Odyssey has raised $310 million in a Series B round that values the AI startup at $1.45 billion, giving the company fresh capital to push ahead on systems it says can simulate the physical world and human behavior.
The financing was led by Natural Capital, with participation from Amazon, GV, AMD Ventures, EQT, IQT and other investors. Odyssey said the round also included support from existing backers and researchers connected to OpenAI, DeepMind, MSL, Recursive and Thinking Machines.
The company said the new funding will speed up its work on what it calls world models, AI systems designed to understand and predict how the world works. Odyssey argues that these models could become a foundational technology with applications in robotics, science, healthcare, education, gaming and defense.
Odyssey was founded in 2023 by Oliver Cameron and Jeff, after the team spent years working on autonomous driving systems. In its announcement, the company said that earlier work on self-driving technology helped it develop a model of the world that captured physics, causality, motion and human behavior, even if only from the limited perspective of the road.
The company pointed to several recent research milestones as evidence of progress. Odyssey-2 Max is described as its most advanced general-purpose world model and one that improved physical accuracy. Starchild-1 was introduced as a real-time multimodal world model. Agora-1 added multi-agent interaction inside a shared simulation, while PROWL used reinforcement learning to improve world-model performance through exploration.
Odyssey said these projects show that world models are moving beyond a narrow research area tied to autonomous vehicles. The company believes the field may be approaching a turning point similar to the one that helped large language models gain wider adoption after GPT-3.
The startup also highlighted a deeper partnership with Amazon Web Services. AWS will become Odyssey’s preferred cloud provider, and the company said it is working with Amazon’s Annapurna Labs to optimize its models on Trainium chips. Odyssey said the collaboration will cover both research and go-to-market efforts.
Amazon executive Ron Diamant said world models are a demanding AI workload because they require high compute throughput and low latency. He said Trainium is designed for that kind of scale and that the partnership could help accelerate applications in robotics, gaming, science and other areas.
Natural Capital partner Jay Zaveri said the firm made Odyssey its largest investment to date because of its conviction in the company’s technical direction, leadership and execution. He said the startup has the potential to help define AI beyond language models.
GV partner Luna Schmid also said Odyssey has been early to a market that is now becoming a large category. She pointed to the company’s long-running focus on using AI to simulate the world as a key reason for continuing to back the business.
Odyssey said the new round will help it scale work at the frontier of world modeling, where speed, quality and infrastructure are central to progress. The company framed the raise as a step toward building AI that can do more than generate text, with systems that can reason about and simulate real-world environments.