Mecka Raises $60M Series B to Scale Physical AI Data

A language model can inherit a civilization's written record before it answers its first prompt. A robot entering a warehouse, laboratory, kitchen, or hospital gets no comparable archive of force, contact, movement, and the small adjustments that make physical work succeed. Mecka is raising capital to build that missing record and carry it into the places where robots are expected to perform.

The New York and Toronto physical AI company announced a $60M Series B on October 7, 2026, led by Sequoia Capital. NVIDIA, M12, Qualcomm Ventures, and Samsung joined as new investors, while Kindred Ventures, Framework Ventures, and Neo continued their support. The company said the capital will expand its data infrastructure, deepen its internal research lab, and build out commercial robot deployments.

The round matters because Mecka is extending its job beyond selling training data. Its ambition is to connect human demonstrations, capture hardware, computer vision, post-training, integration, and ongoing operations into one delivery system for enterprises that want physical AI but do not have an internal robotics organization.

What Mecka Is Building

Mecka was founded in 2025 by Josh Gao, Jason Chong, Mogen Cheng, and Duy Nguyen. Gao serves as CEO and Chong as CTO. The founders came from payments, software, operations, and consumer businesses, then spent months studying robotics research and working with labs around a central premise: robots can learn useful physical behavior from large collections of human activity.

Collecting a video is only the first step. Mecka designs multi-sensor capture hardware, runs collection fleets in homes and commercial environments, and develops computer vision and multimodal systems that turn raw activity into motion tracking, 3D reconstruction, sensor alignment, annotations, and evaluation data. The company says its system reaches sub-centimeter hand-pose accuracy in uncontrolled settings, a technical claim that remains company-reported.

Mecka has also contributed to EgoVerse, a multi-institution study of egocentric human demonstrations for robot learning. The project includes researchers and collaborators from Georgia Tech, Stanford, UC San Diego, ETH Zürich, MIT, Meta, Mecka, and other organizations. Its value is broader than one dataset: it tests how human demonstrations can transfer across labs, tasks, and robot embodiments rather than remaining locked to one machine and one carefully staged environment.

The $60M Series B and the Capital Behind It

Sequoia led the Series B, and the investor group spans several layers of the physical AI stack. NVIDIA supplies accelerated computing, Microsoft operates cloud and enterprise platforms, Qualcomm develops edge and device silicon, and Samsung sits across components, devices, and manufacturing. Their participation does not establish commercial partnerships by itself, but it shows where strategic investors expect the category's infrastructure to connect.

Mecka also named returning investors Kindred Ventures, Framework Ventures, and Neo. Tony Xu, Frank Slootman, and Milan Kovac joined as angels, bringing operating experience from DoorDash, ServiceNow, Snowflake, and Tesla's Optimus program. The company said the new capital will fund more capture instruments, deeper research, larger data operations, and commercial robot deployment.

The financing follows substantial earlier rounds. An $8M seed was announced in August 2025, and DevCuration previously covered another $60M across a $25M Series A and a $35M follow-on led by Framework Ventures, based on Fortune's reporting. Adding the distinct Series B brings publicly disclosed financing to about $128M, although Mecka did not publish a cumulative total in the new announcement.

From Training Data to Enterprise Deployment

Mecka's business is moving toward the customer site. A traditional data supplier can deliver a dataset and stop. Mecka describes a system that captures task-specific activity, helps post-train models, integrates robots into operating environments, and keeps collecting evidence as those deployments encounter new conditions.

That expansion addresses a practical buyer problem. An enterprise may want robots in a factory, laboratory, warehouse, or service operation without wanting to assemble a robotics research group, data operation, hardware integration team, and field-support function from scratch. Mecka is positioning itself as the connective layer across those responsibilities.

The operating challenge grows with that promise. Data volume alone does not guarantee that a robot will generalize safely, and performance in a research setting does not automatically transfer into a customer's messy workflow. Mecka will have to preserve data quality, measure marginal learning value, handle site-specific integration, and show that deployments improve enough to justify their cost.

Revenue Claims and the Evidence Boundary

Mecka says it supplies several leading frontier robotics labs and multiple Mag 7 companies, but the funding announcement does not identify those customers. It also says it surpassed $100M in run-rate revenue in June 2026 and projects a $300M run rate by year-end. Those figures are company-reported, the forward projection has not yet been realized, and no independent revenue audit was disclosed.

The qualification does not erase the commercial signal. A young company attracting this amount of capital from venture and strategic investors while moving from data collection into deployments suggests that physical AI buyers are already spending against the experience gap. The open question is whether Mecka can turn early demand into a repeatable system across customers, robots, and operating environments.

Why This Matters for Physical AI

Digital AI scaled on a record the world had already produced: text, images, code, clicks, and transactions. Physical AI must assemble much of its training record while the category is being built. That changes the economics because capture hardware, consent, operations, quality control, model training, field integration, and customer support become part of the intelligence supply chain.

Mecka is betting that one company can coordinate those layers well enough to become infrastructure beneath many robots rather than a product tied to one form factor. The Series B gives the team more room to test that position in live environments, where the useful data is created by work itself and every deployment exposes the next physical lesson the system still needs to learn.