Mecka AI, a startup founded in 2024 that collects and analyzes human motion data to train humanoid and other robots, announced on Wednesday, October 7, 2026, a $60 million Series B led by Sequoia Capital, according to TechCrunch and BetaKit. Participants included Nvidia, Microsoft’s venture arm M12, Qualcomm Ventures, Samsung and existing investors Kindred Ventures, Framework Ventures and Neo, underscoring broad corporate interest in the physical-AI data stack.
TechCrunch had reported the previous month that Mecka was nearing a new Sequoia-led round at a valuation of about $500 million. The company did not publicly confirm that figure in the October 7 announcement, so the valuation should be treated as a press-reported number, not an official disclosure. BetaKit adds that DoorDash co-founder and CEO Tony Xu, former ServiceNow and Snowflake CEO Frank Slootman, and Milan Kovac, the former head of Tesla Optimus, also participated.
Mecka’s thesis is to do for robotics what Scale AI, Mercor, Surge and other data-labeling companies have done for large language models: supply the human-generated data those systems learn from. The startup pays people to record themselves doing everyday tasks — making coffee, fixing cars, working in workshops or labs — while wearing body sensors and using smartphones. That material, unlike text or images that can be scraped from the open web, does not exist at scale online.
According to BetaKit, Mecka claims it surpassed a $100 million annualized revenue run-rate in June and expects to hit $300 million by the end of 2026. Those figures come from the company and have not been audited in the cited coverage. Series B capital will help, Mecka says, scale its data infrastructure, deepen its internal research lab and build out its commercial robot-deployment layer.
Mecka is U.S.-domiciled, with a presence in Toronto and New York. Three of its four co-founders are Canadian and the vast majority of staff are in Toronto, according to BetaKit. Months earlier, the company acquired Docula, a small Vancouver startup, and announced $60 million in Series A financing split between an initial $25 million round last November and a $35 million extension closed over the summer, backed by Framework, Menlo Ventures, SV Angel, Kindred and others.
Competition in real-world robot data includes XDOF, which according to prior TechCrunch reporting was in talks for a Series B at a $1.2 billion valuation, and human-data platforms that began with language models and are expanding into robotics, such as Scale AI and Micro1. Interest from Sequoia, Nvidia and M12 suggests investors see the motion-data bottleneck as strategic as chips or power.
Nvidia, in particular, has a clear incentive: the more robots are trained and deployed, the greater the demand for its accelerators and robotics software stack. Microsoft, through M12, and Samsung and Qualcomm Ventures also bring potential channels into industrial and consumer-electronics customers. The presence of DoorDash, Snowflake/ServiceNow and Tesla Optimus executives as angel or participating investors reinforces the reading that commercial deployment, not only research, sits at the center of the thesis.
Mecka’s capture mode — first-person perspective with sensors and iPhones in home, culinary, chemistry-lab, metal-fabrication and leather-shop settings — aims for data that multimodal models can turn into control policies for arms and humanoids. The company has built sensor-rich hardware and computer-vision and multimodal models to convert those recordings into training-ready material, according to BetaKit.
The rise of humanoid robotics and “physical AI” has driven demand for exactly this kind of data. Language models can pretrain on massive volumes of public text; robots need demonstrations of how a human opens a drawer, pours a liquid or tightens a screw. Without that corpus, reinforcement learning in simulation or trial-and-error in the real world is slow and expensive. Startups such as Mecka sell the promise of shortening that cycle.
Risks remain. Demonstration quality and diversity matter as much as volume; geographic or task bias can limit generalization across factories, homes and cultures. The costs of recruiting, instrumenting and paying thousands of people who record real tasks are high and operationally complex. And competition from Scale AI, Micro1 and XDOF could compress margins if frontier robotics labs diversify suppliers or build capture capacity in-house. The reported valuation of about $500 million, if confirmed, already prices in aggressive growth.
Canada appears as an unexpected but coherent node: Toronto concentrates computer-vision and machine-learning talent, and the Canadian Robotics Council has recorded a rapid increase in robotics startups. Mecka, though U.S.-domiciled, benefits from that base. The fact that nearly half of Canadian venture dollars last year went to artificial intelligence, according to BetaKit’s account, fits the appetite for embodied AI.
For the market, Mecka’s Series B is another signal that the robotics supply chain is professionalizing by layer: chips, power, motion data, control models and factory or home deployment. Whoever controls human demonstration data can influence which labs train faster and which robots reach commercial reliability first. Sequoia, Nvidia and M12 are betting that layer will prove as indispensable as text-and-image labeling was for large language models over the past several years.
For now, the concrete fact is a $60 million round announced on October 7, with Sequoia in the lead and a roster of corporates and specialist investors. The $500 million valuation remains a prior press figure rather than a company confirmation. If Mecka hits the revenue run-rate it projects by year-end, the next conversation will be whether that data layer can sustain margins at scale or becomes a contested commodity among several suppliers racing to instrument everyday human work.
