Robot data collection accelerates as related orders at Zhenghai Magnetic enter mass production

Why are robotics companies stepping up real-world data collection?

Tesla has added roles in data collection coordination, operations and night-shift management, accumulating training data through motion capture and object manipulation. Model iteration and scenario validation are driving demand for high-quality real-world data.
Figure AI collects human behavior data worldwide through Index and plans to invest more than 1 billion dollars in data and computing power over the next 12 months. Index pretraining raised its success rate in related tests from 9% to 56%; Helix 2.5 achieved zero-shot, full-body autonomous operation in 30 unfamiliar homes.
Genie 4, released by Google DeepMind, supports real-time generation of interactive worlds, scene modification using natural language and synchronized audio generation, potentially enriching simulation training environments for embodied AI.

Soochow Securities believes faster data collection can drive model iteration, creating a flywheel in which data accumulation and model improvement reinforce each other, potentially advancing large-scale robot production.
The brokerage favors companies well positioned to deploy vertical foundation models for physical AI and the supply chains of overseas giants leading in mass production. It also flags the risks of mass-production progress falling short of expectations and intensifying competition.

Founder Securities says Zhenghai Magnetic's related orders for a globally renowned humanoid robot customer have entered an orderly mass-production phase and could contribute incremental earnings in the second half of the year.
The company's rare-earth permanent magnet technology aligns with demand for components such as coreless motors and frameless torque motors. In the first half of the year, the company advanced the development of materials without heavy rare earths and iterations of high-precision magnet processing technology, while optimizing and establishing supporting production lines to enable supply at scale.
Founder Securities forecasts that the company's net profit attributable to shareholders will be 452 million, 723 million and 969 million yuan in 2026–2028, respectively, up 47%, 60% and 33% year on year. The projected growth faces constraints including downstream demand, customer acquisition, development of the robotics industry and export controls.