Why Asia’s Physical AI boom will be decided at the camera, not the model
Robotics and rehabilitation are undergoing a profound transformation that the industry’s own measurement infrastructure has largely failed to keep pace with. For four decades the recording of movement in three dimensions has been organised around a building in a studio. From a wired ring of cameras, a morning of calibration, and a purchase order that […] The post Why Asia’s Physical AI boom will…
Asia is the epicenter of the burgeoning Physical AI boom, with robotics and rehabilitation systems undergoing a significant transformation. Traditionally, movement has been recorded in three dimensions within a studio using a wired ring of cameras, resulting in excellent scientific results but limited scalability. The mismatch between demand for motion data and the industry's capacity to generate it has led to an underpriced opportunity known as the "capture gap."
Asia is uniquely positioned to capitalize on this gap, accounting for more than half of all deals in the robotics and Physical AI sector in the first quarter of 2026, although only around a quarter of the total capital invested. Companies such as LG Electronics in Korea and the J-HRTI consortium in Japan are building data factories to train their robots on large datasets of human motion.
China has also announced over 40 humanoid training centers, while AgiBot in Shanghai has signed a lease agreement with Singtel for robots in Singapore.
The bottleneck in Physical AI is the cost of collecting high-quality teleoperated data, which currently ranges from $340 per hour in early 2024 to $118 per hour by March 2026. To obtain operational readiness, a single manipulation task requires between 200 and 2,000 demonstrations. The industry now relies on three sources of data: teleoperation, simulation, and ordinary video of people at work. However, ground truth for synchronizing multi-camera capture is the most underpriced component of the Physical AI stack.
Existing solutions for capturing human motion are either expensive or lack the necessary synchronization and calibration. However, a new generation of affordable multi-camera rigs is emerging. Stanford's OpenCap is free and runs on two smartphones, Uplift Labs in Palo Alto has trained nearly 20,000 athletes using phone video, and Sportip in Tsukuba reads posture and gait from a tablet.
VALD's HumanTrak has placed a single depth camera in physiotherapy clinics across Australia and Asia, all for a price between $2,000 and $10,000. This price range represents the capture gap, where significant capital can be made by developing affordable motion labs that can operate anywhere in the field.
Written by urgent.news from e27's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
