Mecka AI Nears $500M Valuation in Sequoia-Led Robot Training Data Round

The $500 Million Valuation

Mecka AI, a startup specializing in motion data collection for robotics training, is finalizing a funding round led by Sequoia Capital that values the company at approximately $500 million, according to sources familiar with the deal. The valuation marks substantial growth from the company’s previous funding round announced just three months earlier, when Mecka raised $60 million in a Series A backed by Framework Ventures, with additional support from Menlo Ventures, SV Angel, and Kindred Ventures. The exact size of the current Sequoia-led round has not been announced, and final terms remain subject to potential revision.

The rapid succession of major funding announcements reflects a broader venture capital push to back companies addressing what has emerged as a critical bottleneck in robotics development: access to high-quality training data sourced from real-world human motion. As companies work toward commercializing humanoid robots and deploying general-purpose robotic systems, the availability and cost of training data has become a primary constraint on product development timelines and capability advancement.

The Founders: Unconventional Paths to Robotics

Read also: Former OpenAI executive Fidji Simo joins Nscale board ahead of fall IPO · Listen Labs walks from $1.5B funding round amid Salesforce talks · Utilities Back Fusion Power with First Grid-Connected Deals

A Team Without Robotics Experience

Mecka AI was founded in 2024 by four entrepreneurs whose career trajectories diverge significantly from the traditional robotics founder profile. Josh Gao and Mogen Cheng, both Canadian, had previously built a fintech startup serving restaurants. Jason Chong had sold a cryptocurrency exchange to Coinbase before entering robotics and AI. Duy Nguyen focuses on operations for the startup. Notably, none of the four bring formal backgrounds in robotics engineering, hardware development, or academic robotics research.

Why This Matters

The unconventional composition of the founding team may have been a strategic advantage. Unencumbered by assumptions embedded in traditional robotics, the team identified a fundamental problem that industry incumbents had largely accepted as inevitable: a severe scarcity of training data for physical systems. While machine learning had benefited from massive freely available datasets of text and images, robotics faced a different constraint entirely. Robots needed to learn from video, motion captures, and sensor data showing real humans performing real-world tasks—and such datasets existed in extremely limited quantities.

The Core Insight

The four founders recognized that the path forward followed a pattern already validated in adjacent sectors. Companies like Scale AI, Mercor, and Surge had demonstrated that crowdsourced human-generated training data could power AI systems at scale. Mecka applied this proven model to robotics, positioning itself as a data infrastructure platform rather than a robotics company. The startup’s name itself references “mecha,” the Japanese concept of large humanoid robots controlled by human pilots—a conceptual tie to the company’s mission of capturing human motion for machine learning systems.

The Physical Data Bottleneck

Robotics development faces a constraint that distinguishes it sharply from language model training: robots must learn to operate in physical space, performing tasks that require understanding real-world geometry, friction, balance, and human body mechanics. While text and image datasets can be assembled from the internet at minimal cost, training data for robots must be deliberately captured from humans performing real-world activities.

Mecka operates by paying participants to perform everyday tasks while wearing body sensors and using smartphones to record their movements. Contributors engage in mundane activities—brewing coffee, repairing cars, preparing meals, assembling objects—that roboticists eventually want their machines to learn to perform. This “egocentric” data collection method, which captures video and motion from the first-person perspective of the human operator, generates the training material that robotics AI labs feed into their models. The approach complements other physical data collection techniques already in use, particularly teleoperation systems where humans directly control robot arms or bodies, creating dual streams of training data that inform robot learning.

Building on a Proven Model

Mecka’s strategy mirrors the infrastructure business model that Scale AI refined for language models. Rather than building robots or AI models directly, Mecka focuses on becoming the essential data layer supporting everyone else’s robotics work. The company sources high-quality motion data, manages quality control, and delivers datasets to robotics companies and AI research labs that integrate the training material into their systems. This approach concentrates Mecka’s operational focus on data curation and scale rather than diversifying into multiple product categories.

Market Competition and Demand

Mecka operates in an increasingly crowded sector. XDOF, a competing platform, closed funding last week at a $1.2 billion valuation, according to TechCrunch. Scale AI, which built a multi-billion-dollar business providing training data for large language models, has expanded into robotics data markets. Micro1, another human-data platform, is similarly broadening beyond LLM training into robotics.

Despite the competitive landscape, demand for motion data is accelerating. Robotics companies and AI research organizations widely rely on motion capture and egocentric video to train their systems, though Mecka has not publicly disclosed its customer roster. Companies pursuing humanoid and general-purpose robots—including established players like Boston Dynamics and Tesla, as well as numerous smaller startups—all require access to training data sourced from real-world human motion. The consistency of this demand across the industry, combined with no single company’s ability to meet all supply at required quality standards, has created the conditions for a durable market segment.

Growth Projections and Sector Implications

The company’s growth trajectory has been steep. When announcing its $60 million Series A in June, co-founder Josh Gao disclosed that Mecka was projecting an annual run rate of $100 million by the end of 2026. If achieved, the company would scale from zero revenue at founding to nine-figure annual business in less than three years—a trajectory comparable to the fastest-growing data infrastructure startups ever funded.

The $500 million valuation, assigned to a company that did not exist in its current form before 2024, reflects venture investor conviction that motion data will become as fundamental a resource for robotics as text corpora were for large language models. Whether Mecka can sustain this trajectory depends on its ability to expand its contributor network, maintain rigorous data quality standards, and continue converting the growing roster of robotics companies into paying customers. The broader implication remains clear: as robotics advances, the infrastructure layer supporting that development—the data, tools, and platforms enabling training—may ultimately command as much market value as the robotics companies themselves.