Mecka AI Nears $500 Million Valuation as Robotics Data Race Intensifies

Mecka AI is nearing a new funding round led by Sequoia Capital that would value the robotics data startup at about $500 million, underscoring how quickly investor demand for robot-training data is accelerating. The deal is coming only three months after Mecka disclosed a $60 million raise, highlighting the surge of capital chasing companies that can supply the physical-world data robots need to learn.

The proposed valuation reflects a broader shift in artificial intelligence investing: as large language models have already absorbed vast amounts of internet text, backers are now hunting for the equivalent data infrastructure for machines that move, grasp, and operate in the real world. Mecka sits squarely in that emerging market, collecting human motion data that helps train humanoid robots and other robotics systems.

Why investors are racing into robot-training data

Investors are moving fast because the next major bottleneck in AI may not be computing power or model design, but access to high-quality real-world training data. Robots do not learn effectively from text alone. They need examples of how people interact with objects, navigate spaces, and complete everyday tasks.

Mecka’s business is built around that gap. The startup pays participants to record themselves doing ordinary physical tasks using body sensors and smartphones, creating datasets that can be used to teach robots how human movement looks in practice. In the company’s view, this is the robotics equivalent of the data pipelines that helped large AI model companies scale.

Mecka’s founders believe the shortage of physical-world data is one of the main reasons general-purpose robots have been slower to mature than many hoped.

That thesis is attractive to venture firms because it suggests a repeatable, scalable market. If the robotics industry expands the way AI software has, data collection platforms could become essential infrastructure rather than niche service providers.

What exactly does Mecka AI do?

Mecka AI collects and analyzes human motion data so robotics companies can use it to train humanoid robots and other embodied AI systems. The startup uses an “egocentric” data-collection model, meaning it captures experiences from the perspective of the person performing the task, often with sensors and mobile devices rather than expensive industrial equipment.

Its approach is designed to make real-world behavior machine-readable. Rather than relying only on lab demonstrations or teleoperation, Mecka aims to capture everyday physical actions at scale — the kind of motions a robot would need to understand if it were asked to cook, clean, repair, or assist in homes and workplaces.

How the company positions itself in the AI stack

Mecka wants to be for robotics data what companies such as Scale AI have been for large language models: a supplier of the training material that underpins model development. In that sense, the company is less a robot maker than a data infrastructure business.

This positioning matters because the market for robotics data is still young. While there is significant enthusiasm around humanoids and physical AI, many developers still face a shortage of the datasets required to improve performance outside controlled demonstrations. Mecka is betting that the companies solving that problem first will become indispensable partners to the broader robotics ecosystem.

Who founded Mecka AI?

Mecka was founded by four entrepreneurs in 2024, including Canadians Josh Gao and Mogen Cheng, Jason Chong and Duy Nguyen. Gao and Cheng had previously built a restaurant fintech company, while Chong later joined Coinbase after Coinbase acquired his crypto exchange. Nguyen, the only non-Canadian among the founders, handles operations.

The team does not come from a traditional robotics background, but the founders identified a market opportunity through a different lens: they saw that robots were advancing in software faster than they were in real-world understanding. Their response was to focus on the missing layer — human behavior data.

That outsider perspective may have helped them spot a gap that robotics specialists sometimes overlook. In the current market, technical breakthroughs in models and hardware often get the headlines, but the more mundane work of collecting usable training data can be just as critical.

How fast is Mecka growing?

Mecka appears to be growing quickly enough to attract serious investor attention in rapid succession. After its $60 million funding round in the spring, co-founder Josh Gao told Fortune in early June that the company was on track for a $100 million annual run rate by the end of 2026.

That forecast is notable for a business only a year or so removed from founding. It suggests Mecka is not just building a speculative data product, but already securing demand from customers that need robotics training inputs now.

The company has not publicly named its customers, but the broader market gives a clue about who might be buying. Robotics companies, AI labs and other developers working on embodied systems all need more real-world data to improve performance. Mecka’s data offerings fit that demand.

Why revenue expectations matter in this sector

Revenue projections are especially important in robotics data because the sector is still defining itself. If Mecka can sustain strong demand, it may validate the idea that physical training data is becoming a core AI product category rather than a side service.

For investors, the appeal is not just the near-term revenue but the possibility of being early to a platform that becomes embedded in multiple robotics programs. If humanoid robots and other embodied systems gain traction, the data layer could become a strategic chokepoint.

What does ‘egocentric’ data mean?

Egocentric data is information captured from the participant’s own perspective while they perform a task. In Mecka’s case, that can mean sensors and smartphones recording how a person moves as they make coffee, repair a car or complete other hands-on activities.

