Humyn Labs raises $47M to teach robots from real homes
Humyn Labs, founded this year, raised $47 million in equity and debt to collect human motion data for robots in Brazil, South Korea and India.

Humyn Labs, a startup founded in early 2026, has raised $47 million to supply training data for robots and physical AI. Accel led the $26 million equity part of the round, with Prosus Ventures, Jump Trading and others joining. The other $21 million is venture debt, led by Lighthouse Canton with Trifecta Capital.
The company films people doing ordinary household tasks with head-mounted cameras and says it has built a corpus of up to 200,000 hours of this kind of motion, gathered in more than 20 countries. The idea is that a robot trained only on U.S. homes may struggle in an Indian kitchen, so the data has to look like the places robots will actually work. Humyn also sells face and speech datasets. The new money goes to deployment-testing sites in Brazil, South Korea and India and to offices in San Francisco, London and Seoul.
Why it matters
Robot makers keep saying data, not hardware, is what holds humanoids back, and a small industry is forming to sell it. Humyn's pitch is diversity: real homes in many countries instead of staged lab demos. What it has not shown yet is proof that its footage makes a robot measurably better at a task, or which robot companies are paying for it.