Mecka AI announced a $60 million Series B on October 7, 2026, led by Sequoia Capital, in what amounts to a bet that robotics companies need training data even more urgently than they need chips. NVIDIA, M12 (Microsoft's venture arm), Qualcomm Ventures, and Samsung participated, alongside earlier investors Kindred, Framework Ventures, and Neo.
The timing matters. While digital AI firms scramble for GPUs, robotics labs face a different constraint: gathering enough real-world human behavior to teach machines how to move, grasp, and navigate. Sequoia partners Anas Biad and James Flynn framed the opportunity in sweeping terms, writing that "bringing AI into the physical world will be a wave of comparable importance to digital AI, if not larger." Mecka, they argue, provides the infrastructure to make that happen. The startup says it counts several frontier robotics labs and multiple Mag 7 tech companies as clients, though it declined to name them.
Mecka's pitch centers on capturing human demonstrations through iPhones, body sensors, and proprietary hardware, then processing that footage into datasets suitable for training robot foundation models. The company also supports on-site robot deployments and what it calls "post-training loops" to refine performance. CEO Josh Gao told Fortune earlier that the core thesis is straightforward: "train robots on data collected from humans, not from data collected through teleoperation."
That philosophy diverges from competitors who rely on remote-controlled robots generating their own training runs. Gao believes useful robots can ship now, not years from now, a stance that puts Mecka at odds with hardware-first humanoid makers still focused on prototypes. The Series B proceeds will fund scaling its data infrastructure, deepening its internal research lab, and expanding commercial robot deployments.
Revenue claims and caveats
Mecka claims it surpassed $100 million in annualized run-rate revenue in June 2026 and projects a $300 million run rate by the end of the year. These figures are unaudited. The startup raised $25 million in a Series A last November and a $35 million follow-on a few months later, both led by Framework Ventures, according to Fortune. TechCrunch reported on September 11, 2026, that Mecka was nearing a Sequoia-led deal at around a $500 million valuation; the company did not confirm the final figure.
The founding team brings an unconventional background. Gao, CTO Jason Chong, CPO Mogen Cheng, and COO Duy Nguyen all come from fintech and crypto rather than robotics. Gao and Cheng previously built restaurant fintech software. Chong sold a cryptocurrency exchange to Coinbase. Nguyen has launched multiple ventures. LinkedIn lists the company at 11–50 employees, though 68 profiles appear on its company page and Dealroom estimates 51–200. Headcount data for startups is often messy.
A suddenly crowded field
Mecka operates in what was, until recently, a niche corner of AI infrastructure. That changed fast. XDOF, which specializes in teleoperation and wearable-sensor data, raised a $70 million Series A in June 2026 and was in talks for a Series B at a $1.2 billion valuation as of September 4, 2026, per TechCrunch. Sunday Robotics, maker of an in-home skill-capture glove, secured $165 million at a $1.15 billion valuation in March 2026. Config debuted its HABIT bimanual dataset and CFG-1 foundation model mid-year, while Tacta Systems introduced sensorized industrial gloves in July.
The influx of capital reflects emerging academic consensus. EgoVerse, presented at a recent Robotics: Science and Systems conference, documented reproducible performance gains from co-training robots with human egocentric video across different labs and hardware platforms. Meanwhile, the International Federation of Robotics reported that 5.079 million industrial robots were operating globally as of late 2025, a 9 percent annual increase, with installations hitting 603,000 units.
Wall Street sees the data bottleneck as more than theoretical. Goldman Sachs recently raised its 2035 humanoid robot forecast to approximately 6.5 million units and a market worth roughly $138 billion, while Bank of America Global Research projected 10 million humanoid shipments by the same year. Both firms cited data availability as a gating factor. BMW, which established a Center of Competence for Physical AI in Production in February and piloted Figure AI's humanoid at its Spartanburg plant, said in a statement that "digitalization improves the competitiveness of our production."
The automaker's language was careful, perhaps deliberately vague. But the broader pattern is unmistakable: companies are placing bets that the next wave of AI won't just think faster but move smarter, and that whoever controls the training data controls the infrastructure layer beneath it all.
