As artificial intelligence starts to engage with the physical environment, innovative labs are developing world models intended for controlling physical robotics or simulating objects in tangible space. Unlike large language models, acquiring data for these models is not straightforward, leaving numerous labs in a rush to compile the needed training datasets.

Currently, a startup is emerging with an unexpected data source: the gaming industry.

This is the concept behind Origin Lab, which has just revealed an $8 million seed funding round, spearheaded by Lightspeed Ventures. SV Angel, Eniac, Seven Stars, and FPV also joined in, along with angel investments from Twitch co-founder Kevin Lin and Cruise founder Kyle Vogt.

“The AI systems currently being developed need to grasp how the physical world operates and how objects move,” co-CEO and co-founder Anne-Margot Rodde stated to TechCrunch. “That information essentially resides in video games.” The company’s other co-founders (shown above) are Antoine Gargot and Colin Carrier.

In essence, Origin Lab will function as a marketplace where labs focused on world modeling, such as Yann LeCun’s AMI Labs or Fei-Fei Li’s World Labs, can acquire high-quality licensed data. Conversely, video game companies can generate additional revenue from the digital assets they have already produced. In between, Origin Lab will transform the video game assets into a format suitable for training data — which might be as straightforward as conducting a rendering run or as intricate as automating hours of gameplay footage.

“It became evident that the video game industry was harboring some extraordinarily valuable data, but there was no genuine method or framework to effectively connect AI labs and the gaming sector,” Rodde explains. “So, we effectively constructed that bridge.”

Labs have long been intrigued by video game footage as a potential data resource, but licensing and quality issues have frequently posed challenges. In December 2024, OpenAI sparked a minor controversy when the initial iteration of its Sora video-generation model appeared to replicate footage of popular games and streamers — likely because it was trained on Twitch streams. Amazon has expressed its intent to leverage Twitch footage for training models.

Origin’s triumph in fundraising reflects a burgeoning market — not only for training data but for startups capable of being vital suppliers to prominent AI labs. Faraz Fatemi, a partner at Lightspeed who led the investment in Origin, asserts that the success of companies like Scale AI has made the opportunity too significant to overlook.

“We’ve witnessed how rapid revenue growth can be for data vendors supporting the leading labs,” Fatemi informed TechCrunch. “These are highly capitalized enterprises, and the data is the bottleneck for all of them.”