Accel leads $47 million round for India's Humyn Labs, which films everyday tasks to train US robots

Startups & Funding

Accel Leads $47 Million Round for Humyn Labs' Robot Training Data

Bengaluru's Humyn Labs raised $47 million to film everyday work for robot training: $26M equity led by Accel, $21M debt, with $10M revenue and a $40M pipeline.

By
Rebecca Stone
Filed
Channel
Startups & Funding
Read
3 min read

Humyn Labs has raised $47 million to film people doing ordinary work — and sell that footage to the AI labs teaching robots to copy them. The Bengaluru startup's Series A, reported by TechCrunch, splits into $26 million of equity led by Accel and $21 million of venture debt led by Lighthouse Canton Growth Debt Fund II with Trifecta Capital.

The equity side also drew Prosus Ventures, BeeGlobal, Jump Trading, Dallas Venture Capital, Lighthouse Canton and two Bajaj funds. The debt side, unusually large for a Series A, exists to solve a cash-flow gap that defines the physical-AI data business: Humyn pays its footage collectors and local partners daily or weekly, while the AI labs on the other end take 30 to 60 days to pay their invoices.

The company's product is deceptively simple. Humyn built its own two-camera rig that records people at their actual jobs and household chores — not staged demonstrations. Its software then labels the footage, tracking hand movements and how people grasp, move and manipulate objects. That's the raw material robotics companies and AI labs need to train machines in real-world manipulation.

Where does the footage come from?

Humyn collects in more than 20 countries across Latin America and Asia, with Brazil, India and South Korea as its main hubs. It works through local partners and the contributor network of KGeN, a company its founders started before Humyn. The geographic spread matters commercially: it gives US customers variation in environments, tools and techniques that a single-country dataset can't provide.

The customer base is small but concentrated where the money is. Humyn says it has 12 paying customers — frontier AI labs, robotics companies and developers of world models — though it declined to name any of them. About 90% of its revenue comes from the US.

What do the numbers say?

Co-founder Manish Agarwal told TechCrunch the company booked $10 million in revenue between April and October. On top of that sit $20 million in signed contracts and another $20 million in deals not yet finalized. For a 51-person company, that pipeline suggests the round is about scaling collection capacity rather than proving demand.

Agarwal knows scaling businesses. He was previously CEO of Nazara Technologies, the Indian gaming company. He co-founded Humyn with Ishank Gupta, a former AB InBev executive.

Filming people inside their homes raises the privacy and consent questions now hovering over the entire physical-AI data sector. Agarwal's defense is straightforward: the workers are paid and, in his words, "monitored and supervised for the right input."

Why does debt anchor this round?

The $21 million debt tranche is the structural story here. Data-collection businesses run on working capital — pay collectors now, collect from labs later. Equity investors often resist funding that gap with permanent capital; venture debt matches the timing mismatch instead. Lighthouse Canton leading both the equity-adjacent syndicate and the debt fund signals lender conviction in Humyn's contracted revenue rather than speculative growth.

The broader context: as US frontier labs race to build world models and humanoid-adjacent robotics, demand for real human-motion data has outpaced synthetic alternatives. Companies that can capture, label and ship that data at scale — and finance the float between collection and payment — are becoming picks-and-shovels suppliers to the physical-AI boom.

With $40 million in signed and pending contracts already exceeding its trailing seven-month revenue fourfold, Humyn's next test is whether its 20-country collection network can scale fast enough to convert that pipeline into booked revenue.

Source: mlq.ai

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering technology and product news in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

1,112 articles

Related articles

« Previous article