Prince Mario-Max Schaumburg-Lippe: Sequoia Pours $60M Into the Toronto Startup Teaching Robots by Watching Humans

Picture a kitchen in Toronto on a Tuesday evening. Someone is pouring coffee, wearing a suit stitched with motion sensors, a phone strapped to their chest recording every move from their own point of view. That cup of coffee is not for drinking. It is homework for a robot.

Mecka, a two-year-old Toronto startup, just closed a $60 million Series B led by Sequoia Capital. Nvidia, Microsoft’s M12 fund, Qualcomm Ventures, and Samsung all joined the round, and TechCrunch reported the company is now valued at $500 million. Not a bad two-year run for founders whom, by their own joking admission, no robotics lab would have hired on credentials.

The fintech founders who started over

CEO Josh Gao and Mogen Cheng sold a restaurant-payments startup in 2023. Jason Chong ran a crypto exchange that Coinbase bought. Duy Nguyen built a business reselling sneakers. After the exits, they did something unusual: they spent months reading robotics papers and touring labs, asking the people inside where the real bottleneck was.

The answer they kept hearing was not hardware. Robot hands, arms, and joints keep improving, but there is no giant archive of movement for those machines to learn from. Text and images could be scraped by the terabyte. The way a wrist turns to open a jar, how much force a fingertip applies before a plate slips, when a hand lets go – none of that lives on the internet. So Mecka decided to go make it.

People, not robots, do the demonstrating

Most robot training data comes from teleoperation: a person remotely steering a robot while it records the session. Mecka flipped the model. It pays ordinary people to record themselves doing ordinary things – making coffee, fixing a car, folding laundry, mixing a drink. Demonstrators wear body sensors and use smartphones to capture tasks from their own perspective, at home and at Mecka’s own facilities.

The company puts it bluntly: motion, contact, force, and geometry are not on the internet. You cannot scrape them, license them, or buy them. Someone has to perform them first. And human demonstrations carry something teleoperation does not: intent. The way a grip shifts on a slippery handle, how a pan gets repositioned when the stove is crowded – those tiny adaptations are exactly the physics nobody has ever written down.

The scale is already serious

Mecka’s EgoVerse dataset holds 1,362 hours of recorded demonstrations across 80,000 episodes, performed by 2,087 unique demonstrators. Its sensors capture joint angles, force, and timing at 200 frames per second, with smartphones providing synchronized video from multiple angles. The company designs its own capture hardware and runs reconstruction models with sub-centimeter hand-tracking accuracy.

The business side is moving just as fast. Mecka says it crossed a $100 million annualized revenue run rate in June and projects $300 million by year end. Its customers are described only as several of the top frontier robotics labs and multiple of the world’s biggest tech companies, who pay for the data and for help putting it to work. Robotics researchers from Stanford, MIT, Georgia Tech, UC San Diego, ETH Zurich, and Meta have already used EgoVerse as a research project.

Everyone is hedging into the data layer

Read the investor list and you see the hardware industry placing its bets. Nvidia, Samsung, Qualcomm, and Microsoft’s M12 all want a piece of the dataset that feeds their robotics stacks. The angels tell the same story: DoorDash CEO Tony Xu, former Snowflake CEO Frank Slootman, and Milan Kovac, who previously ran Tesla’s Optimus humanoid program.

The Financial Times reported that robotics and physical-AI companies have raised nearly $48 billion in 2026, with much of it expected to fund the expensive job of generating training data. Competition is heating up: XDOF is reportedly in Series B talks at a $1.2 billion valuation, and Scale AI and Micro1 are expanding from language-model data into robotics. When everyone is racing to the same supply chain, the supply chain is where the demand lives.

Why this matters

There is a familiar pattern here. Scale AI, Mercor, and Surge built the human-data pipelines that made large language models possible. If language models needed armies of annotators, humanoid robots may need armies of demonstrators. The model labs get the headlines and the chipmakers get the revenue, but the quiet money keeps flowing to the picks-and-shovels layer.

So the question for anyone building robots is no longer whether the hardware is affordable. It is whether you have the thousands of hours of human motion your machine still needs to watch. Somewhere tonight, someone is folding laundry in a sensor suit. That laundry is going to teach a robot how hands work.