Z Venture Capital

Investing in Config Intelligence

We are living through the GPT-2.5 moment for robotics. The hardware is ready. The model architectures are maturing. What’s missing is data—and without it, the scaling law that unlocked the LLM revolution cannot take hold in the physical world. Config Intelligence is the company we believe will bridge that gap.

Config is a full-stack data-to-model platform for Robot Foundation Models (RFMs). Its proprietary Action Estimator converts cheap human video into high-precision robot action data at micrometer accuracy—at a dramatically lower cost versus traditional teleoperation. The CFG-1 foundation model, trained on 100K+ hours of in-house data, is hardware-agnostic and already generating revenue.

Robot action data doesn’t exist on the internet and is mostly collected via teleoperation—making scale nearly impossible. Config’s Action Estimator solves this by transforming inexpensive human video into VLA data at ~500-micrometer precision. The flywheel this creates is powerful: more data trains a stronger CFG-1, which attracts more enterprise customers, which generates more data. This is the infrastructure that finally makes the robotics scaling law achievable.

Global leaders like Physical Intelligence and Covariant have all reached the same conclusion: end-to-end ownership of data, model, and deployment is essential. Config does this at a fraction of the cost. Its multi-stage automated pipeline verifies whether data quality actually translates to robot performance via its own RFM—a “loop closure” capability pure-play data providers cannot match. This underpins a clear business model progression: from robotics data platform to RaaS business.

Config has secured commercial contracts with major conglomerates including Samsung, LG, and Hyundai—generating revenue from its first year of operation. These aren’t pilots; robotics is among these conglomerates’ highest corporate priorities, and they need massive, diverse action datasets they cannot build themselves. Korea, with the world’s highest robot density and a world-class manufacturing base, is the ideal proving ground. Deployments in real factory environments create proprietary datasets that global competitors cannot replicate from afar.

The RFM and data infrastructure market is the fastest-growing segment of Physical AI, projected at 50%+ CAGR toward $18B+ by 2030. The window to build a data moat is open—and narrowing.

Config’s founding team is uniquely equipped to seize it:

・CEO Minjoon Seo (KAIST, Meta FAIR, 2,400+ citations)

・CTO Hyungmok Son (Harvard PhD, Waymo)

・Chief Scientist Kimin Lee (KAIST, Google Brain, creator of the Decision Transformer)

・COO Jack Bang (UBS, Glenwood, Socra AI)

We believe Config is the company that will unlock the scaling law for robotics—starting from Korea’s industrial heartland, where the partnerships are real, the data is flowing, and the foundation model is already deployed. That’s why we invested.

・Website：config.ai

・Founding Team: Minjoon Seo (CEO), Hyungmok Son (CTO), Kimin Lee (Chief Scientist), Jack Bang (COO)

・Media：https://techcrunch.com/2026/05/11/koreas-biggest-manufacturers-back-config-the-tsmc-of-robot-data/