AI models that interact with the physical world need data on how things move and exist in space, but unlike language models, there's no convenient pile of internet text to scrape. Enter Origin Lab, which has raised $8 million in seed funding led by Lightspeed Ventures, with participation from SV Angel, Eniac, Seven Stars, and FPV, plus angel checks from Twitch co-founder Kevin Lin and Cruise founder Kyle Vogt.
The startup's pitch is straightforward: video game companies have tons of digital assets already optimized for physics and movement, and AI labs like Yann LeCun's AMI Labs or Fei-Fei Li's World Labs would love to buy that data. Origin Lab acts as the middleman, converting game assets into usable training data - whether that's a simple rendering run or automating hours of walkthrough footage.
"The AI systems being built now need to understand how the physical world works and how things move," co-CEO Anne-Margot Rodde told TechCrunch. "That data essentially lives in video games." The company's other co-founders are Antoine Gargot and Colin Carrier.
The idea isn't new - labs have long been eyeing video game footage, but licensing and data-quality issues have been obstacles. In December 2024, OpenAI's Sora video model caused a minor scandal by apparently regurgitating footage from popular games and Twitch streams, presumably because it had been trained on them. Amazon has also been open about wanting to use Twitch footage for model training.
Origin's $8 million raise signals a growing market for data suppliers to major AI labs. Lightspeed partner Faraz Fatemi, who led the investment, noted that companies like Scale AI have shown how sharply revenue can scale for data vendors serving well-capitalized labs. "The bottleneck for all of them is data," he said.