Mecka AI Nears $500M on the Fight for Robot Training Data
Inside the startup paying people to record themselves making coffee and fixing cars — the physical-world analog of Scale AI.
Published: 2026-09-13 Category: Quick Take Sources: TechCrunch
The Bottleneck Nobody Talks About
Everyone obsesses over GPU supply and inference costs, but there's a quieter bottleneck holding back general-purpose robots: physical-world data. LLMs got their Scale AI, Mercor, and Surge — companies built on human-generated text. Robots have no equivalent pipeline, because there's no torrent of existing human motion data to scrape. Mecka AI is trying to build that pipeline.
The startup is nearing a new round led by Sequoia Capital at roughly a $500 million valuation, according to two people with knowledge of the deal. It comes just three months after Mecka raised $60 million in a round led by Framework Ventures with Menlo Ventures, SV Angel, and Kindred Ventures.
An Unusual Founding Story
Mecka's four co-founders — Josh Gao, Mogen Cheng, Jason Chong, and Duy Nguyen — don't have robotics backgrounds. Gao and Cheng previously built a restaurant fintech; Chong joined Coinbase after it acquired his crypto exchange. What they recognized was a data gap: real-world interactions are the primary bottleneck holding back humanoid robots and other general-purpose machines.
Their model mirrors the human-data companies that powered the LLM boom. Mecka pays people to record themselves performing everyday tasks — making coffee, fixing cars — using body sensors and smartphones. The name itself comes from "mecha," the fictional giant robots from anime.
The Numbers Behind the Hype
As of early June, Mecka was projecting it would end 2026 at an annual run rate of $100 million, CEO Josh Gao told Fortune at the time of the previous fundraise. That's not a validated customer list — Mecka hasn't publicly disclosed who its robotics clients are — but the projection, combined with a Sequoia-led round at $500M, signals investors believe the physical-data market is about to explode.
TechCrunch notes the terms of the new deal aren't final and could still change. Sequoia declined to comment; Mecka didn't respond.
Why This Is the Next Gold Rush
The parallel with the LLM wave is instructive. Before Scale AI, people didn't think of "training data" as a venture-scale category. It became one because models were only as good as their data. Humanoids face the same equation but harder: you can't synthesize real-world physics and dexterity from text.
If Mecka is right that capturing real-world interactions is the primary bottleneck, then the startup building the human-motion pipeline is effectively selling pickaxes in a gold rush. The question is whether the boom in physical-data collection arrives before the burn rate catches up. Either way, the $500M mark says the market's betting yes.
Sources: TechCrunch, "Mecka AI nears $500M valuation in Sequoia-led deal amid rush for robot training data" (Sep 11, 2026).