AiGpu

Industry·

Humyn Labs Secures $47M to Supply Real‑World Training Data for U.S. Robotics AI

Bengaluru‑based Humyn Labs closed a $47 million Series A — $26 million equity led by Accel and $21 million venture debt — to scale its global data‑collection platform that feeds U.S. robotics and world‑model developers with labeled, real‑world footage.

Humyn Labs two‑camera data capture kit in a home setting

Humyn Labs, founded in February by former Nazara Technologies CEO Manish Agarwal and ex‑AB InBev executive Ishank Gupta, has built a two‑camera hardware kit and proprietary software that records people performing everyday tasks across more than 20 countries. The footage is automatically annotated — tracking hand movements, object interactions, and scene context — to create high‑quality training sets for physical AI.

The startup reports 12 paying customers, primarily U.S. frontier AI labs and robotics firms, generating $10 million in revenue between April and October with an additional $20 million in signed contracts and another $20 million in pipeline. To bridge the 30‑ to 60‑day payment cycle from customers while paying contributors weekly, Humyn added a $21 million venture‑debt facility led by Lighthouse Canton Growth Debt Fund II.

Why it matters for GPU / AI infrastructure

Large‑scale, diverse, and accurately labeled real‑world data is the bottleneck for training embodied AI models. As demand for such datasets surges, the compute required to ingest, augment, and train on them grows proportionally, driving higher utilization of GPU clusters. Humyn’s ability to deliver petabytes of curated data directly fuels the next generation of robotics workloads on cloud GPU platforms.

  • aigpu
  • ai gpu
  • ai gpu cloud
  • aigpu dubai
  • robotics
  • training-data
  • india
  • venture-funding

By AiGpu Editorial · Editorial rewrite based on public reporting (TechCrunch AI)

← All articles