Mecka AI Raises $60M Series B to Build Robot Training Data Infrastructure

Funding Round and Valuation

Mecka AI, a robotics data-collection startup founded in 2024, has secured $60 million in Series B funding led by Sequoia Capital. The round includes participation from major tech investors including Nvidia and Microsoft's venture arm M12. The funding values the company at approximately $500 million, according to previous reporting, positioning it as a significant player in the emerging robotics AI infrastructure space.

Business Model and Data Collection

Mecka AI operates by paying people to perform everyday tasks—such as making coffee, fixing cars, or other common activities—while wearing body sensors and using smartphones to capture motion data. This human-generated data is then used to train humanoid robots and other robotic systems. The approach mirrors the data-labeling services that have become critical infrastructure for large language models, with companies like Scale AI and others providing the human-annotated training data that powers modern AI systems.

The startup is attempting to establish the same foundational role in robotics that data-labeling companies have achieved in the LLM space. By systematizing the collection of real-world human motion and behavior data, Mecka AI aims to accelerate the development of physically capable robots that can perform complex, real-world tasks.

Competitive Landscape

Mecka AI is not alone in recognizing the importance of human data for robot training. Competitor XDOF was previously reported to be in discussions for its own Series B funding at a $1.2 billion valuation, indicating strong investor appetite in this sector. Additionally, established data platforms that initially focused on LLM training—including Scale AI and Micro1—are expanding their operations into robotics, creating a crowded but growing market.

The competitive dynamics suggest that data infrastructure for robotics is becoming a critical bottleneck that investors believe requires dedicated solutions. As humanoid robots and industrial automation systems advance, the quality and scale of training data available will likely determine which systems achieve superior real-world performance.