A startup focused on capturing and processing human motion data to train humanoid and other robotic systems is reportedly closing in on a new funding round led by Sequoia Capital that would value the company at approximately $500 million. Two people familiar with the matter confirmed the deal is in progress, though final terms have not yet been agreed upon and remain subject to change.

The company in question, Mecka AI, is moving quickly. This new financing would follow by just three months a $60 million round the startup announced earlier this year, which was led by Framework Ventures and included backing from Menlo Ventures, SV Angel, and Kindred Ventures. The exact size of the new round has not been disclosed. Mecka AI did not respond to requests for comment, and Sequoia declined to offer any statement.

A Founding Team Without Robotics Roots

Mecka AI was established in 2024 by four co-founders, none of whom come from traditional robotics backgrounds. Among them are Canadians Josh Gao and Mogen Cheng, who previously built a fintech startup serving the restaurant industry. Jason Chong, another co-founder, had previously launched a cryptocurrency exchange that was later acquired by Coinbase, after which he joined the company. The fourth co-founder, Duy Nguyen, is the only non-Canadian on the team and handles operational responsibilities at Mecka.

Despite the absence of direct robotics experience, the team identified a critical gap in the development pipeline for general-purpose robots: a severe shortage of physical-world interaction data. They concluded that the inability to capture real-world human movement at scale was the primary obstacle preventing humanoid and other robots from becoming truly capable machines.

The Business Model and Its Inspiration

The company's name draws from "mecha," a term used in science fiction to describe large robots operated by human pilots. That reference is intentional - Mecka AI positions itself as a bridge between human physical behavior and robotic intelligence. Its business model is modeled after what companies like Scale AI, Mercor, Surge, and similar human-data platforms have done for large language models. Rather than training AI on text or images, Mecka focuses on physical movement data.

To gather that data, Mecka pays individuals to record themselves carrying out routine, everyday tasks - things like preparing a meal, performing automotive repairs, or completing household chores. Participants use body sensors and smartphones to capture these interactions, creating what the industry refers to as "egocentric" data: footage and sensor readings taken from the perspective of the person performing the task.

This approach mirrors how robotics companies and AI research labs are increasingly sourcing the real-world physical data their systems need. While Mecka has not publicly named any of its customers, the broader industry's reliance on egocentric data collection - alongside other methods such as teleoperation - is well established among leading robotics developers and AI laboratories.

Revenue Projections and Growth Trajectory

When Mecka announced its previous fundraising round in early June, co-founder Josh Gao shared forward-looking revenue projections with Fortune. At that time, the company was projecting it would close out 2026 at an annualized revenue run rate of $100 million - a figure that underscores the pace at which demand for robot training data is accelerating.

The speed at which Mecka is returning to the fundraising market - just three months after its last announced round - reflects broader momentum in the physical AI and robotics data sector, where investor appetite has grown sharply alongside advances in humanoid robot development.

A Crowded but Expanding Market

Mecka AI is not alone in pursuing this opportunity. Several other startups are competing in the real-world data collection space for robotics training. One notable competitor is XDOF, which was recently reported to be approaching a new funding round at a valuation of approximately $1.2 billion. Additionally, established human-data platforms that originally built their businesses around supplying training data for language models - including Scale AI and Micro1 - are now expanding their offerings to serve the physical robotics market as well.

The convergence of humanoid robotics development and the demand for high-quality physical interaction data has created a rapidly growing segment of the AI infrastructure market. Investors appear to be betting that companies capable of supplying that data at scale will become essential components of the robotics supply chain - much as data annotation and labeling companies became foundational to the rise of modern machine learning over the past decade.

As Sequoia's reported involvement suggests, top-tier venture capital firms are now actively competing to back the infrastructure layer beneath the humanoid robotics boom, treating data collection platforms as critical picks-and-shovels plays in one of the most anticipated technology transitions of the coming decade.