Mifeng Technology Opens Data Collection to the Public with ‘Mifeng Pai’
TREE NEWS reports: On September 23, Chinese startup Mifeng Technology officially launched “Mifeng Pai,” a crowdsourced data platform for embodied AI. The platform allows ordinary people to rent MEgo devices, accept data-collection tasks via an app, and earn commissions based on validated data hours. The company, founded in February 2026, positions itself as a third-party data platform serving the entire embodied AI industry, covering real-machine data, body-less data, and simulation data.
The move comes as demand for training data in embodied AI explodes. Mifeng’s chairman and CEO, Yao Maoqing, noted that million-hour datasets are shifting from an industry-wide supply goal to a per-customer requirement, with some clients requesting tens of millions of hours and top customers demanding 50,000–100,000 hours per week. Traditional centralized data-collection factories, where operators teleoperate robots to perform repetitive tasks, are expensive and difficult to scale linearly.
Mifeng Pai aims to change that by distributing data production across real-world work and living environments. Participants rent a MEgo device for 39 yuan per day (promotional price: 19 yuan) and earn a base return of about 20 yuan per valid hour, plus dynamic subsidies based on city and task quality. Tasks span over 20 domains, including maintenance, catering, logistics, elderly care, and housekeeping. In its first month of beta testing, the platform registered 20,000 users and generated 13,000 task submissions, with top earners making over 5,000 yuan in a month.
Market Implications: Data as the New Oil for Robotics
The launch signals a broader shift in how the AI and robotics industry sources training data. As embodied AI models grow more capable, they require not just more hours but also more diverse scenarios, higher spatial precision, and richer action semantics. Mifeng’s crowdsourcing model could lower the cost of data collection and accelerate the development of general-purpose robots. For investors, this underscores the growing importance of data infrastructure in the AI value chain—a segment that has attracted significant venture capital and could see consolidation as scale becomes critical.
Mifeng also announced a “Scenario Data Alliance” with over 50 enterprises and institutions in hotels, retail, manufacturing, and other sectors. The company envisions a network that can dispatch data-collection tasks on demand, akin to a “Meituan for data.” However, the model faces challenges: ensuring data quality from dispersed workers, managing complex operations, and competing with robot makers that may eventually collect data themselves as their fleets scale.
Key Takeaways for Investors
- Data infrastructure is a critical bottleneck: Embodied AI companies need massive, diverse datasets. Platforms that can efficiently aggregate and process such data could become essential suppliers.
- Crowdsourcing may reduce costs but adds operational complexity: While it avoids the overhead of dedicated data-collection centers, quality control and task standardization remain significant hurdles.
- Consolidation likely: As demand grows, customers may prefer fewer, larger suppliers capable of delivering millions of hours with consistent quality. Mifeng anticipates an oligopolistic market.
- Long-term risk: As robots become more autonomous, they may generate their own training data, potentially reducing reliance on human-collected data for mature tasks.
Mifeng’s bet is that the world still has countless scenarios robots have yet to master, and that a distributed network of human workers can reach them faster and cheaper than any centralized factory. Whether that bet pays off will depend on execution and the pace of robotics adoption.
