Mecka AI Raises $60M Series A to Build the Data Layer Behind Physical AI

Mecka AI, a New York-based physical AI infrastructure company, has raised $60M across a $25M Series A and a $35M follow-on round, both led by Framework Ventures, with participation from Menlo Ventures, SV Angel, Kindred Ventures, and Ted Xiao. Mecka AI builds robotics training data infrastructure for physical AI systems. Rather than building robots directly, the company focuses on collecting, structuring, and deploying large-scale human movement data used to train robotics and embodied AI systems.

The company was founded by Josh Gao (CEO), Jason Chong (CTO), Mogen Cheng, and Duy Nguyen (COO). Their thesis is simple: the next wave of AI will not be constrained by model intelligence alone. It will be constrained by access to high-quality real-world data. The funding reflects a broader shift taking place across the AI market, where investors increasingly believe the winners in robotics may not be the companies building robots themselves, but the infrastructure companies supplying the data required to train them.

What Happened

Every technology cycle develops its own invisible economy. The internet needed hosting providers before it needed SaaS giants. Cloud computing needed infrastructure before it needed enterprise software empires. Generative AI needed datasets before it needed chatbots. Physical AI appears to be following the same pattern.

Mecka AI announced $60M in new financing to expand its platform for collecting, evaluating, and deploying human movement data used by robotics and embodied AI companies. Framework Ventures led both financings, while Menlo Ventures, SV Angel, Kindred Ventures, and Ted Xiao participated in the round. The company's approach centers on capturing real-world human activity through custom body sensors and iPhones, turning those demonstrations into training data for robotics systems learning how to navigate physical environments and perform tasks that humans execute instinctively.

That sounds simple until you realize how difficult it is. Large language models can learn from trillions of words scraped from the internet. Robots cannot learn how to stock shelves, organize inventory, manipulate objects, or navigate dynamic environments from a Wikipedia page. Physical AI requires experience, and that experience has to come from somewhere.

Why This Matters

One of the stranger characteristics of the current AI market is how much attention flows toward models while data remains largely invisible. Yet every major AI breakthrough eventually collides with the same reality: models are only as useful as the information used to train them. For robotics companies, that challenge becomes exponentially harder because the physical world is messy, unpredictable, and filled with countless edge cases.

Humans operate in environments full of interruptions, changing conditions, and contextual decisions that rarely fit neatly into a dataset. Capturing that complexity at scale is becoming one of the most valuable problems in AI. Mecka AI's thesis is that robotics companies need access to vast libraries of real-world human behavior before they can achieve meaningful deployment at scale.

According to company-linked research collaborations, EgoVerse, Mecka AI's flagship dataset, contains more than 64,000 human demonstration episodes developed alongside Georgia Tech, Stanford, Meta Reality Labs, and Scale AI. Those partnerships matter because they highlight an emerging reality within embodied AI: data quality is becoming a competitive advantage.

Market Context

The robotics market has entered a new phase. For years, investors focused primarily on hardware. Then attention shifted toward foundation models capable of controlling robots. Now the conversation is increasingly moving toward training infrastructure. That shift mirrors what happened in AI infrastructure, where infrastructure providers often became some of the most valuable companies in the ecosystem because they sat underneath entire markets rather than competing within a single application category.

Mecka AI operates within the emerging physical AI infrastructure category, which includes robotics training platforms, embodied AI systems, simulation environments, and deployment tooling. As physical AI moves closer to commercial deployment, the infrastructure supporting data collection and model training is becoming strategically important.

According to investor commentary associated with the financing, the company is projecting approximately $100M ARR based on signed contracts. While those figures remain company-reported projections rather than independently audited financial results, they help explain investor enthusiasm around the category. More importantly, they suggest that demand for robotics training data may already be moving from research budgets into commercial spending, which is often the point where a market begins to mature.

Competitive Landscape

Mecka AI operates within a rapidly expanding physical AI ecosystem that includes robotics developers, foundation model companies, simulation platforms, and infrastructure providers. The company is frequently compared to the role that Scale AI played in the large language model ecosystem. The comparison is not perfect, but it highlights how investors increasingly view data infrastructure as a standalone category rather than a supporting service.

The broader competitive landscape includes companies focused on embodied AI, robotics foundation models, synthetic data generation, and physical-world training systems. The distinction for Mecka AI is its emphasis on real human demonstration data collected at scale rather than relying primarily on simulation or synthetic environments.

That focus reflects a growing belief across robotics that synthetic environments alone may not capture enough real-world complexity to train highly capable physical systems. Whether that proves true remains one of the most important questions in robotics over the next decade.

What This Signals

The most interesting part of this announcement is not the funding amount. It's where the money is going. The market is beginning to fund the infrastructure behind physical AI with the same conviction previously reserved for foundation models and cloud infrastructure. That shift suggests investors increasingly view robotics as a data problem as much as a hardware problem.

Josh Gao, Jason Chong, Mogen Cheng, and Duy Nguyen are effectively making a wager that whoever controls the highest-quality physical-world training data may control one of the most valuable positions in the robotics stack. Investors appear willing to fund that thesis.

The Bigger Industry Shift

Every major technology wave eventually discovers that intelligence alone is not enough. The internet needed networks. Cloud computing needed data centers. Generative AI needed datasets. Physical AI appears to need something different: a scalable system for capturing how humans interact with the real world.

That is where Mecka AI is placing its bet. While much of the robotics industry remains focused on what robots will eventually do, Mecka AI is focused on how robots learn to do it. The distinction sounds subtle, but history suggests it may be enormously important.

Frequently Asked Questions

What is Mecka AI?

Mecka AI is a New York-based company that builds robotics training data and deployment infrastructure for physical AI and embodied AI systems.

How much funding did Mecka AI raise?

Mecka AI raised $60M, consisting of a $25M Series A and a $35M follow-on financing.

Who led Mecka AI's funding round?

Framework Ventures led both the $25M Series A and $35M follow-on round, with participation from Menlo Ventures, SV Angel, Kindred Ventures, and Ted Xiao.

Who founded Mecka AI?

Mecka AI was founded by Josh Gao (CEO), Jason Chong (CTO), Mogen Cheng (Co-Founder), and Duy Nguyen (COO).

What is EgoVerse?

EgoVerse is Mecka AI's large-scale human demonstration dataset containing more than 64,000 episodes used to train robotics and embodied AI systems.

What does Mecka AI do?

Mecka AI collects and structures human movement data that robotics companies use to train physical AI systems and deploy embodied AI applications.

Why is robotics training data important?

Robotics systems require real-world behavioral data to learn navigation, manipulation, decision-making, and physical task execution in complex environments.

What is physical AI?

Physical AI refers to AI systems that operate in the real world through robots, autonomous machines, embodied agents, and intelligent physical systems.