In Xinhua's 'AI in Focus', DaoAI lays out its read on physical AI: 2026 is widely seen as the starting point for physical AI at scale. Unlike the purely virtual narrative of the metaverse, physical AI already has a clear cost-down/efficiency-up business loop, real hardware to run on, and genuine industrial demand — a business logic that is verifiable.
Source / Source: Xinhua News · 2026-07-20
The report first clears up a misconception: physical AI is not the same as 'embodied intelligence', nor another rerun of a pure concept. It stresses that AI is moving from the virtual into the physical, and the crux is whether it can truly understand 3D space, reason by physical laws, and drive hardware to interact in the real world.
From 'seeing' to 'understanding' the physical world
Manufacturing is full of hazardous, highly repetitive steps that demand extreme consistency. DaoAI replaces human work in harsh conditions with AI-vision and robot-vision systems, so machines not only see but understand space and causality — inspecting and operating with precision.
Underpinning all of this is the company's own DaoAI World model — a 'brain' that understands the physical world, distilling spatial intelligence into a foundation reusable across scenarios.
Already running on real lines
- Assembly and appearance inspection on air-conditioner lines
- Micron-level defect detection and classification for semiconductors and chips
- Solder-joint and component quality control for PCBA
- ~30-minute line deployment, 99%+ accuracy, go-live from 1–20 good samples
The value of physical AI is not how novel the concept sounds, but whether the factory can actually put it to work.
Xinhua's 'AI in Focus' column argues that local companies like DaoAI are turning physical AI from vision into repeatable, at-scale productivity — a key link in Chinese smart manufacturing's next stop.
FAQ
Is physical AI the same as embodied intelligence?
No. Per Xinhua, physical AI is about AI moving from virtual to physical — understanding 3D space and reasoning by physical laws to drive hardware. It is neither embodied intelligence nor mere hype.
Is physical AI already commercialized?
Yes. 2026 is seen as its scale-up starting point, with a clear cost-down/efficiency-up loop and real hardware, already running on air-conditioner, semiconductor and PCB lines.
What powers DaoAI's physical AI?
The self-developed DaoAI World model — a foundation that understands the physical world and distills spatial intelligence into cross-scenario capabilities for inspection, robot vision and more.