As artificial intelligence begins to interact with the physical world, a new bottleneck has emerged: training data for world models. Unlike large language models, which can draw from vast text corpora on the internet, physical AI systems require data that captures how objects move, interact, and behave in real space. That data has proven difficult to source at scale. Now, a startup called Origin Lab is proposing an unexpected solution: the video game industry.
Origin Lab announced an $8 million seed funding round led by Lightspeed Ventures, with participation from SV Angel, Eniac, Seven Stars, and FPV. Angel investors include Twitch co-founder Kevin Lin and Cruise founder Kyle Vogt. The company plans to act as a marketplace connecting video game studios with AI labs building world models, such as Yann LeCun’s AMI Labs or Fei-Fei Li’s World Labs.
Also read: Anthropic's Cat Wu on staying ahead in AI: 'We don't think about competitors'
Bridging two industries
The core premise is straightforward. Video game companies have spent years creating detailed digital environments with realistic physics, object interactions, and movement patterns. That data, Origin Lab argues, is precisely what world-model builders need. However, licensing and data formatting have historically made such collaborations difficult.
Origin Lab co-CEO and co-founder Anne-Margot Rodde explained the company’s approach: “The AI systems that are being built now need to understand how the physical world works and how things move. That data essentially lives in video games.” The startup will convert game assets into usable training data, ranging from simple rendering runs to automated walkthrough footage that captures hours of physics-consistent interactions.
“It became clear that the video game industry was sitting on some incredibly valuable data, but there was no real way or infrastructure to basically connect AI labs and the video game industry,” Rodde said. “So essentially, we built that bridge.”
A growing market for data vendors
The funding round signals growing investor confidence in the market for specialized AI training data. Faraz Fatemi, a partner at Lightspeed who led the investment, drew comparisons to companies like Scale AI. “We’ve seen how sharp the revenue scaling can be for data vendors that are serving the major labs,” Fatemi said. “These are very well-capitalized businesses, and the bottleneck for all of them is data.”
The need for clean, licensed data has become more pressing following recent controversies. In December 2024, OpenAI faced criticism when its Sora video-generation model appeared to reproduce footage from video games and streamers, raising questions about training data provenance. Amazon has also expressed interest in using Twitch footage for model training, highlighting the industry’s hunger for game-derived data.
Why this matters for the AI industry
World models represent a frontier in AI research, aiming to give machines an understanding of physical space and causality. Unlike text-based AI, which can be trained on the open web, world models require structured data that captures motion, collision, gravity, and object permanence. Video games offer a controlled, annotated environment where these properties are already simulated.
For video game companies, the arrangement offers a new revenue stream from existing assets. For AI labs, it provides a legally clear path to high-quality training data. Origin Lab positions itself as the intermediary that handles conversion, licensing, and quality assurance.
Conclusion
Origin Lab’s $8 million seed round reflects a maturing understanding of the data challenges facing physical AI. As world-model builders race to train systems that can operate robots, deal with spaces, and understand physical causality, the video game industry may become an unlikely but essential supplier. The startup’s success will depend on its ability to negotiate licensing agreements, ensure data quality, and scale its conversion pipeline. But the investment from Lightspeed and prominent angel investors suggests confidence in the model.
FAQs
Q1: What is a world model in AI?
A world model is an AI system designed to understand and simulate the physical world, including object movement, interactions, and spatial relationships. Unlike language models, world models require training data that captures real-world physics and behavior.
Q2: Why is video game data useful for training world models?
Video games contain detailed digital environments with realistic physics, object interactions, and movement patterns. This data is already structured and annotated, making it valuable for training AI systems that need to understand physical space.
Q3: How does Origin Lab plan to make money?
Origin Lab will act as a marketplace, connecting video game companies with AI labs. It will charge fees for converting game assets into usable training data and for facilitating licensing agreements between the two parties.
