Humyn Labs Raises $47 Million to Supply Real-World Training Data for Robotics AI

Humyn Labs, a Bengaluru-based company that collects real-world training data for robotics and AI developers, has raised $47 million in Series A financing, according to TechCrunch. Of that total, $26 million is equity led by Accel. The other $21 million is venture debt, a loan extended to a venture-backed company, led by Lighthouse Canton Growth Debt Fund II, with participation from Trifecta Capital.
The equity round also drew Prosus Ventures, BeeGlobal, Bajaj Finance's Finserv Intelligence, the Bajaj Alts AI Opportunities Fund, Dallas Venture Capital, Lighthouse Canton and Jump Trading.
Humyn Labs was founded in February by Manish Agarwal and Ishank Gupta. Agarwal was formerly CEO of Nazara Technologies, and Gupta was formerly an executive at AB InBev. The two founders had previously started Kratos Studios, now known as KGeN.
The company sells data, not a model. Humyn Labs built its own two-camera rig to record people performing everyday tasks, then uses its software to process and label the footage so that robot-learning teams can train on it. Collection runs across more than 20 countries in Latin America and Asia, with Brazil, India and South Korea as the main hubs.
Demand is concentrated in one market. The company has 12 paying customers, including frontier AI labs, robotics companies and developers of world models, which are AI systems that simulate how physical environments change over time. About 90% of its revenue comes from the U.S.
The company generated $10 million in revenue between April and October. It also reports $20 million in signed contracts and a further $20 million in deals not yet finalized. The two pipeline figures are separate, and only the first is contracted.
The competitive set includes other startups that collect real-world data for robotics training, among them Mecka AI, XDOF and MicroAGI.
The broader context here is a change in what robot learning depends on. Large language models were trained largely on text scraped from the web, a corpus that already existed. Manipulation policies for physical robots, which govern how a robot grips and moves objects, have no comparable pool of data. Teams have therefore commissioned or recorded human demonstrations themselves, and the cost and quality of that collection now sit on the critical path of development. A vendor that can capture, label and deliver such footage at volume becomes part of the development pipeline itself.
In my view, two numbers deserve closer attention than the headline total. The first is the revenue mix. With roughly nine-tenths of revenue from the U.S., Humyn Labs' fortunes are tied to a small number of American buyers, and a single lab's procurement decision could move the business materially. The second is the gap between the $10 million recognized in roughly seven months and the $40 million in contracted and pending work. Whether the unsigned $20 million converts will say more about the company's trajectory than the round size does.
The investor list is also worth reading closely. Accel's lead on the equity side, and Lighthouse Canton's role on both the equity and debt sides, suggest the company is being financed as infrastructure rather than as a research project. Venture debt at this stage usually signals that the company expects revenue to scale faster than its equity can fund, which makes the round a bet on execution as much as on the market.
If the collection model holds, the practical effect would be lower friction for robotics teams that need varied, real-world examples of human movement, along with a wider geographic base for that work than a single-country operation could provide. The open question is quality control at that scale. Labeling pipelines for physical-world data are harder to audit than text, and buyers will judge Humyn Labs on whether its footage improves policy performance, not on how many hours it has recorded.


