Mecka AI raises $60M Series A to train robots on human-motion data
What's the deal? New York-based Mecka AI has raised $60 million across two previously unannounced fundraises to build a business around collecting human physical data — hand gestures, walking gaits, body movements — and using it to train robots. The startup closed US$16.2M Series A in November, followed by US$22.6M follow-on investment.
Framework VenturesDealroom has a profile for this one. Try Dealroom →, a VC firm best known for crypto investments, led both rounds. Menlo Ventures, SV Angel, Kindred VenturesDealroom has a profile for this one. Try Dealroom →, and angel investor Ted Xiao — a former Google DeepMind researcher and founding member of Jeff Bezos's AI startup Project Prometheus — also participated.
Mecka AI collects data from humans using custom body sensors and iPhones, then supplies it to robotics companies. The startup projects US$64.7M annual run rate based on signed contracts, though it declined to name customers or disclose its valuation.
Why now? The robotics industry is hitting an inflection point. "Many folks are now looking and realising that robotics is going to be one of the most important waves that's happening," said cofounder and chief executive officer Josh Gao.
Mecka AI's core bet is unconventional: train robots on data captured from everyday human movement rather than through "teleoperation," where humans manually control a robot. That approach had mostly lived in research papers until Gao's team partnered with robotics labs to prove it could work at scale. "Nobody really popularly believed this would be possible," he said.
The four cofounders — Gao, Mogen Cheng, Jason Chong, and Duy Nguyen — come from fintech and e-commerce backgrounds, not prestigious AI labs. Gao and Cheng previously sold a restaurant payments startup; Chong sold his company to Coinbase. They spent months studying robotics research before founding Mecka AI in 2025.
What could go wrong? The company faces real risks. It hasn't disclosed any customers, making its $100 million run-rate projection hard to verify independently. Its founders lack deep robotics pedigrees, and the approach of training robots on human-sourced data rather than teleoperation remains unproven at commercial scale.
Competition is building fast. Wayve trains autonomous driving models on car camera footage, and MicroAGI recently began recording human cleaners to gather embodied AI data. If human-motion data becomes a commodity, Mecka AI's edge could erode quickly.
The signal: Framework Ventures leading both rounds marks a notable pivot for a firm rooted in crypto, underscoring how investor capital is migrating toward physical-world AI infrastructure. With Mecka AI still at the early growth stage, the $60 million bet signals that backers see the supply of high-quality embodied data — not model architecture or hardware — as the decisive competitive moat in the next phase of robotics.
Read more: fortune.com
