From $60M to a $500M Valuation in One Summer: Sequoia Leads the Land Grab for Mecka AI's Robot Training Data

Three months. That is all the time it took for Mecka AI to go from announcing a $60 million fundraise to shopping — successfully, it appears — a round that values the startup at roughly $500 million. According to a TechCrunch report published September 11, the New York-based company, which collects and analyzes human motion data to train humanoid robots and other robotics systems, is nearing a new financing led by Sequoia Capital at a valuation of about half a billion dollars, according to two people with knowledge of the deal.

The precise size of the new round has not been learned, and the terms are not final and could still change. Mecka AI did not respond to a request for comment, and Sequoia declined to comment. But if the round closes near the reported terms, it will be one of the fastest valuation step-ups of the current AI cycle — and the clearest signal yet that the venture capital world has decided the next great data scarcity is not text, images, or code. It is physical motion.

What Mecka actually does

Mecka’s premise is simple and, in hindsight, obvious. Large language models were built on the internet — trillions of tokens of text that already existed, waiting to be scraped, cleaned, and scaled. Robots have no equivalent corpus. There is no internet of physical interactions: no archive of what a human hand feels when it grips a coffee cup, no dataset of how a mechanic’s weight shifts when a bolt refuses to turn. The primary bottleneck holding back general-purpose robots, including humanoids, is that almost nobody has systematically captured real-world interactions at scale.

Mecka’s answer is to pay people to record themselves performing everyday tasks — making coffee, fixing cars, folding laundry — using custom body sensors and smartphones. The company structures this “egocentric” capture (so named because the data is recorded from the performer’s own point of view) into datasets that robotics companies and AI labs use to train their models, alongside other physical data collection methods like teleoperation.

The company’s name is a nod to “mecha,” the fictional giant robot controlled by humans — a fitting metaphor for a business whose product is, in effect, a stream of human bodies teaching machines how to move.

Founders with no robotics pedigree

The origin story is unusual for a robotics-adjacent startup. Mecka was co-founded in 2024 by four entrepreneurs, none of whom have backgrounds in robotics. Canadians Josh Gao and Mogen Cheng previously built a restaurant fintech startup. Jason Chong joined Coinbase after it acquired his crypto exchange. Duy Nguyen, the only non-Canadian on the team, runs operations.

What the founders lacked in robotics credentials they made up for in pattern recognition. They saw that Scale AI, Mercor, Surge, and other human-data companies had built enormous businesses supplying labeled human intelligence to LLM developers — and that the same playbook, aimed at physical AI instead, had no clear leader. Mecka set out to do for robotics what those companies did for language models.

That thesis is now being validated by the money. As of early June, when the startup announced its previous fundraise, CEO Josh Gao told Fortune that Mecka was projecting it would end 2026 at an annual run rate of $100 million. The company has not publicly disclosed its customer list, but the scale of that projection implies meaningful contracts already in place with labs hungry for motion data.

A category forming in real time

The Sequoia-led round does not exist in a vacuum. It is the second nine-figure-plus valuation in the robot-data category in a single week. TechCrunch reported on September 4 that XDOF — a direct competitor that is just three months out of stealth — is in talks for a Series B at a $1.2 billion valuation, after claiming 20 customers including several frontier AI labs and a partnership with UC Berkeley to release a massive open robot-training dataset. Beyond the specialists, human-data platforms that grew up serving LLM developers, such as Scale AI and Micro1, are expanding into physical data as well.

The rush reflects a structural shift in where AI’s constraints live. Model architectures for physical AI are increasingly available — several labs have open-sourced vision-language-action models — and compute, while expensive, is procurable. What cannot be synthesized is authentic human motion at scale. Simulated data has improved dramatically, but sim-to-real transfer remains the graveyard where many promising robot policies go to die. Real egocentric recordings of messy, unstructured human activity are the asset that cannot be faked.

Seen through that lens, a $500 million valuation on a startup projecting a $100 million run rate — roughly five times forward revenue, if the projection holds — is aggressive but not irrational by 2026 standards, especially for a category-defining data supplier with Sequoia’s backing.

The risks behind the multiple

There are, of course, real questions. A valuation negotiated at roughly eight times the total capital Mecka has raised to date assumes the category winner-take-most dynamics of LLM data suppliers will repeat in robotics. That is not guaranteed. incumbents like Scale AI have deeper enterprise relationships and could bundle physical data into existing contracts. Robotics labs may verticalize, building in-house capture programs once they understand the unit economics — the same way many LLM developers ultimately internalized data operations that they once outsourced.

There is also the data-supply chain’s perennial fragility: the quality and consent of the humans in the loop. Mecka’s model depends on a distributed workforce wearing sensors and filming themselves for pay. Scaling that workforce while maintaining data quality, diversity of tasks, and defensible labor practices is operationally brutal — precisely the kind of unglamorous execution problem that venture capital tends to underprice.

And the terms of the Sequoia deal, as TechCrunch noted, are not final. Deals coming together at this speed in a category this hot can come apart, or be restructured, just as quickly.

Why it matters anyway

Even with those caveats, the direction of travel is unmistakable. The most sophisticated capital in the industry — Sequoia, and Framework Ventures, Menlo Ventures, SV Angel, and Kindred Ventures before it — is converging on the judgment that embodied AI’s scaling curve will be gated by physical data, and that the companies which own the capture pipeline will tax the entire robotics stack the way data suppliers taxed the LLM stack.

For the robotics industry, a well-funded Mecka means more and better human-motion data flowing to whoever can pay for it — likely accelerating humanoid development in the process. For everyone else, it is one more reminder that in this AI cycle, the shovel sellers keep getting richer faster than the gold miners. This time, the shovels are worn on the body.