Mifeng Technology announced Aug. 17 that it has closed a new financing round worth "hundreds of millions" of yuan, with China Telecom as the lead investor and participation from Zhangjiang Group. The round also saw follow-on commitments from existing shareholders including HSG and Yuanqi Innovation, according to the company.
Funding to expand embodied AI data supply and hardware production
The company said the fresh capital will be used to build out a platform-based supply system for embodied AI data, expand mass production of its MEgo product line and deepen capabilities across data governance and evaluation.
- Infrastructure: accelerate efforts to overcome bottlenecks in physical-world interaction data and create standardized, scalable data provisioning.
- Products: increase output of MEgo devices — described by the company as grippers and head-mounted units for lightweight, multimodal data capture in factory and home settings.
- Capabilities: reinforce full-stack abilities in data collection, preprocessing, annotation, feature extraction and closed-loop evaluation to support large-model iteration.
Mifeng positions itself as a one-stop supplier of standardized embodied AI data for robotics manufacturers, operating under a B2B model to address what it describes as a persistent industry shortage of high-quality interaction datasets. The company launched its physical AI data platform in April 2026 and has rapidly pursued growth since its February 2026 spinout from the restructuring of Zhiyuan.
Rapid fundraising after corporate spinout
Since separating from Zhiyuan, Mifeng has completed three financing rounds in roughly six months, reflecting strong investor interest in physical-world data capabilities. The most recent round, led by a major state-affiliated telecom, underscores a strategic alignment between large infrastructure players and data suppliers for embodied AI.
| Investor | Role |
|---|---|
| China Telecom | Lead investor |
| Zhangjiang Group | Participant |
| HSG, Yuanqi Innovation | Existing investors; oversubscribed follow-on investments |
Company statements indicate the funds will enable Mifeng to scale toward producing "tens of millions of hours" of physical interaction data capacity, which it says is necessary to support iteration of physical-world large models. The MEgo devices are intended to collect lightweight, multimodal datasets in both industrial and domestic contexts, feeding into the companys data governance pipeline for annotation and evaluation.
Industry implications and context
The investment highlights two broader trends in China's AI ecosystem. First, major telecommunications and infrastructure firms are moving to secure upstream sources of specialized data deemed essential for next-generation models and robotics. Second, startups focused on embodied data collection are attracting rapid capital as the market for physical-world AI training sets emerges.
For robotics companies, more accessible, standardized datasets could shorten development cycles for perception and control models. For investors and policymakers, the expansion of scaled data production raises questions about data quality standards, governance practices and the industrial pathways for deploying AI in factory and home environments.
Mifeng's financing follows its statement that it will continue to invest in platform architecture and closed-loop evaluation systems. The company says these efforts aim to ensure systematic, reproducible data pipelines that can support both customers and model developers as embodied AI applications multiply.
The company did not disclose an exact valuation tied to this round, nor did it provide a precise monetary figure beyond the characterization of the financing as "hundreds of millions" of yuan. It likewise did not publish a timeline for when expanded MEgo production would reach full capacity.
As embodied AI moves from research demonstrations to commercial deployments, the ability to supply high-quality, large-scale interaction datasets will be a critical enabler for robotics makers and physical-world model training across industrial and consumer sectors.
