Mecka AI is attracting a new wave of investor interest as the robotics industry confronts one of its biggest bottlenecks: high-quality physical-world training data.
The two-year-old startup is nearing a Sequoia Capital-led funding round at a valuation of about $500 million, according to people cited in the source. The exact size and terms were not final.
The prospective Sequoia deal follows a $60 million financing announced just three months earlier, led by Framework Ventures with Menlo Ventures, SV Angel and Kindred Ventures also participating. That pace of fundraising shows how quickly capital is moving toward the data layer behind humanoid and general-purpose robotics.
Mecka was founded in 2024 by four entrepreneurs, including Josh Gao, Mogen Cheng, Jason Chong and Duy Nguyen. The founders did not come from traditional robotics backgrounds. Their thesis was that the scarcity of real-world behavioral data, rather than model architecture alone, was limiting robot performance.
The company pays people to record themselves performing everyday tasks such as making coffee or repairing cars. Body sensors and smartphones capture the movements, creating datasets that robotics companies can use to train models on how humans interact with the physical world.
The model resembles what companies such as Scale AI did for large language models, but with a harder data problem. Text and images already exist in enormous digital quantities. High-quality physical-action data is much more expensive to collect, label and standardize.
That scarcity is driving competition. XDOF was recently reported to be raising capital at a $1.2 billion valuation, while broader human-data platforms such as Scale AI and Micro1 are also expanding beyond language-model workflows.
Mecka’s growth expectations are aggressive. As of early June, the company was projecting that it would end 2026 at a $100 million annual run rate. The customer list has not been publicly disclosed, but many robotics companies and AI labs use egocentric data and teleoperation to train systems.
The investment case depends on whether training data remains a durable bottleneck. If robotics companies eventually generate enough synthetic data or collect their own information internally, external data providers could face margin pressure.
The bull case is that physical-world data behaves more like infrastructure. Every new task, environment and robot design creates additional training needs, making collection an ongoing requirement rather than a one-time dataset.
A $500 million valuation implies investors are leaning toward that second view. Mecka now has to prove that data collection can scale quickly, remain high quality and produce recurring customer demand. In the robotics boom, compute and models get most of the attention. Mecka is betting that the scarce asset will be human behavior itself.
Customer concentration will be another issue once Mecka discloses more commercial detail. A data supplier serving a small number of large robotics labs can scale quickly but may face heavy bargaining power. A broader customer base would improve both resilience and valuation quality.
Data quality is likely to become the moat. Collecting large volumes of human motion is relatively straightforward compared with ensuring the recordings are diverse, accurately labeled and useful for specific robot tasks. Customers will pay more for datasets that reduce training time or improve real-world performance. Mecka’s valuation therefore depends on whether its collection network produces information that is difficult to replicate, not merely whether it can recruit more people to record videos. If the company can build proprietary standards around quality and task coverage, the data layer could become sticky infrastructure rather than a commodity service.