As artificial intelligence systems increasingly move towards interacting with the physical world, a new demand has emerged for sophisticated 'world models.' These models are crucial for operating physical robotics and accurately simulating objects within physical space. However, unlike large language models which benefit from vast readily available text datasets, world models face a significant challenge: a scarcity of suitable, easily accessible training data. This data gap has left numerous research labs scrambling to compile the necessary datasets to advance their work.

Addressing this critical need, a new startup named Origin Lab has emerged with an innovative and somewhat unexpected solution: leveraging the rich digital environments of the video game industry. Origin Lab recently announced an $8 million seed funding round, spearheaded by Lightspeed Ventures, with significant participation from SV Angel, Eniac, Seven Stars, and FPV. Angel investors included prominent figures such as Twitch co-founder Kevin Lin and Cruise founder Kyle Vogt, underscoring the potential and industry confidence in Origin Lab's approach.

According to Anne-Margot Rodde, co-CEO and co-founder of Origin Lab, the fundamental premise is clear: "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." Alongside Rodde, the company's other co-founders are Antoine Gargot and Colin Carrier, forming a team focused on bridging this vital data divide. In essence, Origin Lab functions as a specialized marketplace, enabling world-model-focused laboratories, such as Yann LeCun’s AMI Labs or Fei-Fei Li’s World Labs, to acquire high-quality, licensed data.

On the supply side, video game companies stand to gain an additional revenue stream by monetizing the extensive digital assets they have already created. Origin Lab's pivotal role in the middle involves converting these complex video game assets into a format that is directly usable as AI training data. This conversion process can range from simple rendering runs to more intricate tasks like automating hours of walkthrough footage, ensuring the data meets the stringent quality and format requirements for AI model training.

Rodde emphasized the untapped potential, stating, "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. So essentially, we built that bridge." While AI labs have long recognized the value of video game footage as a potential data source, significant hurdles related to licensing and data quality have historically impeded widespread adoption. A notable incident in December 2024, where the initial version of OpenAI’s Sora video-generation model appeared to reproduce footage from popular video games and streamers (likely due to training on Twitch streams), highlighted the imperative for properly licensed and curated data.

Origin Lab's successful fundraising is indicative of a burgeoning market—not merely for raw training data, but for specialized startups capable of acting as crucial suppliers to major AI development labs. Faraz Fatemi, a partner at Lightspeed who led the investment in Origin Lab, drew parallels to the success of companies like Scale.AI, noting, "We’ve seen how sharp the revenue scaling can be for data vendors that are serving the major labs. These are very well-capitalized businesses, and the bottleneck for all of them is data." This sentiment reinforces Origin Lab's position as an essential player in the evolving landscape of AI infrastructure.