Mecka AI Series B: $60M From Sequoia for Robot Training Data

Mecka AI, a two-year-old company that pays people to record themselves doing ordinary physical tasks, has closed a $60 million Series B led by Sequoia Capital. The round is a bet that the main bottleneck in robotics is not smarter models but the scarcity of data showing how the physical world is handled. Nvidia and Microsoft's venture arm M12 participated, and TechCrunch had earlier reported that the round would value the startup at around $500 million. 1

What happened

The round was led by Sequoia. New investors include Nvidia, M12, Qualcomm Ventures and Samsung. 23 Kindred, Framework Ventures and Neo also took part. Individual backers include DoorDash CEO Tony Xu, former ServiceNow and Snowflake chief Frank Slootman, and Milan Kovac, who previously led Tesla's Optimus humanoid program. 3

Mecka says it will use the money to build more of its data-capture hardware, expand its research work and increase the number of robots it deploys for customers. 3

The business: recording what was never written down

Mecka was founded in 2024. It pays people to perform everyday tasks, such as brewing coffee or repairing cars, while wearing body sensors and using smartphones. It then turns those recordings into training material for humanoid and other robots. 1 The company uses its own multi-sensor rigs and processes the footage through motion tracking, 3D reconstruction and sensor alignment so that it becomes usable training signal. 3

WOWTALE's coverage explains the reasoning behind the company most clearly. Language models learned from the huge written record humanity left online. No comparable archive exists for how a hand grips, folds or pours, how much pressure to apply, or when to let go. Information about motion, contact, force and geometry was never published on the internet, so it cannot be scraped, licensed or purchased. In Mecka's view, robots are short on experience rather than intelligence. 2 To test the idea, Mecka runs EgoVerse, a human-to-robot transfer study conducted with researchers. 23

TechCrunch compares the company to Scale AI, Mercor and Surge, the data-labeling firms that supplied human-generated data for large language models. Mecka wants to fill the same role for robotics. 1

The revenue claim

The AI Insider reports the most striking figure: Mecka says it passed $100 million in annualized run-rate revenue in June 2026 and expects to reach $300 million by the end of the year. 3 These are company-reported numbers and run-rate metrics can flatter a business, so they deserve some caution. Still, if they hold, they would make the roughly $500 million valuation look modest next to many AI deals.

The projected growth depends on a change in strategy. Mecka plans to move beyond selling robotics data into robot integration, post-training and ongoing deployment services. 3 That widens the opportunity, but it also changes what kind of company Mecka is. Selling data is relatively light on capital. Deploying and maintaining robots for commercial customers involves hardware, operations and support costs.

A crowded field

Mecka has competition. XDOF also collects real-world data for robot training and was reportedly in talks to raise a Series B at a $1.2 billion valuation. 1 TechCrunch also notes that human-data platforms that started out serving LLM developers are now moving into robotics. 1 That means Mecka could end up competing with well-funded incumbents that already run large networks of paid contributors.

Reading the round

The investor list says a lot. Nvidia, Qualcomm and Samsung make chips and devices and have obvious reasons to want the robotics ecosystem to grow. Kovac's involvement adds credibility from inside the humanoid industry. This looks like strategic backing as much as financial backing.

The outlets diverge mostly in emphasis rather than in facts:

• TechCrunch treats Mecka as one entrant in an emerging category and focuses on the valuation and competitors. 1
• WOWTALE focuses on the underlying argument that robots need recorded experience. 2
• The AI Insider highlights the revenue figures and the move toward deployment. 3

Taken together, the round shows investors betting that the "data layer" model, which paid off during the LLM boom, will repeat in physical AI. That is a reasonable thesis, since human demonstration data really is scarce. The bigger question is whether Mecka's advantage lies in its data or in its services. Data collection can be copied by competitors with more money and larger contributor networks. Close integration with customers' robot fleets is harder to copy, but it is also harder to scale.

Mecka's decision to move into deployment suggests it sees that risk. Whether its ambitious revenue targets turn into lasting returns will depend on how well it executes in the second business, not the first.