This kind of footage and sensor data is useful because it reflects how humans actually interact with the world, not just how they appear from an external camera. For a robot, those details can help with understanding hand motion, object manipulation and sequence of actions.

Some robotics teams also use teleoperation, a method in which a human remotely controls a machine to produce training examples. Mecka’s model sits alongside those approaches as part of a broader push to gather physical-world examples at scale.

• Human sensors can capture body movement in detail.
• Smartphones can record task execution in everyday environments.
• Egocentric viewpoints can provide more useful training examples for embodied AI.
• Teleoperation and other physical data methods complement these datasets.

How does Mecka fit into the wider data economy?

Mecka is part of a larger trend in which AI firms are expanding beyond text and images into richer forms of training data. In the language-model era, companies built pipelines to label, refine and supply text. In the robotics era, the equivalent product may be motion data, task demonstrations and physical interaction records.

That transition is drawing comparisons to the rise of human-data companies in the LLM market. The logic is straightforward: whoever gathers the most useful data, at the right quality, can become a critical supplier as downstream model builders scale.

Mecka is not alone in pursuing that opportunity. The market has begun to attract multiple startups trying to package real-world information for robotics labs and AI teams. The competitive environment suggests investors believe the category may be large enough for more than one winner.

Who are Mecka’s rivals?

Mecka faces competition from several startups and data platforms chasing the same robotics opportunity. One closely watched peer is XDOF, which TechCrunch reported last week was also nearing a new financing round, with a valuation of about $1.2 billion. That figure suggests the market has started to assign large premiums to companies that can capture and monetize physical-world training data.

Mecka is also competing indirectly with established human-data firms that are broadening their offerings beyond language-model training. Scale AI and Micro1 are among the companies expanding into related areas, reflecting how quickly the data business is moving from text toward robotics and other embodied applications.

The competitive picture matters because it shows this is no longer an experimental niche. A growing number of firms are trying to become the default supplier of training data for AI systems that operate in the physical world.

Why Sequoia’s involvement matters

Sequoia Capital’s reported role as lead investor gives the deal added weight because the firm remains one of the most influential names in venture capital. Its participation often signals that a category is becoming strategically important, not just technically interesting.

For Mecka, a Sequoia-led round would likely do more than provide capital. It could also increase credibility with robotics customers, researchers and future investors, especially in a market where data infrastructure companies benefit from network effects and trust.

Sequoia declined to comment on the deal, and Mecka did not respond to a request for comment.

That silence leaves some details unresolved, including the exact size of the financing. TechCrunch reported that the terms were not final and could still change, meaning the valuation and round structure remain subject to negotiation.

What comes next for Mecka AI?

The next phase for Mecka will likely depend on whether the company can convert strong investor appetite into durable customer adoption. The robotics data market is promising, but it is also early, fragmented and highly dependent on where embodied AI development goes next.

If humanoid robots and other general-purpose machines gain broader deployment, demand for physical training data could rise sharply. That would make companies like Mecka more central to the AI supply chain, particularly if they can continually refresh datasets and cover a wide range of tasks and environments.

Still, the company will need to prove that its datasets are not only large, but also accurate, diverse and useful enough to improve model performance. In robotics, bad data can be as limiting as no data at all.

Key risks ahead

Despite the optimism, several challenges remain. The sector must still determine which data formats are most valuable, how to standardize collection across environments and whether buyers will pay enough for recurring access to support premium valuations.

1. Robotics training needs are still evolving.
2. Competition is increasing quickly.
3. Data quality and consistency are hard to maintain at scale.
4. Customer concentration could become a risk if the market remains narrow.

Why this deal is part of a larger AI pattern

Mecka’s rise fits a familiar pattern in AI markets: when one layer gets crowded, investors move to the next bottleneck. First came model builders, then tool providers, then data suppliers, and now the physical-world equivalents are starting to emerge for robotics.

That pattern helps explain why a startup founded in 2024 can quickly attract a valuation near half a billion dollars. The bet is not only on Mecka itself, but on the idea that embodied AI will require an industrial-scale data economy just as language models did.

For now, the reported deal puts Mecka among the most closely watched newcomers in robotics infrastructure. Whether it becomes a category-defining company will depend on how fast the robotics market grows — and on whether high-quality human motion data proves to be the key ingredient investors think it is.

Timeline of Mecka AI’s rapid rise

Mecka’s brief history shows how fast enthusiasm around robotics data is building. The company has gone from startup formation to major financing talks in a matter of months.

As the robotics industry searches for the next unlock, Mecka’s bet is clear: the road to better robots runs through better data about how humans actually move, work and solve everyday problems.