Robot Data Firm Config Raises $27M Seed Funding

Config, a startup aiming to standardize the creation of training data for robotic foundation models (RFMs), has successfully closed a $27 million seed funding round. The investment, which was oversubscribed and values the company at over $200 million, was led by Samsung Venture Investment. This significant capital infusion positions Config as a critical infrastructure provider in the rapidly evolving robotics sector, drawing parallels to semiconductor foundries like TSMC that supply essential components without directly competing with chip designers.

The company's strategic vision is to become the indispensable data supplier for the burgeoning field of robot AI. Unlike large language models that leverage existing internet text, robot AI requires meticulously collected, real-world interaction data. This process is inherently resource-intensive, demanding physical robots, specialized facilities, and skilled personnel. Config addresses this bottleneck by focusing on producing and refining this crucial data, enabling developers of robotic systems to accelerate their model training and deployment.

Config's impressive roster of investors underscores the market's confidence in its approach. Alongside lead investor Samsung Venture Investment, the round saw participation from prominent strategic and financial backers including ZER01NE (the venture arm of Hyundai Motor Group), LG Technology Ventures, SKT America, Mirae Asset Venture Investment, Korea Development Bank, GS Futures, Kakao Ventures, and ZVC. Noteworthy angel investment also came from Pieter Abbeel, a distinguished professor at UC Berkeley and co-founder of Covariant AI.

Founded in January 2025 by CEO Minjoon Seo, who brings a wealth of experience from his tenure as an AI researcher at Meta and chief scientist at Twelve Labs, Config boasts a founding team with deep expertise from companies like Waymo, Google, and Naver. The company currently operates data collection facilities in Seoul and Hanoi, Vietnam, employing nearly 300 individuals. To date, Config has amassed over 100,000 hours of human motion data, a substantial volume compared to existing open-source datasets, and is actively working to expand this to one million hours.

Config's core technological advantage lies in its proprietary methods for transforming raw human motion data into a format optimized for robotic learning. This "data conversion" technology, as described by CEO Seo, is crucial for bridging the gap between human actions and robot capabilities, akin to translating between languages. This focus on data transformation, rather than model development, allows Config to serve a wide array of RFM developers without direct competition, a strategy that has already begun generating revenue from clients in manufacturing, agriculture, and defense sectors.

The fresh capital will fuel Config's ambitious growth plans, including expanding its data collection operations, scaling its enterprise platform towards a $10 million ARR target by the end of 2026, and launching a cloud-based Robot-as-a-Service (RaaS) offering. This move into RaaS will democratize access to advanced robotic AI training, allowing customers to leverage Config's foundation models without requiring significant on-premise hardware investments. The company's success highlights a critical trend in the AI hardware and software ecosystem: the increasing demand for specialized, high-quality data infrastructure to power the next generation of intelligent machines.