Mecka AI Nears $500M Valuation as Sequoia Makes Another AI Bet on Robotics

Sequoia Capital is reportedly leading a new funding round for Mecka AI at a valuation of about $500 million, putting the young robotics-data company within striking distance of the $1 billion mark to become a unicorn.

Mecka AI is back in the fundraising market only months after raising $60 million, with Sequoia Capital now lined up to lead a deal that could value the company at roughly $500 million, according to two people familiar with the negotiations. The size of the new round has not been disclosed, and the terms could still change before the deal closes.

The proposed financing would be another sharp step up in a remarkably short period for Meck AI as the startup announced its $60 million financing in June, with Framework Ventures leading and Menlo Ventures, SV Angel and Kindred Ventures among the participating investors.

The bigger story may be what Sequoia sees in the company.

Sequoia is betting on the data behind robots

Mecka AI is not building humanoid robots itself. Its business is built around a less visible piece of the robotics race: collecting the human movement data that can be used to train machines.

The company pays people to record ordinary activities, including tasks such as making coffee and repairing cars, using smartphones and body sensors. That information can then be processed into data for robotics companies and AI labs working on systems that need to understand how humans move and interact with physical objects.

That puts Mecka in a part of the AI market where investors are increasingly looking for businesses that can supply the raw material needed by model builders.

The company has not publicly named its customer list, but its founders have argued that real-world data remains a major constraint for general-purpose robotics.

Sequoia has been putting serious money behind AI and physical-world computing this year.

Its portfolio already includes robotics AI companies such as Skild AI and Physical Intelligence, while the firm has also backed Waymo and recently announced a $100 million investment in battery maker Form Energy aimed at the power demands surrounding the AI economy.

Against that backdrop, a Mecka investment would look less like a one-off robotics bet and more like another piece of Sequoia's wider AI strategy.

From $60 million to $500 million territory

Mecka AI's last publicly reported financing was substantial on its own. Fortune reported in June that the company had raised $60 million across a $25 million Series A completed in November and a $35 million follow-on investment. Framework Ventures led both financings.

At the time, CEO Josh Gao said Mecka was projecting a $100 million annual run rate by the end of 2026 based on signed contracts. The company had about 40 employees and was expanding the systems it uses to gather physical-world data.

A valuation near $500 million would give those growth claims a much higher price tag, especially for a startup founded in 2024.

It would still be a long way from a unicorn valuation. But for a company that has only been operating for about two years, getting close to the halfway point on the road to $1 billion would be a notable funding milestone.

Four Founders Without a Robotics Pedigree

The team behind Mecka did not come out of Stanford’s robotics lab or Google DeepMind. Josh Gao and Mogen Cheng sold a restaurant payments startup in 2023. Jason Chong sold a crypto exchange to Coinbase. Duy Nguyen made money reselling sneakers before running an agency.  They knew each other through friends, read papers for months, visited labs, and landed on a bet that most robotics researchers had treated as a research curiosity rather than a commercial plan: train robots on data captured from humans, not from teleoperation.

Teleoperation produces clean data tied to a specific machine’s joints and movements. But it is slow. Every hour of training data requires an operator and a robot. Mecka flipped the model. It sends smartphones and body sensors to contributors, pays them to perform tasks, and then processes the footage into labeled episodes that robotics labs can use. The approach treats billions of people as potential teachers. 

The translation problem is real. A human demonstration captures hands, objects, movement and context. A robot needs force readings, joint states and action labels. Mecka has expanded beyond raw collection into processing, evaluation and deployment support, positioning itself between labs, hardware makers and workplaces that want robots doing defined jobs.

The Cost Curve Investors Are Watching

The margin question is where the business lives or dies. Video is expensive. Before a robot can learn from a person folding laundry, hours of footage have to be encoded, split, stored, labeled and verified. Most companies in this space run on general-purpose GPUs by default and treat the cost as a fixed line item. 

Mecka hired Basil Yusuf from Google in March to own that cost curve. He moved production video encoding off GPUs onto dedicated video-transcoding chips and rebuilt the pipeline around them. Encoding an hour of video now costs roughly a quarter of what it did on the GPU path, and the savings compound downstream because labeling, verification and delivery all consume processed video. 

That engineering work matters more than any single product announcement. A data company’s entire margin structure rests on the cost of turning raw footage into usable training material. If Mecka can keep that cost falling faster than the price it charges, it has a business. If it cannot, it has a video storage company with a robotics pitch.

Mecka is not alone, as XDOF, which is another robot-training-data startup, was reported last week to be nearing a round at a $1.2 billion valuation. Scale AI and Micro1 are pushing beyond language models into physical data.  The category is filling up fast, and investors are placing bets before a clear winner emerges.

The race for robot training data is getting crowded

Mecka is entering a market where other startups are chasing the same shortage of useful physical-world data.

TechCrunch recently reported that XDOF was nearing a financing that could value it at $1.2 billion. Companies such as Scale AI and Micro1 are also expanding their human-data businesses beyond the language-model market and into areas tied to physical AI and robotics.

That competition could make the quality, variety and ownership of training data more important as robotics companies push toward systems capable of operating outside controlled demonstrations.

Mecka's approach is centered on human activity captured from a first-person perspective, rather than relying solely on teleoperation or data generated directly from robot fleets. The bet is that what humans naturally do in the physical world can provide useful material for teaching machines how to behave in it.

A Sequoia bet that could change Mecka's trajectory

There is still no completed deal to announce. Mecka has not commented on the proposed financing, and Sequoia declined to comment to TechCrunch. The reported $500 million valuation is also not final.

Yet the reported terms are revealing on their own.

Sequoia is making another sizable bet on AI at a time when investors are looking beyond chatbot companies and toward the infrastructure, data and physical systems needed to make AI useful outside the screen. Mecka sits directly in that middle layer.

That makes this more than a fresh funding round for a two-year-old startup. It could be another Sequoia AI bet that pushes Mecka much closer to unicorn status, provided the company can turn its growing demand for human-motion data into the revenue scale investors are already expecting.

For now, the $500 million figure is still a deal under negotiation. If it closes around that level, Mecka will have moved a long way in a very short time.