Mecka AI raises $60M Sequoia-led Series B for human motion data to train robots

Mecka AI, a startup that gathers human motion data to train robots, has raised a $60 million Series B round led by Sequoia, according to TechCrunch. Nvidia, Microsoft's venture fund M12 and other investors also participated in the round.

A data-labeling playbook, applied to robots

Founded in 2024, Mecka AI is applying the data-supply playbook that served large language models to physical machines. For LLMs, companies such as Scale AI, Mercor and Surge built businesses supplying the human-generated training data those systems learn from. Mecka AI wants to occupy the equivalent position in robotics.

Its method differs from text labeling: the startup pays people to record themselves carrying out routine activities, such as making coffee or repairing cars, while wearing body sensors and using smartphones. The motion data captured this way is then collected and analyzed to train humanoid robots and other kinds of machines.

What we know about the round

Sequoia led the Series B, with Nvidia, M12 and other unnamed backers joining. The announcement did not include a valuation figure, but TechCrunch had previously reported that Mecka AI was nearing a new funding round at a valuation of about $500 million.

A crowded field

Mecka AI is not the only company betting that real-world human data is what robots currently lack. According to TechCrunch's earlier reporting, XDOF, another startup collecting real-world data for robot training, was in talks to raise a Series B at a $1.2 billion valuation.

At the same time, data platforms that began by serving LLM developers are pushing into robotics. TechCrunch names Scale AI and Micro1 as examples of companies expanding from language-model data work into robot training data, which means Mecka AI faces competition both from fellow robotics-native startups and from larger incumbents with existing delivery infrastructure.

Why it matters

Large language models advanced quickly once an entire industry emerged to supply them with human-generated training data, and investors appear to be betting the same dynamic will play out in robotics. As humanoid robots move closer to practical deployment, the scarce input is likely to be high-quality demonstrations of physical tasks — data that is far harder to collect at scale than text scraped from the web.

A $60 million round led by Sequoia, with strategic capital from Nvidia and Microsoft's M12 alongside, signals conviction that motion data collection could become a foundational layer of the robotics stack rather than a niche service. It also points to a growing gig economy around physical demonstrations: people being paid to brew coffee or fix a car so that a robot can eventually learn to do the same.