Mecka AI Targets $500 Million Valuation in New Funding Round

Mecka AI is approaching a new financing round led by Sequoia Capital at a valuation of approximately $500 million. The deal follows a $60 million raise just three months earlier, highlighting the rapid growth of companies focused on gathering physical-world data to train humanoid robots.

The startup, co-founded in 2024 by former fintech and crypto executives, aims to solve a critical bottleneck in robotics: the lack of real-world interaction data. By paying individuals to record everyday tasks like making coffee or fixing cars, Mecka generates the specific type of movement data that general-purpose robots require to learn functional skills.

Founders identify data gap in robotics

Josh Gao, Mogen Cheng, Jason Chong, and Duy Nguyen did not come from robotics backgrounds. Instead, they observed that while artificial intelligence models for language have ample training data, physical robots lack similar datasets. They recognized that capturing human actions in real environments was the primary missing piece for developing versatile automation.

The company’s approach mirrors how data companies like Scale AI supported large language model development. However, Mecka focuses on egocentric video and sensor data from the wearer’s perspective. This method allows algorithms to understand how humans manipulate objects and navigate spaces, which is essential for robots to perform complex manual tasks safely and effectively.

Market competition intensifies in data collection

The push for this type of data is becoming a competitive race. Other startups, such as XDOF, are also securing high valuations while expanding their data collection networks. Established players in the AI data space are expanding their services to include physical world interactions, recognizing that text-only training is insufficient for embodied intelligence.

According to reporting by GN technics/ai (en-US), Mecka projected an annual run rate of $100 million by the end of 2026. This financial trajectory suggests that demand for high-quality motion data is growing rapidly among robotics firms and AI labs seeking to build more capable autonomous systems.

Trade-offs in human data collection methods

While this model accelerates data availability, it relies heavily on human labor and sensor consistency. The quality of the training data depends on the diversity and precision of the tasks performed by the participants. There is an inherent trade-off between the speed of data acquisition and the rigorous standardization required for reliable robot behavior.

Additionally, the reliance on third-party individuals to generate data introduces variables in privacy and data ownership that differ from traditional software development. As the valuation climbs, the company must balance rapid scaling with maintaining the high fidelity of the motion data that its clients depend on for building trustworthy robotic systems.