Humyn Labs raises $47M to generate more physical AI data for robots
$47 million for physical AI data collection
On November 18, Humyn Labs announced a $47 million Series A. The goal: generate physical AI data for robot companies.
The lead investor is Accel. Prosus Ventures also joined the equity round, alongside Bajaj Finance, Jump Trading, Dallas Venture Capital, and angel investor Anita Pramoda, the former CFO of Epic. The company was founded less than ten months ago.
Forty-five percent of the total — $21 million — is venture debt, led by Lighthouse Canton, with Trifecta Capital participating.
And that is very interesting ...
The payment timing problem
Humyn Labs recruits and deploys human physical AI data contributors: people who wear proprietary two-camera rigs and perform real-world tasks in factories, warehouses, retail stores, homes, and farms while the system records them.
These contributors are paid daily or weekly.
But the frontier AI labs and robotics companies that buy the resulting datasets pay on 30-to-60-day net terms.
That gap — between the daily cash out and the monthly cash in — is the operational constraint that shapes how many physical AI data company will have to finance themselves, as the market matures. Humyn Labs co-founder Manish Agarwal says they structured the raise deliberately to bridge it without further equity dilution.
The debt is essentially working capital.
What Humyn Labs actually sells
As most in robotics know, robots do not learn a lot from the internet. They are not LLMs.
Where a language model can read every manual ever written about warehouse picking; a robotic arm still needs to watch a person who has picked items in that specific aisle, with that specific bin layout, under those specific lighting conditions, for years. That gap is physical AI.
(Of course, OpenAI is betting frontier-level LLMs won't need physical AI.)
The company describes its mission as "the study of humans for machines."
"A robot does not learn a job from the internet," says co-founder Ishank Gupta. "It learns it from the person who has done it for twenty years, on the floor where it will be done."
Humyn's platform converts raw recordings into 12 structured signal primitives across four sensory modalities including:
- 21-point dexterous hand keypoints
- egocentric gaze tracking
- stereo depth maps
- action segmentation
- scene geometry
- paralinguistic affect detection
Output is delivered in MCAP, RLDS, LeRobot v3, and ROS Bag formats: the native formats of the robotics development stack.
When the company launched, Agarwal says 40 to 60 percent of field-collected physical AI data was being discarded due to quality failures ... which of course increases the cost of the remaining data.
Humyn's validation and labeling pipeline is designed to close that rejection rate.
Revenue and customers, six months in
Humyn Labs launched commercial operations in April 2026. By October, it had generated $10 million in revenue ... incredibly quick growth.
It had signed $20 million in additional contracts and was in advanced discussions on a further $20 million. The company says its annualized run rate crossed $50 million within six months of first revenue.
The customer list has 12 types of clients including frontier AI labs, humanoid robotics companies, world-model developers, and voice AI builders.
Approximately 90 percent of revenue comes from the United States. The company operates across more than 20 countries, with primary collection hubs in Brazil, India, and South Korea, where labor is cheap.
The network behind the network
Agarwal and co-founder Ishank Gupta previously built Kratos Studios, later rebranded as KGeN (Kratos Gamer Network), a contributor-network platform spanning more than 60 countries.
That network — built to distribute gaming tasks to contributors across emerging markets — now provides the operational backbone for Humyn's data collection in private homes and local commercial spaces.
Agarwal previously served as CEO of Nazara Technologies, India's first listed gaming company, from 2015 to 2022. Gupta is a Wharton alumnus and former AB InBev executive. The pairing is unusual for a deep-tech data company: a consumer internet operator and an FMCG logistics executive, both focused on the last-mile challenge of getting humans into the environments where robots need to learn.
The benchmark play
Alongside its data business, Humyn Labs publishes BRIDGE, an independent evaluation benchmark series for physical AI.
BRIDGE ASR 2.0, launched in September 2026, tests 23 voice models across 23 languages on six metrics using real-world conversational audio — including dual-speaker overlap, interruptions, and code-switching — with particular depth in Indic languages, Latin American Spanish, Brazilian Portuguese, and Vietnamese.
The raise in context
"Physical AI is going to be one of the defining technology shifts of the next decade, and data and deployment are where the real moat will be built," says Accel partner Pratik Agarwal. "Humyn Labs has the technical depth, the customer pull, and the ambition to be a category leader."
Humyn Labs competes with Mecka AI, XDOF, MicroAGI, and India-based Human Archive and BuildAI. The funding will go toward scaling collection infrastructure in Brazil, South Korea, and India; opening offices in San Francisco, London, and Seoul; hardware development; and a stated $20 million commitment to building physical AI deployment systems. The company has 51 full-time employees.
What the raise signals, more than its size, is that the physical AI data market is now large enough for a dedicated infrastructure company to reach a $50 million run rate in its first operating year. The debt structure — an unusual but logical response to daily collection costs and monthly customer payments — may be the template others in this category adopt as they scale.
Sources
In this story
Follow to get an alert when they're in the news
Humanoid (almost) Daily
The humanoid robot news that matters, with context from HumanoidMegaHub's tracking. Free, almost every day.
Keep reading
Written by
I tell stories with data and build startups. I've raised angel and VC funds, led software companies and startups, and chronicled the rise of the mobile economy. Currently I write for Forbes, host several podcasts, and am starting up HumanoidMegaHub. I was TUNE's mobile economist and Singular's VP of Insights, built the research team at VentureBeat, and managed teams creating software for partners like Intel and Disney. In past lives I built websites, e-learning platforms, and mobile apps, and, a long time ago, was named to Folio's top 100 of the media industry's "most innovative entrepreneurs and market shaker-uppers.”