Mecka AI Just Raised $60M From Sequoia to Teach Robots How Humans Move

Sequoia is betting $60M that robot data is the next gold rush

Mecka AI, a startup founded in 2024, announced a $60 million Series B round led by Sequoia on October 7, at a valuation of around $500 million. Nvidia and Microsoft's venture fund M12 joined the round. What does Mecka do? It pays ordinary people to record themselves doing everyday tasks — making coffee, fixing cars, folding laundry — while wearing body sensors and using smartphones. The company then turns that human motion data into training material for humanoid robots and other machines.

The pitch is simple: Mecka intends to do for robotics what Scale AI, Mercor, and Surge did for large language models. Those companies supplied the human-generated data that LLMs learned from. Mecka wants to be the equivalent supplier for the physical world.

Why motion data is the hard part

Training a chatbot is cheap in comparison: the internet is full of text. Training a robot to pick up a mug requires knowing exactly how a human wrist rotates, how much force a grip needs, how weight shifts when you reach. Simulation helps, but the sim-to-real gap — the difference between a physics engine and a cluttered kitchen — is where most robot demos die. Teleoperation, where a human remotely pilots a robot to generate training episodes, works but is slow and expensive.

Mecka's bet is that recording real humans at scale, with sensors capturing the full body, produces the volume and variety of physical data that makes generalist robot policies possible. Paying people to live their lives with sensors on is a clever way to get data diversity that no lab environment can match.

The land grab is on

Mecka isn't alone. XDOF, another startup collecting real-world data for robot training, was reportedly in talks to raise a Series B at a $1.2 billion valuation. Meanwhile the human-data platforms that started with LLMs are expanding into robotics: Scale AI and Micro1 both see physical AI as the next frontier. When the data-labeling giants and the specialists all point at the same market, you know the land grab has started.

The open question is defensibility. Motion data is valuable, but is it a moat? The first company to build a large, clean, well-labeled corpus of human movement has an edge — until the second company records the same coffee-making motion. Mecka's advantage will come from the pipeline: the sensors, the quality control, the labels, and the relationships with robot makers who need a continuous feed, not a one-time dataset.

There's a deeper reason the money is flowing now. Humanoid robot pilots are moving from lab demos to paid deployments — warehouses first, then factories — and every serious program needs training data yesterday. The startups that lock in data-supply contracts with the robot builders become infrastructure, not vendors. That's the bet Sequoia, Nvidia, and M12 just placed: that Mecka becomes the Scale AI of the physical world before the giants build their own pipelines in-house.

What to watch

Two things decide whether this $60M was well spent. First, whether humanoid timelines hold — the entire robot-data market assumes humanoids reach commercial deployment in the next few years; if that slips, the data business gets a longer, leaner runway. Second, whether Mecka can sign the robot builders as customers before the data becomes commoditized.

The funding environment for this kind of bet is clearly hot — compare Ramp's $1.85B round or TypeSafe's $870M Series A for how concentrated 2026's mega-rounds have become. For the week-by-week view of where the money is going, our startup funding roundup tracks every major raise.