Cortex AI
Startup collecting real-world robot and human data to train embodied AI models.
On the record
Checkpoint: Cortex Marketplace public launch and first exclusivity test
The call
Cortex AI has one real frontier-lab win with Ai2 but faces richer rivals shipping similar products and a synthetic-data threat that could shrink the whole category.
Cortex AI collects and sells data to train artificial intelligence systems that control robots in real workplaces. The company's main potential edge is a marketplace connecting workplaces with robot companies, built by a founder who previously ran a successful two-sided marketplace. However, larger competitors like Scale AI (backed by Meta with a $15 billion valuation) and Encord (which raised $158 million) are already shipping similar products, and nothing stops them from copying Cortex's approach. Two things could upend the picture: first, whether Cortex can keep exclusive rights to hard-to-find training data collected from real robot deployments (which would create a lasting advantage), or whether customers own all that data instead (which would make Cortex just another contractor); second, whether the founder can focus fully on Cortex while also running a separate health care startup, since beating well-funded rivals requires serious attention and capital.
Why
01
AI pressure: Genesis AI (sim-to-real synthetic data). If synthetic sim-to-real data closes the real-world gap, demand for Cortex's hand-collected data shrinks fast. Could AI do the core job? Yes, likely.
02
Bigger players: Scale AI (post-Meta, ~49% stake) and Encord (~$158M raised). Encord already ships a dedicated Physical AI data suite; Scale AI is active in robot demonstration data with far more capital. Could a giant copy it? Yes, easily.
03
What protects it. A workplace-hosted data marketplace the founder plans to build, echoing his prior two-sided marketplace exit at Carousell, but not yet public.
What would change this call
Cortex Marketplace publicly launches with a workplace-hosted data-collection model. Would show Cortex is building its only credible moat before rivals copy it.
Cortex lands 2-3 more named frontier-lab logos beyond Ai2 within 12 months. Would show the Ai2/MolmoAct2 deal is repeatable, not a one-off.
Encord or Scale AI ships a comparable workplace-hosted marketplace first. Would show a better-funded rival closed Cortex's distribution window.
Cortex lands 2-3 more named frontier-lab logos beyond Ai2 within 12 months. Would show the Ai2/MolmoAct2 deal is repeatable, not a one-off.
Encord or Scale AI ships a comparable workplace-hosted marketplace first. Would show a better-funded rival closed Cortex's distribution window.
Key numbers
- Total funding reported $6M to $12.5M depending on source
- Encord raised roughly $158M for physical AI data tools
- MolmoAct2 dataset: 700 hours of bimanual teleoperation data
Named risks
- Sim-to-real synthetic data could close the real-world data gap and reduce demand for Cortex's core product.
- Frontier labs (Physical Intelligence, Figure AI, DeepMind) insourcing their own data collection.
- Cortex Marketplace not yet publicly launched as of October 2025, no exclusivity secured over workplace data.
- Founder Lucas Ngoo concurrently operating a second startup, CareSense, as of October 2025.
- Funding and traction figures across trackers conflict, no verified revenue or valuation disclosed.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Encord | gaining on | Shipped dedicated Physical AI data suite, vendor-claimed $158.1M raised, as of May 2026 |
| Scale AI | contesting | Post Meta's ~$15B, 49% stake deal (June 2025), reorganizing under new leadership near same data niche |
| Genesis AI | gaining on | Sim-to-real synthetic data approach could reduce need for real-world data Cortex sells, noted March 2026 |
| Skild AI | contesting | Well-funded robot-foundation-model rival competing for same robotics-lab customers as of 2026 |
| Physical Intelligence, Figure AI, Google DeepMind | contesting | Frontier labs run substantial in-house data collection/teleop fleets, reducing reliance on outside vendors |
Questions people ask
- Is Cortex AI profitable in 2026?
- No revenue, ARR, or customer counts have been publicly disclosed for Cortex AI as of mid-2026. The San Francisco startup, founded 2025 out of YC F25, has one confirmed engagement (Ai2's MolmoAct2, May 2026) but no disclosed commercial revenue figures.
- Could AI replace Cortex AI's business model?
- Possibly. Cortex AI sells real-world egocentric and robot data, but sim-to-real synthetic data (e.g., Genesis AI) could reduce demand for real-world teleoperated data collection, and automated AI annotation tools could compress its curation layer, both flagged as risks as of March-May 2026.
- Who is Cortex AI's biggest competitor?
- Encord and Scale AI are the closest named rivals. Encord has shipped a dedicated Physical AI data suite (vendor-claimed $158.1M raised, as of May 2026), while Scale AI operates post Meta's roughly $15 billion investment for a 49% non-voting stake in June 2025.
- Has Cortex AI worked with any major AI labs?
- Yes. The Allen Institute for AI (Ai2) credited Cortex AI with helping curate the 700-hour MolmoAct2-BimanualYAM dataset and conducting an independent benchmark evaluation, published alongside Ai2's MolmoAct2 model release on May 4, 2026.
- How much funding has Cortex AI raised?
- Cortex AI has not confirmed an official total. Multiple third-party data trackers report conflicting figures in the single-digit to low-double-digit millions, with a seed round reportedly closing around December 2025; no verified figure or valuation has been officially disclosed by the company.
- Does Cortex AI have a real competitive moat?
- Not yet confirmed. Its planned 'Cortex Marketplace,' which would pay workplaces to host robot data collection, was described as not yet publicly launched as of October 2025, and no exclusivity or proprietary dataset scale has been independently verified.
Timeline
- 2026-05-04Ai2 releases MolmoAct2 model; credits Cortex AI for dataset curation and benchmark evaluation
- 2025-12Cortex AI reportedly closes seed funding round
- 2025-11-08Cortex AI publishes YC Launch post detailing product and marketplace model
- 2025-10Cortex AI exits stealth as part of Y Combinator Fall 2025 batch
- 2025-06Meta invests roughly $15 billion for a 49% non-voting stake in rival Scale AI
Revision history
26 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
- Products, Competitors, Financials, Employees, Headquarters Locations (2025-10-09)
- Cortex AI: Large-scale real-world robot & human data for embodied AI
- Cortex AI - Headcount, Funding & Open Jobs (2026-06-21)
- Valuation, Funding Rounds & Stock Price (2026-05-15)
- MolmoAct 2: An open foundation for robots that work in the real world (2026-05-04)
- MolmoAct2 Action Reasoning Models for Real-World Deployment (2026-05-02)
- Robot Demonstration Data Providers
- Why AI Teams Choose Encord (2026-05-19)
- Cortex AI Alternatives & Competitors (2026)
- 20 Physical AI Companies to Watch in 2026 (2026-03-13)
- Top Robotics Foundation Model & Embodied AI Companies 2026 (2026-05-29)
- Top Data Labeling Platforms for Physical AI in 2026 (2026-05-28)