Origin Lab secures $8M to bridge video game data gap for AI world models
The startup will operate a marketplace connecting gaming studios with AI laboratories, converting digital assets into licensed training sets for robotics and spatial simulation.
Origin Lab has closed an $8 million seed funding round led by Lightspeed Ventures, establishing a new marketplace designed to connect video game companies with artificial intelligence laboratories. The platform aims to monetise existing digital assets from the gaming industry by selling high-quality, licensed data to developers building world models for robotics and physical space simulation.
Unlike large language models, which benefit from abundant text data, world models require complex information on how physical objects move and interact. Co-founder and co-CEO Anne-Margot Rodde stated that the venture bridges a significant infrastructure gap, noting that the video game industry sits on valuable data but historically lacked a mechanism to share it with AI developers. Origin Lab will convert these assets into training formats, ranging from simple rendering runs to automated walkthrough footage.
The funding round included participation from SV Angel, Eniac, Seven Stars, and FPV. Angel investors in the round include Twitch co-founder Kevin Lin and Cruise founder Kyle Vogt. Potential buyers for the data include labs such as Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs, which are actively seeking to assemble the necessary datasets for their physical-world simulations.
The initiative addresses long-standing licensing and quality hurdles that have previously hindered the use of video game footage for AI training. Interest in such data has grown following scrutiny of OpenAI’s Sora model in December 2024, which appeared to regurgitate footage from games and Twitch streams. Amazon has also publicly expressed interest in utilising Twitch footage to train its own models.
Faraz Fatemi, a partner at Lightspeed Ventures, highlighted the commercial viability of the sector, citing the success of data vendors like Scale.AI as a precedent. He noted that the revenue scaling potential for companies supplying data to major labs is significant, as these well-capitalised businesses face a persistent bottleneck in acquiring high-quality training material.
