Mecka, a company building data infrastructure for robot learning, announced on October 7, 2026, that it has raised a $60 million Series B led by Sequoia Capital. New investors in the round include NVIDIA, Qualcomm Ventures, Samsung and M12, Microsoft’s venture fund, the company said, with continued support from Kindred, Framework Ventures and Neo.
Angel investors participating include DoorDash founder and CEO Tony Xu, former ServiceNow and Snowflake CEO Frank Slootman, and former Tesla Optimus head Milan Kovac, according to the company. Mecka said the capital scales its data infrastructure, deepens its internal research lab, and builds out commercial robot deployment, work the company describes as the data and deployment layer for robotics.
The company’s thesis is that what robots lack is experience. No one ever wrote down how a hand grips, folds or pours, how hard to press, or when to let go. “Motion, contact, force and geometry aren’t on the internet. You can’t scrape them, license them or buy them,” the company wrote in its Series B announcement. Mecka describes that gap as the biggest bottleneck in robotics, calls robotics the largest productivity opportunity ahead, and predicts data will become a trillion-dollar market.
On its website, the company describes a series of challenges between raw activity and usable training data. Real-world tasks are noisy, so models need large-scale edge cases and task-specific demonstrations to generalize. Raw video is not enough: teams need annotation, context, quality assurance, motion understanding and evaluation workflows before real-world activity becomes useful training data. Finally, the lab is not the finish line: robotic systems must be tested against real workflows, environments and constraints before they can work in the field.
The Capture and Research Stack
Mecka said that to digitize the physical world it had to build the entire stack: capture hardware, reconstruction models and quality infrastructure. The company designs and manufactures its own multi-sensor hardware for each signal class, with sensor selection and synchronization driven by what its models need downstream. It runs a global production operation, with capture fleets recording human demonstration in homes and commercial environments.
An internal lab builds the computer vision and multimodal models that turn raw, noisy reality into structured signal, covering motion tracking, 3D reconstruction and sensor alignment. The company reports state-of-the-art sub-centimeter hand-pose accuracy on in-the-wild data.
The company’s website lists signal and action primitives, modules it describes as ways of capturing or interacting with physical reality: pixel understanding, depth and geometry, object tracking, segmentation, tactile force, sound waves, speed and motion, point cloud, surface contour, and localization. Mecka also offers an iOS app for capturing environment- and task-specific data, plus web tools for browsing, visualizing and querying data, collaborating on datasets with teams, and running inference APIs.
The EgoVerse Dataset and Transfer Study
Mecka said it tested its bet on real-world human demonstration with researchers at Georgia Tech, Stanford, UC San Diego, ETH Zürich, MIT and Meta through EgoVerse, a human-to-robot transfer study replicated across labs, tasks and robots. The EgoVerse paper, first submitted on April 8, 2026, and last revised on July 7, 2026, describes a collaborative platform that unifies data collection, processing and access under a shared framework, enabling contributions from individual researchers, academic labs and industry partners.
The paper’s current release includes 1,362 hours of human demonstrations across 80k episodes, spanning 1,965 tasks, 240 scenes and 2,087 unique demonstrators, with standardized formats, manipulation-relevant annotations and tooling for downstream learning. Mecka’s FAQ describes EgoVerse as its large-scale human interaction dataset, built from real-world first-person activity across tasks, spaces and environments.
Beyond the dataset, the authors conducted a large-scale study of human-to-robot transfer, with experiments replicated across multiple labs, tasks and robot embodiments under shared protocols. The paper reports that policy performance generally improves with increased human data, but that effective scaling depends on alignment between human data and robot learning objectives. The author list includes Josh Gao and Jason Chong.
Customers, Revenue and the Integrator Model
Mecka said it supplies several of the top frontier robotics labs and multiple Mag 7 companies. The company said it surpassed $100 million in run-rate revenue in June 2026, within only a few months of operations, and projects a $300 million run rate by the end of the year.
The same stack powers the company’s commercial deployments. Mecka describes itself as a modern robotics integrator: where a traditional integration ships once and stays fixed, its deployments capture data on site, post-train on it, and improve with every hour they run, according to the company. For enterprises that want physical AI but have no robotics team, Mecka said it brings the hardware, data capture, post-training, integration and ongoing operations. Its FAQ states that Mecka does not build robots; instead it plugs into the hardware, model and commercial ecosystems and owns the data, integration and evaluation layer between them, an arrangement the company calls the Mecka Loop.
Mecka said the financing lets it build more data faster: more instruments, deeper research, and robots deployed where the work is. The announcement is signed by Josh, Jason and the Mecka team, and the company said it is hiring across research, hardware and operations.
