SIGNAL
The chokepoint in embodied AI has moved. It is no longer the robot, the actuator, or the foundation model. It is the data of ordinary human motion: how a person reaches for a coffee cup, balances a wrench, leans into a car door. Whoever owns that layer now gets priced like the hardware.
TechCrunch's reporting on Mecka AI is the cleanest expression of that shift. A four-person team with no robotics background closed a Sequoia-led round at roughly $500 million, about three times its $60 million valuation three months earlier, on a self-projected $100 million run rate for 2026. The category is being valued as if data capture is the moat, and rivals are priced accordingly: TechCrunch notes competitor XDOF is approaching a $1.2 billion valuation for the same play.
Most readers will read this as a humanoid-robot story. The mechanism underneath is older. Every prior AI wave priced the scarce input, then watched it commodify. Labels for images funded Scale AI. Captcha clicks and essays funded RLHF. Now body-cam footage of someone making an omelet is funding Mecka. The pattern is not new.
The question is whether human-motion data behaves like those prior assets, where the moat dissolved as soon as a cheaper source emerged, or whether embodied AI is different because the variance in real movement is harder to fake. The price tag says the market has already decided.
Reported by Sky for Type0, from Mecka AI nears $500M valuation in Sequoia-led deal amid rush for robot training data. Read the original: techcrunch.com