XDOF and Mecka Chase the Same Robot-Data Prize

Two companies that pay humans to perform ordinary tasks so a robot can eventually learn to copy them are being priced by investors within days of each other this month, at valuations that assume the same unproven business scales the way data-labeling did for language models.

XDOF is in late-stage talks for a Series B at a roughly $1.2 billion valuation led by 8VC, TechCrunch reported Sept. 4, three months after the Berkeley-founded startup's Series A closed. Mecka AI is nearing a roughly $500 million valuation in a round led by Sequoia Capital, TechCrunch reported Sept. 11, also about three months after its own prior round. Neither deal is final. Both are explicitly described by TechCrunch as subject to change before signing.

The business both companies are selling

XDOF, founded in 2024 by UC Berkeley researchers Philipp Wu and Fred Shentu, builds data pipelines, collection tools and annotation systems that let frontier AI labs and robotics companies buy real-world robot-interaction data instead of building the infrastructure to gather it themselves. Its collectors combine remote teleoperation with a low-cost control rig called GELLO, and the company is assembling a dataset it calls ABC, described in its own materials as the largest collection of high-quality robot training data yet assembled, in partnership with UC Berkeley's AI research lab.

Mecka AI, also founded in 2024, takes a lower-tech approach: it pays people to record themselves performing everyday tasks, such as making coffee or fixing a car, using body sensors and smartphones. TechCrunch reported the company positions itself as doing for robotics what Scale AI, Mercor and Surge have done for large language models, a comparison Dealroom's own coverage of XDOF makes explicit in the other direction, naming Mecka AI and Scale AI as XDOF's closest rivals.

The numbers do not agree with each other

How big XDOF's own Series A actually was depends on which outlet is read. TechCrunch, in the same September article and in a companion piece carried on Yahoo Finance, put the round at $70 million, with Thrive Capital, Andreessen Horowitz, Lux Capital and Spark Capital participating. Dealroom's independent write-up of the same financing lists it at $45.3 million. Neither outlet has corrected the other, and this article could not resolve which figure is accurate.

Mecka's numbers carry less ambiguity, because only one outlet has reported them: a $60 million round led by Framework Ventures, with Menlo Ventures, SV Angel and Kindred Ventures participating, closed roughly three months before the Sequoia talks. As of early June, according to TechCrunch, Mecka's own founders were projecting the company would exit 2026 at a $100 million annualized run rate. XDOF's current run rate is closer to $50 million annualized, TechCrunch reported, with about 20 paying customers that include frontier AI labs.

What the same-week timing implies

That two data-supply startups founded the same year are being re-priced within a week of each other, by different lead investors, is itself informative: it means the capital chasing this niche is large enough, and impatient enough, to bid up more than one company at once rather than consolidating around a single winner. Robotics investment broadly is concentrated rather than broad based. The Robot Report tracked $4.87 billion in disclosed robotics funding across 162 deals in August 2026, of which the two largest deals, Unitree's Shanghai IPO and XPeng's robotics unit raise, accounted for more than $1.8 billion between them. Against that backdrop, two data-supply startups pulling in a combined valuation approaching $1.7 billion in one week is a bet that the bottleneck for robot capability is data, not hardware or deployment.

What would prove the bet wrong

Neither company has published evidence that the human-recorded data it sells actually improves a deployed robot's performance, in the way that internet-scale text demonstrably improved language models. TechCrunch quoted the company's own framing of the problem: "Unlike LLMs, which initially trained on the entirety of the internet, physical robots don't have an equivalent real-world dataset to draw from." That is a description of a gap, not proof that either company's data closes it. If a robotics customer publicly credits XDOF's or Mecka's data with a specific capability jump, the bet looks vindicated. If the Series B and the Sequoia round both close months late, or below the reported numbers, at the reported valuations, that would suggest the same investor enthusiasm now bidding up two competitors at once cooled before either raise reached a signature.

Metric	XDOF	Mecka AI
Founded	2024	2024
Prior round	$70M (TechCrunch) / $45.3M (Dealroom)	$60 million
Prior round lead	Thrive Capital	Framework Ventures
New round in talks	About $1.2 billion, led by 8VC	About $500 million, led by Sequoia
Reported annualized revenue	About $50 million	$100 million (2026 target, per founders)
Data method	Teleoperation plus GELLO rig	Body sensors and smartphones

Both companies now sit inside the same funding wave that has lifted SoftBank's robotics bet on the Robotics and AI Institute and pushed Cornelis Networks to a $205 million raise for the infrastructure underneath it. The next concrete marker for either company is a signed term sheet: until one appears, both valuations remain what TechCrunch called them, numbers that are "not final and could still change."