gather data from sensors, interpret that data through advanced models, adjust their internal control strategies, and then execute a new action. The learning that occurs is structural rather than purely statistical, meaning that the robot’s control models evolve as it experiences different forces and movements in the real world.

Early Commercialisation and Rapid Growth

Commercial deployment of Physical AI began in 2026, with early adopters spanning discrete manufacturing, healthcare, mining, energy, and transportation. The technology is already being used to automate repetitive tasks in factories, assist surgeons with precision tools, and navigate autonomous trucks across long haul routes. By 2027, the cost of dexterous actuation and edge computing hardware is expected to drop significantly, while large language models (LLMs) mature, creating a perfect storm that will accelerate adoption across industries.

GlobalData’s Strategic Intelligence report, titled “Physical AI,” highlights that the technology’s growth will be driven by a combination of falling hardware costs and the increasing sophistication of AI models. The report notes that the integration of LLMs will allow Physical AI systems to understand natural language commands and contextual information, further expanding their usability in complex environments.

Competitive Landscape and Global Investment

Several countries are investing heavily in Physical AI, with Japan, China, South Korea, Taiwan, Singapore, and the United States leading the way. Japan alone has earmarked JPY10 trillion (about $63 billion) for advanced robotics and AI research. In the United States, Nvidia has formed a coalition with ten companies and organizations that bring deep expertise in robotics, signalling a coordinated push toward commercial viability.

China’s 15th Five‑Year Plan explicitly commits to nurturing Physical AI, establishing funding mechanisms and risk‑sharing arrangements between central and local governments. The plan underscores the technology’s strategic importance, positioning it as critical national infrastructure rather than a niche research project.

Regulatory Gaps and Liability Challenges

Regulation of Physical AI remains fragmented. There is no dedicated legal framework that addresses the unique challenges of embodied intelligence. Instead, existing machinery law, product‑safety law, product‑liability law, and horizontal AI law are applied in a piecemeal fashion. Technical standards attempt to bridge gaps, but unresolved liability questions persist, especially when a Physical AI system causes harm or fails to perform as expected.

GlobalData analysts note that harmonising these overlapping regulations is a major focus for the industry. Until a comprehensive legal framework is in place, manufacturers and operators will need to navigate a complex patchwork of rules that vary by region.

What’s Next for Physical AI?

The next few years will see Physical AI move from early pilots to mainstream deployment. As hardware costs continue to fall and AI models become more capable, we can expect to see autonomous drones delivering medical supplies, robots performing complex surgeries, and vehicles navigating congested city streets without human oversight. Meanwhile, regulators will need to catch up, creating clear guidelines that protect consumers while fostering innovation.

For now, the technology’s promise is clear: by blending perception, reasoning, learning, and action, Physical AI can transform industries that rely on precision, safety, and adaptability.

Key facts

Physical AI blends IoT, generative AI, and LLMs to create autonomous, learning systems
Commercial use began in 2026, with rapid expansion expected as hardware costs fall
Japan, China, and the US are leading investment, with Japan committing $63 billion
Regulation is fragmented, relying on existing laws rather than dedicated frameworks
The technology promises safer, more adaptable automation in manufacturing, healthcare, and transportation

Why it matters

Physical AI enables machines to learn from real‑world experience, reducing reliance on pre‑programmed instructions and opening the door to safer, more efficient automation across critical sectors.

Frequently asked questions

What is Physical AI?

Physical AI refers to embodied systems that combine perception, reasoning, learning, and action, allowing them to operate autonomously and adapt through real‑world experience.

When did Physical AI start commercialising?

Commercial deployment began in 2026 across industries such as manufacturing, healthcare, mining, energy, and transportation.

Which countries are investing most in Physical AI?

Japan, China, South Korea, Taiwan, Singapore, and the United States are the leading investors, with Japan allocating JPY10 trillion ($63 billion).

Is there a specific law for Physical AI?

No, Physical AI is regulated under a mix of existing machinery, product‑safety, product‑liability, and horizontal AI laws, with no dedicated legislation yet.