Config Robotic Data Platform Raises 27 Million Dollars In Seed Funding To Expand Foundation Model Data Operations And Scale Industrial AI Applications
Config has secured $27 million in a seed funding round, which Samsung Venture Investment led, for its specialized data layer designed for robotic foundation models. According to TechCrunch interview, the deal gives the Seoul based and San Jose based startup a valuation north of $200 million. The company also got backing from venture arms of major Asian manufacturing companies such as Hyundai Motor ZER01NE Ventures, LG Tech Ventures, and SKT America, as well as Mirae Asset Ventures Korea Development Bank GS Futures Kakao Ventures and Z Ventures and angel investor Pieter Abbeel.
The startup was founded in January 2025 by CEO Minjoon Seo, a former researcher at Meta and chief scientist at Twelve Labs, and its founding team includes researchers from Waymo, Google, and Naver. While companies such as this may manufacture hardware, the company says that it focuses on what it sees as the prerequisite raw material for machine learning, specifically the data which robots will use to interact with the world around them.
"To make it clear, for robotics AI, the amount of investment necessary to create the models is much higher than software chatbots," Seo told TechCrunch. "This is because robots need physically acquired data using actual robots and specialized environments."
The startup hopes to be the "neutral supplier of data" for the robotic industry in much the same way TSMC does for the semiconductor industry. Config will provide data layers to manufacturers and system integrators who want to build their own robotic AIs instead of sourcing them from third parties. Currently, the company's clients are found in the defense industry, agriculture industry and industrial sector, with competitors including Physical Intelligence, Skild AI and Generalist AI.
The company employs around 300 people to produce data across operations centers in Seoul and Hanoi, and it claims to have collected over 100,000 hours of human motion data. This dataset is more than 30 times bigger than AgiBot World, the largest public equivalent with nearly 3,000 hours. One of Config's key technological differentiators is that it says that it can transform the data before it even reaches training stages to better align how robots interact with their environments.
The new $27 million will be used to massively scale data operations in Vietnam and South Korea, to reach 1 million hours of data collection, and expand the enterprise platform business to $10 million ARR by the end of 2027. Config will also be releasing a cloud based Robot as a Service, which will enable companies to take advantage of its foundational models over the cloud without requiring their own powerful on board hardware.
Minjoon Seo emphasized that his company will focus on converting the data itself, not the models, to achieve desirable results.
"Training a robot using raw human data is no different than teaching a new language by using only materials that comes from another language," he said.
With a processed, converted data layer, Config plans to enable a lower barrier to entry for industrial businesses who want to integrate physical AIs into their existing supply chains.
