Mecka AI has raised a $60 million Series B round led by Sequoia. The startup collects human motion data to train humanoid and other robots, according to TechCrunch's October 7 report.
Mecka AI is a robot-data supplier founded in 2024. It pays people to record everyday tasks with body sensors and smartphones. It then structures those recordings into labeled motion datasets for sale to robot developers.
Table of Contents
- How Mecka turns chores into training data
- What EgoVerse gives developers
- Who is backing the scale-up
- What buyers should verify first
How Mecka turns chores into training data
Demonstrators wear body sensors and use smartphones during normal tasks. Recorded jobs include making coffee, fixing cars, and folding laundry. The approach captures routine motion outside a lab.
Mecka cleans, labels, and packages the recordings as motion datasets. Robot teams can license the packages instead of running their own capture programs. The trade shifts spending from rigs and staffing to purchased data.
What EgoVerse gives developers
Its EgoVerse dataset holds 1,362 hours across 80,000 episodes from 2,087 demonstrators, according to AI Weekly's dataset breakdown. An episode is one recorded run of a task by a demonstrator. The size matters because robot learning needs varied repeats of the same job.
Mecka sells motion, contact, and force signals tied to those tasks. Developers can use the signals to train and test humanoid and other robots. Buyers should match the task mix to their own deployment needs.
Who is backing the scale-up
The round included Nvidia, Microsoft's venture fund M12, Qualcomm Ventures, and Samsung. Kindred, Framework Ventures, and Neo also joined. The mix ties chip suppliers, platform investors, and robotics backers to one data bet.
People familiar with the terms put the Series B valuation at about $500 million. Total disclosed funding now stands at roughly $128 million, according to Bytevyte's funding breakdown. The new capital supports larger collection and wider deployment.
What buyers should verify first
Mecka describes customers only as several top frontier robotics labs and multiple Mag 7 companies. It has not named the systems using the data or published measured performance gains, as noted in Dataist's coverage.
Buyers therefore cannot judge fit from customer proof alone. Humanoid builders facing a data bottleneck can buy motion data rather than build capture rigs, operators, and annotation pipelines. That option saves time when the needed tasks overlap Mecka's catalog.
- Compare Mecka's task list with your target tasks.
- Ask for label schema and sensor details.
- Test one skill in a pilot before scaling.
