Mecka AI Nears $500 Million Valuation as Demand for Robot Training Data Surges

Mecka AI is reportedly nearing a valuation of around $500 million in a new funding round led by Sequoia Capital, highlighting the growing demand for real-world data needed to train humanoid robots and other advanced robotic systems.

The reported financing comes only three months after the two-year-old startup announced a $60 million funding round led by Framework Ventures. The terms of the new deal, including its exact size, have not been finalized and could still change, according to TechCrunch, which cited two people familiar with the deal.

Neither Mecka AI nor Sequoia Capital provided comment on the reported financing.

Mecka AI Is Building a Data Business for Robotics

The company’s strategy is based on a relatively simple idea: robots need enormous amounts of real-world information to learn how to interact with their surroundings.

Large language models benefited from companies that collected, labeled, and organized huge quantities of human-generated data. Mecka AI is attempting to apply a similar model to robotics by collecting information about how people physically interact with the world.

The startup pays people to record themselves performing everyday activities, including tasks such as making coffee and repairing vehicles. The recordings can be captured using smartphones and body sensors, creating what the company describes as egocentric data from a first-person perspective.

That information can then be used to help robotics companies develop systems capable of understanding and performing physical tasks.

The approach addresses one of the major differences between AI software and physical robots. A chatbot can learn from enormous quantities of text and digital information, but a robot needs to understand movement, objects, environments, and the physical consequences of its actions.

Why Physical-World Data Matters for Humanoid Robots

Humanoid robots are being developed to operate in environments designed for people, from factories and warehouses to homes and commercial spaces.

That creates a significant training challenge.

A robot needs to learn much more than what an object looks like. It needs to understand how an object can be picked up, how much force is required to move it, how a person’s body moves while performing a task, and how different environments affect those actions.

Real-world demonstrations can provide information that is difficult to reproduce through purely synthetic environments.

This is where companies such as Mecka AI see an opportunity. By collecting large quantities of human movement and interaction data, the company aims to provide robotics developers with the raw material needed to train increasingly capable machines.

A Startup Moving Quickly

Mecka AI was founded in 2024 by Josh Gao, Mogen Cheng, Jason Chong, and Duy Nguyen.

Interestingly, the founders did not come from traditional robotics backgrounds. Gao and Cheng previously worked together on a restaurant-focused fintech startup, while Chong joined Coinbase after the cryptocurrency exchange acquired his company.

Their experience illustrates how quickly the robotics industry is attracting entrepreneurs from other areas of technology.

The founders identified a potential bottleneck in robotics development: while AI companies have access to massive amounts of digital information, robotics developers have far less physical-world data available for training.

Mecka AI was created around that gap.

The company’s name comes from “mecha,” a term commonly associated with fictional giant robots controlled by humans.

The Robotics Data Market Is Heating Up

Mecka AI is entering an increasingly competitive market.

Other companies are also working to provide data for robotics and AI systems, including XDOF, while established data businesses such as Scale AI and newer platforms are expanding into physical-world data collection.

The broader trend mirrors what happened with generative AI.

As large language models became more capable, the demand for high-quality training data created a major industry around collecting, labeling, evaluating, and organizing information. Robotics could be moving toward a similar model, but with a much harder data-collection problem.

Physical actions are more difficult to capture than text. They also vary enormously depending on the person, environment, object, tools, and task involved.

For robotics companies, that makes large and diverse datasets potentially valuable.

Mecka AI Has Ambitious Growth Targets

The startup’s recent fundraising also suggests that investors see significant commercial potential in the robotics data market.

When Mecka AI announced its previous $60 million financing in June, co-founder Josh Gao told Fortune that the company was projecting an annual run rate of $100 million by the end of 2026.

That would represent rapid growth for a company founded only two years earlier.

The reported Sequoia-led financing, if completed at the rumored valuation, would provide another indication that investors believe the market for robotics training data could become a significant part of the broader AI economy.

However, the financing is not yet final. The reported valuation and other terms could change before the deal closes.

From Digital AI to Physical AI

The rise of companies like Mecka AI reflects a broader shift in artificial intelligence.

The first major wave of generative AI focused largely on digital tasks: generating text, creating images, writing software, analyzing information, and interacting with users.

The next stage increasingly involves physical AI — systems that can perceive and act in the real world.

Humanoid robots are one of the clearest examples.

For these machines to become genuinely useful, they will need to move beyond carefully controlled demonstrations and learn from the enormous variety of situations encountered in everyday life. That requires data representing real human behavior and real physical environments.

Mecka AI’s business is built around supplying part of that missing layer.

The Bigger Challenge for Robotics

Collecting data is only one piece of the robotics puzzle.

Robotics companies still need advances in hardware, computer vision, machine learning, sensors, motion planning, safety, and computing infrastructure. Even a massive dataset does not automatically produce a robot capable of reliably performing a task.

There is also the question of data quality.

Training information must accurately represent the environments and tasks robots will encounter. If datasets are too narrow or lack sufficient diversity, robotic systems may perform well in testing but struggle when conditions change.

That makes the quality, scale, and variety of physical-world data increasingly important.

A New AI Data Economy Could Be Emerging

Mecka AI’s reported funding round is significant not only because of the company’s potential valuation, but because it points to where the AI industry may be heading.

The companies that build advanced robots will need hardware and AI models, but they will also need vast amounts of information about how the physical world works.

That could create a new data economy around robotics, with businesses collecting human demonstrations, labeling physical actions, evaluating robot behavior, and building datasets specifically designed for machines.

Mecka AI is positioning itself in the middle of that emerging market.

If humanoid robots become widely deployed, the demand for high-quality physical-world training data could grow dramatically. For now, the reported Sequoia deal is another sign that investors are betting heavily on that future.