Human Archive
Multimodal data provider for robots: video, tactile, IMU capture for world models.
On the record
Checkpoint: 12-month window to land 2+ paid multi-hundred-K contracts with named frontier humanoid labs
The call
Human Archive has real capture infrastructure but no disclosed paying customers, while Scale AI already ships a rival product in the same market.
Human Archive pays workers in multiple countries to perform tasks while wearing cameras and sensors, then sells that video and movement data to AI companies training robots. The company has real equipment and workers, but Scale AI launched a competing product in March 2026 with much more money and partnerships, making the gap between them look closeable in about 18 months if Scale AI keeps investing. The biggest unknown is whether AI labs like OpenAI or Anthropic will actually buy this data from outside companies instead of building their own recording setups with robots and workers, which would change everything; if they do build in-house, Human Archive's advantage disappears fast.
Why
01
AI pressure: Frontier robotics labs building in-house teleop data farms (e.g. NVIDIA GR00T, World Labs sim-to-real). Labs may skip buying data if sim-to-real tools close the tactile gap or in-house teleop farms scale cheaper. Could AI do the core job? Only partly.
02
Bigger players: Scale AI (Physical AI Data Engine, Universal Robots partnership). Scale AI already ships a competing physical-AI data product with a named robotics partner, backed by far more capital. Could a giant copy it? Yes, easily.
03
What protects it. Live network of 1,000+ headsets, custom tactile rigs, and multi-country labor already collecting data today.
What would change this call
Scale AI discloses a named humanoid/robotics lab customer for Physical AI Data Engine within 12 months. Would show the incumbent capturing deals Human Archive needs to prove its wedge.
Human Archive announces 2+ named paid contracts worth multi-hundred-K with frontier labs within 12 months. Would validate that frontier labs buy external data rather than build in-house.
Human Archive's next funding round size (Series A/B) compared to the ~$100M+ moat-build cost. Would show whether investors believe the company can fund the capital gap versus Scale AI.
Human Archive announces 2+ named paid contracts worth multi-hundred-K with frontier labs within 12 months. Would validate that frontier labs buy external data rather than build in-house.
Human Archive's next funding round size (Series A/B) compared to the ~$100M+ moat-build cost. Would show whether investors believe the company can fund the capital gap versus Scale AI.
Key numbers
- $8.2M seed/Series A led by Wing VC and NVP Capital
- 1,000+ active headsets, 50+ device types deployed
- Zero named enterprise customers per third-party reporting
Named risks
- Capital gap: moat estimated at hundreds of millions to replicate vs only $8.7M raised.
- No named paying enterprise customers confirmed as of July 2026.
- India MEITY regulatory scrutiny of worker data consent practices.
- Sim-to-real quality improvements from World Labs, Genesis AI could erode real-data advantage.
- Below-market pay to gig workers (~$1/hr) risks labor supply and reputational backlash.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Scale AI | losing to | Shipped competing Physical AI Data Engine, partnered with Universal Robots by March 2026 |
| World Labs | contesting | Synthetic sim-to-real data could substitute real-world tactile capture, flagged May 26 2026 |
| Genesis AI | contesting | Sim-based contact/tactile data generation threatens HA's real-world-capture wedge as of May 2026 |
| Spatial AI | contesting | Similarly strategized YC F25 peer in robotics data sourcing per Harvard CAP report |
| NVIDIA GR00T / in-house model labs | losing to | Frontier labs may build teleop farms in-house rather than buy external data, unvalidated assumption |
Questions people ask
- Is Human Archive profitable?
- No revenue is publicly disclosed for Human Archive as of May 2026, per CB Insights financial data showing undisclosed figures. The company is under one year old, pre-revenue by public accounts, and monetizes through B2B dataset licensing to robotics labs, but no named paying customer has been confirmed by third parties.
- How much funding has Human Archive raised?
- Human Archive has raised roughly $8.7M total, including an $8.2M seed round announced May 26, 2026 led by Wing Venture Capital and NVP Capital, plus an earlier ~$0.5M pre-seed led by Y Combinator in early 2026, per inforcapital.com and TechCrunch.
- Who is Human Archive's biggest competitor?
- Scale AI is Human Archive's biggest named threat, having shipped a competing Physical AI Data Engine and partnered with Universal Robots in March 2026, per roboticsbiz.com, July 6 2026. Scale AI is far better capitalized and could close Human Archive's hardware/data moat within an estimated 18 months.
- Could AI replace Human Archive's data collection business?
- Possibly. Synthetic-data and world-model rivals like World Labs and Genesis AI could substitute for real-world tactile and contact data if simulation fidelity improves enough, a risk flagged by SiliconAngle on May 26, 2026. Human Archive's edge is its physical hardware fleet, not defensible software or IP.
- Does Human Archive have any confirmed customers?
- No named enterprise customer has been confirmed by third-party reporting as of July 6, 2026, per roboticsbiz.com. The company states its customers are unnamed frontier robotics labs and humanoid foundation model teams, and investor Wing VC claims broad lab interest, but this is a promotional, vendor-adjacent statement.
- Who owns Human Archive?
- Human Archive is privately held and venture-backed with no parent company, founded by Raj Patel (CEO), Shloke Patel, Samay Maini (CTO), and Rushil Agarwal (COO), per CB Insights, June 14 2026. It is part of Y Combinator's W26 batch with no disclosed valuation as of May 25, 2026.
Timeline
- 2026-07-06roboticsbiz.com names Scale AI, Spatial AI, Asimov as closest peers; no named HA customer confirmed
- 2026-06-19Third-party analysis scores Human Archive 71/100 replicability, calls moat capital-and-time
- 2026-06-18zaka.vc reports 1,000+ active headsets, 50+ device types across multiple countries
- 2026-05-30Company confirmed hiring across Shenzhen, San Francisco, Germany, China, India
- 2026-05-26$8.2M seed round announced, led by Wing Venture Capital and NVP Capital
- 2026-05-26TechCrunch reports MEITY scrutiny of egocentric data-collection consent practices in India
- 2026-05-25Company listed with no disclosed valuation and $0/undisclosed revenue on CB Insights
- 2026-03-08Human Archive launches on Y Combinator, claims 8,000 hours of data collected per day
- 2026-03Scale AI ships competing Physical AI Data Engine, partners with Universal Robots
- 2026-01Pre-seed round of ~$0.5M led by Y Combinator (early 2026)
Revision history
27 July 2026
First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- Human Archive - Robotics Startup, $8.7M Raised (2026-05-18)
- This startup is betting India's gig economy can train the world's robots (2026-05-26)
- Human Archive: Building Data Infrastructure for Physical AI (2026-06-18)
- Human Archive
- Claude's Corner: Human Archive, Building Common Crawl for Robot Hands (2026-06-19)
- Best data labeling companies for robotics & physical AI in 2026 (2026-07-06)
- Human Archive Careers, Perks + Culture
- Launch YC: Human Archive: The World's Largest multimodal robotics dataset (2026-03-08)
- Launch YC: Human Archive: Large-Scale multimodal datasets for robotics and world modeling (2026-03-08)
- README.md · humanarchive/HA-Multi-Samples
- AI training data provider Human Archive raises $8.2M (2026-05-26)
- From Bits to Atoms: Physical AI (2026-04-08)