Korea’s top manufacturers back Config, the TSMC of robot data
▼ Summary
– Config, a startup building a data layer for robotics foundation models, raised a $27 million seed round led by Samsung Venture Investment, with strategic backing from Hyundai, LG, and SKT.
– Founded in January 2025 by ex-Meta researcher Minjoon Seo, Config focuses on providing training data for robots rather than building robots themselves.
– Robotics AI development is costly because training data must be physically collected via robots and human operators, unlike easily sourced text for language models.
– Config aims to be a neutral data supplier for robot AI, similar to TSMC’s role in chip manufacturing, as large manufacturers seek to build proprietary robot AI.
– The startup has over 100,000 hours of human motion data, uses a proprietary data conversion process before training, and plans to scale to 1 million hours and $10 million ARR by 2027.
Asia’s manufacturing giants are channeling their industrial strength into a new frontier: physical AI. In nations like South Korea, Japan, China, and Taiwan, large-scale production and export-led growth have long defined economic success. Now, that same structural foundation is reshaping the adoption of artificial intelligence, steering investment toward companies that can bridge the gap between hardware and intelligent automation.
That shift makes the latest funding round for Config particularly telling. The startup, headquartered in both Seoul and San Jose, is building the critical data infrastructure for robotics foundation models (RFMs). It has just secured backing from the venture arms of some of South Korea’s most prominent manufacturers.
Samsung Venture Investment led an oversubscribed $27 million seed round, valuing Config at over $200 million. This brings the company’s total capital raised to $35 million. Additional strategic investors include Hyundai Motor’s ZER01NE Ventures, LG Technology Ventures, and SKT America, the VC unit of South Korea’s telecom giant. Other notable backers include Pieter Abbeel (a UC Berkeley professor and co-founder of Covariant) and financial investors such as Mirae Asset Ventures, Korea Development Bank, GS Futures, Kakao Ventures, and Z Ventures.
Founded in January 2025 by CEO Minjoon Seo, a former Meta researcher and chief scientist at TwelveLabs, Config brings together co-founders with experience at Waymo, Google, and Naver. The company’s mission is deliberately focused: rather than building robots, it supplies the data those machines need to learn and operate effectively. The core belief is that superior data is the key to unlocking more useful robots.
Training large language models is expensive due to computing demands, but the raw material , text scraped from the internet , is abundant. Teaching robots to move is a fundamentally different challenge, Seo explained in an exclusive interview. Every piece of training data must be physically collected. That requires a robot, a facility to run it, and people to operate it. As a result, robotics AI is far more costly to develop than software-only chatbots. As companies pursue more capable robots, the expense of gathering and labeling data escalates quickly.
Config aims to be the company that makes everyone else’s robot AI possible. The startup compares its role to TSMC, the Taiwanese chipmaker that manufactures for Apple, Nvidia, and AMD without competing with any of them. Config wants to play a similar role in robotics by supplying data. This approach is gaining traction as large manufacturers increasingly seek to build their own proprietary robot AI rather than relying entirely on outside vendors , precisely the market Config is betting on.
The company is already generating revenue, according to COO and co-founder Jack Bang. Current customers include large manufacturers, system integrators, and companies in the agriculture and defense sectors. Competitors in the space include Physical Intelligence, Generalist AI, and Skild AI.
Config records humans performing physical tasks in controlled studio environments and in the field. The startup operates out of Seoul and Hanoi, where a workforce of nearly 300 people handles data production. To date, it has accumulated over 100,000 hours of human motion data, more than 30 times the size of AgiBot World, the largest comparable open source dataset at roughly 3,000 hours.
Most robotics teams train AI models on human motion data and then adapt those models for a robot. Config is taking a different approach, Seo said. The company focuses on transforming the data before training begins so it is better suited to the way robots move and interact with the world. Seo compared the process to language translation. Training a model on one type of data and expecting it to work seamlessly in another setting, he said, is like trying to teach Korean using only English-language materials.
“The data must be converted, not the model. This conversion technology is Config’s core technical differentiator,” Seo said.
The new funding will support three priorities: scaling its data operation in Vietnam and Seoul toward 1 million hours of collected data, growing its enterprise platform business to $10 million in ARR by the end of 2027, and launching a cloud-based robot-as-a-service product that lets companies run Config’s foundation model without requiring onboard hardware.
(Source: TechCrunch)