Mecka raises $60M Series B led by Sequoia to build robotics training data

Mecka.ai has raised a $60 million Series B round led by Sequoia Capital. The startup’s goal is to build an internet-scale dataset for training robots.

The round, announced October 7, 2026, also brought in NVIDIA, M12 (Microsoft’s venture fund), Qualcomm Ventures and Samsung. The round also included returning backers Framework Ventures, Kindred Ventures and Neo.

The angel list is notable too. Tony Xu, Frank Slootman and Milan Kovac all participated as individual investors.

The company did not disclose a post-money valuation for this round. Earlier discussions suggested a valuation close to $500 million, though that figure has not been tied to the new financing.

This is not Mecka’s first trip to the well. It raised an $8 million seed round in August 2025, then a $60 million Series A split into tranches in November 2025 and June 2026.

Add it up and the company has now disclosed $128 million in funding across three rounds.

Mecka reported an annual run-rate revenue of more than $100 million as of June 2026. It is expected to reach $300 million by the end of the year.

A startup of 40 to 60 employees reporting nine-figure run-rate revenue is an unusual sight in a sector better known for burning cash than collecting it.

Why record people instead of robots

Mecka collects egocentric data, meaning footage and motion captured from the first-person perspective of the person doing the task. To gather it, the company uses body sensors, smartphones and specialized hardware. Recording happens in ordinary real-world settings rather than controlled labs.

Participants are paid to perform demonstrations of everyday activities.

Mecka’s in-house video-understanding lab focuses on motion tracking and 3D reconstruction, turning flat video into spatial information a machine can learn from.

The company has also built a dataset called EgoVerse. The target customers are developers working on humanoid robots and embodied AI.

The team behind it

Mecka was founded by Josh Gao, Jason Chong, Mogen Cheng and Duy Nguyen. The company operates out of New York and Toronto. Headcount sits somewhere between 40 and 60 people.

What this means for the robotics data race

There are open questions worth watching. The first is whether the company hits its expected $300 million run-rate by year end. Growing from more than $100 million in June to triple that in six months would require serious momentum.

The second is how well egocentric human data actually transfers to machines. A human hand and a robot gripper are not the same tool, and motion tracking and 3D reconstruction have to bridge that gap.

The third is valuation. Without a disclosed post-money figure, outsiders are left with earlier talk of roughly $500 million.