Skip to content

Topic

Robot Training Data Pipelines

Collection, verification and refinement of unstructured robot interaction data as the input constraint on embodied AI.

Current stories

leadershipOne report1 publisher

Microagi's Shift offers free private chefs in San Francisco in exchange for footage of their work

Microagi's CEO says the company pays more than 35,000 people in over 15 countries to film physical tasks, Business Insider reported. Robotics is short of training data, so startups are buying footage of skilled work such as private cooking.

Reality

Evidence55
Adoption50
Hype gap+25
Incentives70
Confidence60
investConfirmed6 publishers

Mecka raises $60 million Series B led by Sequoia; prior ~$500M valuation talk not tied to this round

Sequoia led a $60 million Series B in Mecka AI, joined by Nvidia, Qualcomm and Microsoft's M12. TechCrunch's reported $500 million valuation is at most five times the $100 million-plus run-rate Crypto Briefing cites, a low price for data sold as scarce.

Perspective Coverage

7 publishers
Builder
Builder 34%
Operator
Operator 19%
Investor
Investor 47%

Reality

Evidence55
Adoption45
Hype gap+30
Incentives60
Confidence60
leadershipOne report1 publisher

OpenAI's highest robotics salary band goes to the engineer who moves the training data

OpenAI's careers page now lists 27 robotics openings, up from 11 in May, with published base pay running from $177,000 to $500,000. The top of that range attaches to data infrastructure work, and the robotics software job tops out at $325,000.

Publishers:entrepreneur.com

Reality

Evidence58
Adoption
Not scored
Hype gap+30
Incentives62
Confidence52
investOne report1 publisher

China's robot data centres are outrunning their buyers

Twenty-two provincial innovation centres opened within 14 months, and Wuhan alone logs 24,000 training entries a day. But Beijing's Shijingshan centre dropped its robot partner over weak data sales, so subsidy is still paying the bill.

Publishers:en.sedaily.com

Reality

Evidence45
Adoption55
Hype gap+20
Incentives70
Confidence45
leadershipOne report1 publisher

OpenAI expands robotics hiring, with one engineer role paying up to $500,000

OpenAI has more than doubled its robotics listings since May and is hiring people to design joints, run test labs and handle the legal work. The published pay is the part rival recruiters will notice.

Reality

Evidence66
Adoption12
Hype gap+10
Incentives68
Confidence62
leadershipOne report1 publisher

An Airbus SVP would split the robot stack into three models under independent safety controls

Greg Ombach, a senior vice president at Airbus, writes that vision-language-action models force a choice between capability and edge deployment, and that industrial buyers should validate one operating domain before expanding.

Publishers:forbes.com

Reality

Evidence30
Adoption
Not scored
Hype gap+15
Incentives68
Confidence40
investOne report1 publisher

UBTECH's 10,000-robot year in Liuzhou fits inside a single shift

UBTECH put a capacity number on the record for its new Liuzhou humanoid plant and left out the price, the shipments and the order book, so the cost case rests on its claim about Chinese component supply.

Publishers:en.sedaily.com

Reality

Evidence36
Adoption30
Hype gap+33
Incentives76
Confidence46
investOne report1 publisher

Robot makers buy the inference before they sell the robot

Where a robot's inference runs decides what the machine costs to build. SemiAnalysis puts the industry's upfront compute bill in the trillions at a billion robots, paid by manufacturers on every unit before it does any work.

Reality

Evidence52
Adoption30
Hype gap+18
Incentives55
Confidence45
productOne report1 publisher

Sequoia-led round values Mecka AI near $500M for footage of people making coffee

Mecka AI pays people to wear sensors while they make coffee or fix cars, and two people told TechCrunch that Sequoia is leading a round at about $500 million, three months after a $60 million round. The terms are not final.

Publishers:techcrunch.com

Reality

Evidence38
Adoption
Not scored
Hype gap+30
Incentives68
Confidence35
buildOne report1 publisher

Figure sizes its humanoid training run at up to 1,389 Vera Rubin racks

Figure is buying compute the way a frontier lab buys it. The announcements skip the numbers that matter most: how much of the ceiling the initial money covers, and how many megawatts Barstow will draw.

Publishers:runtimewire.com

Reality

Evidence50
Adoption20
Hype gap+38
Incentives76
Confidence57
buildOne report1 publisher

Lightwheel lets its downloaders shape the next 90,000 hours of egocentric footage

The 10,000 hours released on August 26th arrive in LeRobot v3 and MCAP with licensing set subset by subset, while the remaining 90,000 land in stages on no published date. The plannable corpus is the small one.

Publishers:runtimewire.com

Reality

Evidence38
Adoption20
Hype gap+22
Incentives78
Confidence44
buildOne report1 publisher

Figure is buying its data moat by the minute, and it just published the price

Index pays crowd contributors for first-person household video. Figure's own figures put contributor spend near $3.75m a month, and they do not reconcile into an hourly wage.

Publishers:runtimewire.com

Reality

Evidence32
Adoption54
Hype gap+34
Incentives82
Confidence46
leadershipOne report1 publisher

Robotics won the funding argument. The expo hall shows it has not won the hardware one.

Actuate went from 250 attendees in 2024 to 1,500 this year and could have sold 2,000. Almost none of them brought a working robot, which tells you where the bottleneck sits.

Reality

Evidence55
Adoption45
Hype gap+30
Incentives70
Confidence52
investOne report1 publisher

LG and Nvidia put a number on physical AI: 100,000 hours of robot data in one year

LG Electronics says it will train on 100,000 hours of robot data inside the year at its Seoul DataFactory. The target turns a chairman-level MOU into a fleet purchase and a floor plan.

Publishers:en.sedaily.com

Reality

Evidence30
Adoption20
Hype gap+40
Incentives78
Confidence40