India’s Humyn Labs has raised $47 million in Series A financing to supply real-world human activity data to U.S. AI and robotics companies training models to operate in the physical world, TechCrunch reported November 18. Why it matters: the bottleneck for household and workplace robots is increasingly not the model but the footage — and a new industry is forming to record everyday human tasks at scale.
November 18, 2026
What happened
The $47 million round combines $26 million in equity led by Accel, with participation from Prosus Ventures, BeeGlobal, Bajaj Finance’s Finserv Intelligence and Bajaj Alts AI Opportunities Fund, Dallas Venture Capital, Lighthouse Canton and Jump Trading, plus $21 million in venture debt led by Lighthouse Canton’s Growth Debt Fund II with participation from Trifecta Capital. Bengaluru-based Humyn Labs was founded in February by Manish Agarwal, the former CEO of Indian gaming company Nazara Technologies, and Ishank Gupta, a former AB InBev executive. The pair previously founded Kratos Studios, now known as KGeN, whose contributor network across emerging markets helps Humyn collect data.
How the data business works
Humyn built its own two-camera hardware to record people performing everyday tasks, then uses software to process and label the footage — tracking hand movements, interactions with objects and other details so AI models can learn from it. Collection runs across more than 20 countries in Latin America and Asia, with Brazil, India and South Korea as the main hubs. Agarwal told TechCrunch the company records people doing their usual work or household chores rather than staged activities, working with local partners who identify collection sites and oversee the process.
The company says it has 12 paying customers, including frontier AI labs, robotics companies and developers of world models, with about 90% of revenue coming from the U.S. It generated $10 million in revenue between April and October, has another $20 million in signed contracts, and a further $20 million in deals not yet finalized. Humyn has 51 full-time employees, most based in Bengaluru, plus its local partner network.
Why debt is half the round
The $21 million venture debt component is a working-capital story. Humyn pays its workers and data-collection partners daily or weekly, but typically waits 30 to 60 days to be paid by its AI lab customers. Borrowing bridges that gap and is less dilutive than selling more equity, Agarwal said. The structure is a useful reminder that AI training-data businesses are operations-heavy: the costs are human and immediate, while the revenue arrives on enterprise payment terms.
Why it matters
Physical AI is creating a data economy that looks less like web scraping and more like a global filming operation, with real questions attached — TechCrunch notes the practice of recording inside people’s homes has raised privacy and consent concerns, and India has emerged as a hub for this collection work. Humyn is not alone: startups including Mecka AI, XDOF and MicroAGI are chasing the same demand. For readers, the takeaway is that the next constraint on capable robots may be mundane — who owns enough well-labeled footage of humans folding, cooking and fixing things, and on what terms it was recorded. Everything AI, tested applies to datasets too: scale, diversity and quality have to work together, as Agarwal himself put it.
Frequently asked questions
What does Humyn Labs sell?
Labeled real-world data: footage of people performing everyday tasks, recorded on Humyn’s own two-camera hardware and processed to track hands and objects, sold to AI labs, robotics companies and world-model developers.
How much did Humyn Labs raise?
$47 million in Series A financing — $26 million in equity led by Accel and $21 million in venture debt led by Lighthouse Canton’s Growth Debt Fund II.
Who are Humyn’s customers?
The company says it has 12 paying customers, including frontier AI labs and robotics firms, and that about 90% of its revenue comes from the United States. It has not named specific customers.
Sources: TechCrunch

