Asia is winning the physical AI race: Japan pledges $63 billion while China builds a state funding machine for robots

Robots that perceive, reason and act on their own have moved from lab demos to real deployment, and Asia is setting the pace. Physical AI began commercializing in 2026 and will accelerate from 2027, according to a new report from research firm GlobalData.

Japan alone has committed JPY10 trillion ($63 billion) to advanced robotics and AI. No dedicated law governs these systems yet.

What is physical AI?

Physical AI refers to embodied systems that perceive, reason and act autonomously. It can be deployed in robots, autonomous vehicles, drones and industrial systems.

GlobalData’s Strategic Intelligence report, “Physical AI,” describes it as a fusion of technologies rather than a single one:

• Internet of Things (IoT)
• Generative AI and agentic AI
• Machine learning
• World models
• Vision-language models (VLMs) and vision-language action (VLA) models

These systems run a four-step continuous loop of perception, reasoning and context, learning, and action. GlobalData says the loop relies on localized feedback and iterative adaptation, not the recursive learning that has been in the news recently.

“Intelligence is no longer confined to the digital realm,” said William Rojas, Research Director, Strategic Intelligence at GlobalData. “Physical AI learns through structural adaptation, not just statistical training.”

Asia-Pacific leads the race

Japan, China, South Korea, Taiwan, Singapore and the US lead in both R&D investment and deployment, the report says. Several moves stand out:

• Japan: JPY10 trillion ($63 billion) committed to advanced robotics and AI
• China: central and local governments are coordinating funding and support, and the 15th Five-Year Plan commits to nurturing the technology and building funding and risk-sharing mechanisms
• Nvidia: has formed a coalition with 10 companies and organizations with robotics expertise

“This is a global race, and it is being led out of Asia-Pacific,” Rojas said. “Physical AI is being treated as critical national infrastructure, not a science project.”

Commercialization is under way in discrete manufacturing, healthcare, mining, energy and transportation. GlobalData expects growth to accelerate from 2027 as the cost of dexterous actuation and edge computing hardware falls and large language models mature.

No physical AI law yet

Regulation is trailing the technology. Physical AI is governed by overlapping existing frameworks stretched to fit embodied systems, with technical standards holding them together and liability questions unresolved.

“Globally, no specific physical AI law exists,” said Nilesh Raghoo, Associate Analyst, Strategic Intelligence at GlobalData. “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 those rules, Raghoo said.