Zhiyuan Robotics has spun out its data business for standalone financing, with China Telecom taking over as lead investor — the third round in six months.
On August 17, Mifeng Technology (Mifeng Tech), a one-stop physical AI data service platform, announced completion of a new funding round worth several hundred million RMB. Though only established in February 2026, the company has closed three rounds within half a year: a seed plus angel round led by Sequoia China, an angel+ round led by Guofang Capital in June, and this round led by China Telecom with Zhangjiang Group participating, with Sequoia and Yuanqi Innovation over-subscribing as existing shareholders.
The 'One-Split-Into-Four' Strategy
Mifeng did not appear from nowhere — it is the data business entity carved out of Zhiyuan Robotics' 'one-split-into-four' strategy from late 2025. Zhiyuan split four business lines into independent entities: dexterous hands (Lindian), data (Mifeng), rentals (Qingtian Zu), and quadruped robots (Zhiyuan Kutuo). Qichacha records show Zhiyuan Innovation (Shanghai) Technology Co., Ltd. remains Mifeng's largest shareholder.
Why spin data out separately? Mifeng Chairman and CEO Yao Maoqing put it bluntly: the team judged that 'what will truly constrain industry development and form moats is data — embodied intelligence models will not have strong moats.' The numbers make this stark: GPT-5's training corpus equates to roughly 10 billion hours, while robots cannot learn directly from the internet, and the entire industry's aggregate high-quality embodied data amounts to only about 500,000 hours — a mismatch of more than an order of magnitude.
Three-Sided Business Built on a 'Data Desert'
- Collection: The self-developed MEgo series of body-free collection terminals are in mass production and delivery. Hardware-independent of robot bodies, lightweight portable devices reach real scenarios across factories, logistics, supermarkets, homes, hotels, restaurants, security, elder care, and tourism, capturing macro-environment and micro hand-operation data simultaneously. The product line covers full-body teleoperation and body-free data collection, forming a tiered data supply pyramid.
- Governance: The MEgo Engine data governance platform provides an automated closed loop of preprocessing, spatial reconstruction, multimodal annotation, and quality evaluation. Its self-developed VIO high-precision pose capture algorithm, human-machine skill transfer, long-sequence action auto-splitting, and atomic operation intelligent annotation improve traditional manual data processing efficiency by more than 10x.
- Evaluation: A closed-loop evaluation system deploys a data flywheel connecting task execution, data backflow, case mining, and data updates. The company claims each data loop can improve task success rates by 5% to 10% in some tested application scenarios.
- Whether Mifeng actually breaks through tens-of-millions-of-hours data capacity within the year
- Whether MEgo terminal deployment ROI in real factories and retail pays back for operators
- Whether the other three Zhiyuan spin-offs (Lindian dexterous hands, Qingtian Zu rentals, Zhiyuan Kutuo quadrupeds) each complete independent financing — only when all four succeed is the Zhiyuan ecosystem fully validated.
Mifeng plans to reach tens-of-millions-of-hours annual data capacity, has innovation centers in 20+ Chinese cities, and has deployed 5+ overseas nodes. Customers include embodied foundation models such as Ant Lingbo's LingBot-VLA 2.0.
Why China Telecom's Lead Matters
China Telecom leading this round signals more than capital. All three major Chinese telecom operators have now placed bets on embodied data: China Telecom co-invested in Domfoe Robotics with Inovance and set up a 2 billion RMB AI industry fund; China Mobile's chain-leader fund also invested in Domfoe; China Unicom built its own embodied intelligence pilot base with an 80TB dataset. Operators view embodied data as the next-generation infrastructure entry point after 'cloud-network + compute,' elevating the strategic value of platform-type data suppliers.
The path overlaps with Yuanli Wuxian's 'Ego data route + AtomBrain' direction, but Mifeng differentiates by not betting on any single robot body: MEgo body-free terminals deploy independently, making collection lighter and cheaper — closer to a 'China Mobile of the data era.' By splitting data from an embodied-brain built-in capability into a platform business, Mifeng takes a To-B service provider route rather than a To-B equipment maker route.
Data has been repeatedly underestimated over the past 12 months. Model weights can be fine-tuned, distilled, and re-architected — but physical-world data can only be collected on site. When China Telecom pays for 'tens-of-millions-of-hours data capacity,' the deciding battleground of the embodied race is shifting from the model layer down to the data layer.