Reporters learned from Mifeng Technology, a one-stop physical AI data service platform, that the company has completed a new financing round of hundreds of millions of yuan. This round was led byChina TelecomLead investment by (6.630, 0.06, 0.91%), with follow-on investment from Zhangjiang Group and over-subscriptions from existing investors including Sequoia China and Yuanqi Innovation. This round will be used to break through the industry's "data desert" bottleneck, build an embodied AI data platform infrastructure, scale up mass production of MEgo series sensorless data collection products, strengthen full-stack capabilities in data acquisition, governance, and closed-loop evaluation, accelerate the deployment of physical interaction data capacity at the tens-of-millions-hour level, and redefine the iteration paradigm for physical world foundation models.
Since its founding earlier this year, MiFeng Technology has completed multiple rounds of financing within approximately six months. Investors include state-backed industrial platforms, top-tier market funds, and leading industry players. This sustained and intensive capital support underscores the strong confidence of both the financial markets and the industry in the long-term growth potential of the physical AI data sector, as well as high recognition of MiFeng's full-stack technology approach and platform-based business model.
Live data collection in real-world scenarios: the ultimate data platform built for models and application needs.
For physical AI to truly reach its "GPT moment"—mastering environmental changes, physical laws, and higher-order causal reasoning—large-scale, high-quality real-world physical interaction data is essential. By comparison, mainstream large language models have been pre-trained on 100 trillion tokens, equivalent to 100 billion hours of data. Physical AI, however, requires even more data due to its lower effective information density and higher noise levels. To break through this industry bottleneck, MiFeng Technology has built a full-stack, self-developed infrastructure for embodied intelligence data supply by strengthening three key areas: data collection, data governance, and evaluation. This creates a platform-based supply system tailored for embodied intelligence.
At the data collection stage, Mefeng Technology's self-developed MEgo series of body-free collection products has achieved mass production and delivery. These products have already collected data in real-world scenarios including factories, logistics centers, supermarkets, homes, hotels, restaurants, security systems, elderly care facilities, cultural tourism sites, and more. The MEgo series enables precise synchronized collection of macro-environmental perception and micro-hand operation data without reliance on external infrastructure.Robot(15.650, 0.23, 1.49%)本体, leveraging lightweight portable hardware to penetrate diverse scenarios, achieves full-domain, high-fidelity, low-cost scalable data collection. Building on this foundation, MiFeng Technology has constructed a layered data supply pyramid, covering all types including real-device teleoperation and non-bodily data acquisition.
At the governance level, Meifeng Technology's MEgo Engine data governance platform has established a fully automated closed loop covering preprocessing, spatial reconstruction, multimodal annotation, and quality evaluation, boosting traditional manual data processing efficiency by over 10 times. Built on our proprietary VIO high-precision pose capture algorithm, the platform supports human-machine skill transfer, long-sequence action segmentation, and intelligent atomic operation labeling, significantly reducing data processing costs for enterprises and research institutions.
On the evaluation side, we build a closed-loop iteration system using multi-dimensional data insights to optimize data collection strategies and continuously improve deployment ROI. By deploying a data flywheel, Mifeng Technology connects task execution, data return, case mining, and data updates into an automated loop—transforming data supply from a one-time delivery into a continuous update cycle alongside model deployment. In select application scenarios, each data feedback loop boosts task success rates by 5% to 10%. This is why Mifeng Technology is recognized as the "physical AI data infrastructure platform that best understands model and scenario needs."
Currently, MiFeng Technology has established over 20 domestic innovation centers and deployed more than 5 overseas nodes. We have formed strategic partnerships with leading internet, embodied AI, and enterprise AI companies both domestically and internationally to provide high-quality training data for Embodied Foundation Models such as Ant LingBot-VLA 2.0.
Collaborating cloud networks, scenarios, and ecosystems to accelerate the development of physical AI data infrastructure.
Investors in this round bring extensive industry and scenario advantages, further strengthening MiFeng Technology's ecosystem collaboration capabilities across data collection, governance, and application to build a pool of authentic, compliant, and high-quality embodied AI scene data.
China Telecom is a global leader in integrated intelligent information services, boasting robust cloud-network infrastructure, extensive customer channels, and strong government and enterprise resources. Leveraging China Telecom's strengths in 5G networks, cloud storage, computing power scheduling, and scenario connectivity, combined with MiFeng Technology's capabilities in end-to-end embodied intelligence data collection, dataset production, data governance, and integrated hardware-software solutions, the two companies plan to jointly develop standardized "Cloud-Network + Embodied Data" integrated solutions for humanoid robot manufacturers, smart manufacturing enterprises, research institutions, smart parks, and government/enterprise clients in smart home sectors. This initiative aims to drive efficient data collection, secure storage, scalable governance, and standardized supply of embodied data.
Zhangjiang Science City is a world-class hub for science and technology innovation, with deep expertise in industrial clustering, open application scenarios, and the allocation of innovative resources. As the primary force behind Zhangjiang Science City's development, Zhangjiang Group will open up diverse real-world application scenarios—including parks, industrial zones, and commercial spaces—alongside access to industry policies, spatial carriers, and upstream-downstream enterprise networking resources. Together with MiFeng Technology, we will break through bottlenecks in embodied AI data supply across hardware, technology, scenarios, and ecosystem dimensions. The two parties will co-build a public service platform for embodied AI data, jointly develop high-quality industry datasets, and drive the implementation of data standards and the集聚 of ecosystem enterprises.
Lin Chen, Deputy General Manager of Zhangjiang Group, stated: "MiFeng Technology is a rare domestic platform-based infrastructure provider for embodied AI data, offering end-to-end capabilities spanning hardware data collection, automated governance, and scalable data supply. Its technical roadmap aligns closely with industry development trends. We look forward to deepening the value extraction from data elements through continued collaboration and accelerating the transformation and commercialization of more embodied AI innovations in Zhangjiang."
Yao Maoqing, Chairman and CEO of MiFeng Technology, stated that the transition of embodied AI from labs to the real world hinges on the ability to continuously acquire authentic, compliant, high-quality data. This funding round brings not only capital support but also deep synergy across cloud-network infrastructure, industrial scenarios, and ecosystem resources. Leveraging its end-to-end capabilities spanning "collection, governance, supply, and evaluation," MiFeng Technology will enhance large-scale data supply and scenario service capabilities, accelerating the transformation of high-quality data into model capabilities and tangible industry value.