Mecka AI is close to closing a new funding round led by Sequoia Capital at a valuation of roughly $500 million, according to two people with knowledge of the deal. The round comes just three months after the startup raised $60 million led by Framework Ventures, with participation from Menlo Ventures, SV Angel, and Kindred Ventures. The precise size of the new round has not been disclosed, and terms remain subject to change.
The company collects and analyzes human motion data to train humanoid robots and other robotics systems. The core idea is straightforward: general-purpose robots need to understand how humans interact with the physical world, and the best way to get that data is to pay people to record themselves doing everyday tasks. Mecka's workers wear body sensors and use smartphones to capture movements like making coffee, fixing cars, or navigating cluttered rooms. The resulting datasets are then sold to robotics companies and AI labs building the models that power their machines.
The Data Bottleneck
The robotics industry has a well-understood problem. Software models have improved dramatically in recent years, but the physical-world data needed to train them remains scarce. Language models train on text scraped from the internet. Vision models train on images and video. Robot models need something harder to obtain: high-quality recordings of human bodies interacting with objects, surfaces, and environments in three dimensions.
Mecka was co-founded in 2024 by four entrepreneurs who recognized this gap. Josh Gao and Mogen Cheng are Canadians who previously built a restaurant fintech startup. Jason Chong joined Coinbase after it acquired his crypto exchange. Duy Nguyen, the only non-Canadian on the team, focuses on operations. None of the co-founders have backgrounds in robotics, but they identified the data problem as the primary bottleneck holding back general-purpose robots and set out to solve it.
The company's name derives from mecha, the fictional giant robots controlled by humans in anime and science fiction. The reference is apt. Mecka is trying to do for robotics what Scale AI, Mercor, and Surge have done for language models: build the human data infrastructure that makes the underlying AI systems work.
The Economics of Human Motion Data
Mecka pays people to record themselves performing tasks. The recordings capture body movements through sensors and smartphone cameras, producing what the industry calls egocentric data. This is data captured from the perspective of the person performing the action, which is the same perspective a humanoid robot would have as it navigates a kitchen or a workshop.
The approach scales better than teleoperation, where human operators control robots remotely to generate training data. Teleoperation produces high-quality data but requires expensive hardware and specialized operators. Mecka's method uses consumer devices and pays ordinary people, which keeps costs lower and volume higher.
As of early June, Mecka was projecting that it would end 2026 at an annual run rate of $100 million. That figure suggests strong demand from robotics companies, though the startup has not publicly disclosed its customer list. Many robotics companies and AI labs rely on egocentric data captured through this approach alongside other physical data collection methods to build their models.
A Crowding Market
Mecka is not alone in this space. XDOF, another startup collecting real-world data for robot training, was reported by TechCrunch last week to be nearing a new round at a $1.2 billion valuation. Scale AI, which built its business labeling data for language models, is expanding into physical-world data. Micro1 is another human-data platform making similar moves.
The competition reflects a broader shift in the AI industry. As language models mature and their training data becomes increasingly commoditized, the frontier of AI development is moving into the physical world. Robots need to grasp objects, walk across uneven surfaces, and interact with humans in unpredictable environments. The models that enable these capabilities are only as good as the data they train on, and that data is expensive and difficult to produce at scale.
The rush for robot training data mirrors the earlier scramble for language model training data. Companies that positioned themselves early in that cycle, like Scale AI, became foundational infrastructure providers. The question for Mecka and its competitors is whether the robotics data market will consolidate the same way, or whether the diversity of physical-world tasks will support multiple specialized providers.
For now, the funding rounds suggest that investors believe the bottleneck is real and that the companies solving it will be valuable. A $500 million valuation for a two-year-old startup with no public customer list is a bet on the future of robotics, not a reflection of current revenue. If humanoid robots do become mainstream, the companies that provided the data to train them will have been early to a market that did not exist when they started.