Mecka AI Nears $500M Valuation in Sequoia-Led Deal as Robotics Labs Race for Real-World Training Data

In a vivid demonstration of how venture capital is migrating from digital software models to physical automation infrastructure, robotics data startup Mecka AI is nearing a new financing round led by Sequoia Capital that values the two-year-old company at approximately $500 million. The transaction, reported by sources familiar with the negotiations, arrives a mere three months after Mecka announced a $60 million Series A led by Framework Ventures alongside Menlo Ventures, SV Angel, and Kindred Ventures.

The rapid valuation markup highlights an acute operational crisis across embodied artificial intelligence: the physical data wall. While large language models scaled effortlessly by ingesting trillions of tokens of publicly available text from the open internet, physical-world robots cannot learn complex manipulation, dexterity, and navigation from Wikipedia articles or Reddit archives. To train general-purpose humanoid robots and agile warehouse systems, AI models require millions of hours of rich, high-frequency physical interaction telemetry—a dataset that simply did not exist at scale.

Founded in 2024 by four entrepreneurs—Canadians Josh Gao and Mogen Cheng (who previously built restaurant fintech ventures), former Coinbase engineer Jason Chong, and operations lead Duy Nguyen—Mecka approached robotics without traditional mechanical engineering backgrounds. Instead, they recognized that the primary bottleneck stifling robotics foundation models was an infrastructure problem identical to the early days of language models.

Taking inspiration from the term 'mecha'—human-controlled mechanical suits—Mecka set out to build for robotics what Scale AI, Mercor, and Surge built for large language models: a human-in-the-loop physical data collection engine. The company contracts thousands of distributed human contributors equipped with body-worn inertial measurement units (IMUs), haptic gloves, and egocentric smart glasses or smartphones. These workers record themselves executing ordinary, fine-motor tasks across domestic and industrial settings, ranging from brewing espresso and operating power tools to folding irregular fabrics and assembling automotive sub-assemblies.

This egocentric video, paired with synchronized multi-axis kinematic and proprioceptive sensor streams, is transformed into structured tokenized demonstrations. When fed into vision-language-action (VLA) foundation models, the data bridges the sim-to-real gap, teaching robotic actuators how to adapt to physical resistance, surface friction, and unmapped spatial clutter.

Demand for real-world kinematic demonstrations has surged as well-funded robotics ventures and hyperscale AI labs race to deploy commercial pilots. While Mecka maintains confidentiality regarding its client roster, industry analysts point to the aggressive expansion of humanoid programs at Tesla, Figure AI, Boston Dynamics, 1X Technologies, and frontier foundation labs building multimodal spatial models.

The commercial appetite has translated into explosive revenue velocity. Following its Series A fundraise, CEO Josh Gao projected that Mecka would exit 2026 with an annualized run rate approaching $100 million. If achieved, the milestone would position Mecka as one of the fastest-growing data infrastructure platforms in venture history, justifying Sequoia’s premium valuation multiple.

Mecka’s escalating valuation reflects a broader venture contest for the physical AI data layer. Competitors such as XDOF are reportedly finalizing rounds valuing their platforms north of $1.2 billion, while established LLM labeling giants like Scale AI and Micro1 are aggressively deploying teleoperation rigs and kinematic data capture divisions.

As hardware components like actuators, harmonic drives, and solid-state lidar standardize and commoditize, competitive moats in robotics are consolidating around training data density. By industrializing human physical task capture into scalable, pre-packaged training datasets, Mecka is establishing itself as an indispensable utility for the physical AI revolution.

• Mecka AI nears $500M valuation in Sequoia-led deal amid rush for robot training data — TechCrunch
• Mecka AI Approaches $500 Million Valuation in Sequoia-Backed Funding Round for Robot Training Data — KillBait Archive
• Mecka AI Nears Sequoia-Led Funding at a Reported $500 Million Valuation — Superpower Daily