Odyssey raises $310m at $1.45b valuation to build AI world models

Odyssey, an AI lab founded by self-driving car veterans, raised $310 million in Series B funding at a $1.45 billion valuation, the company said Wednesday, in one of the largest investments yet in a technology that aims to move artificial intelligence beyond language and into physical world simulation.

Natural Capital led the round, with participation from Amazon, AMD Ventures, GV, EQT and In-Q-Tel. Existing investors including Jeff Dean, Google's chief scientist, and Kyle Vogt, founder of Cruise, also joined. The company simultaneously announced a strategic partnership with Amazon Web Services, which will become Odyssey's preferred cloud provider and supply Trainium chips for its compute-intensive workloads.

"World models represent a new class of foundation model — AI that can understand and simulate the world itself," said Oliver Cameron, co-founder and chief executive officer of Odyssey. "The last few years have seen major breakthroughs in scaling, interactivity, multimodality and physics accuracy, and the field is now advancing extremely quickly."

Odyssey has released four major research systems over the past three years. Odyssey-2 Max advanced physics accuracy for general world simulation. Starchild-1 introduced the first real-time multimodal world model, combining visual and audio understanding. Agora-1 enabled multiple humans and AI agents to interact within a shared simulated environment. PROWL demonstrated how world models can improve through active exploration — learning from their own experiences rather than static training data.

World models differ fundamentally from large language models by attempting to predict how environments evolve over time and how objects interact, rather than simply predicting the next word. This makes them computationally far more demanding. AWS Trainium chips are purpose-built for exactly this kind of workload, said Ron Diamant, vice president and distinguished engineer at Amazon, describing world models as "one of the most demanding workloads in AI" with requirements for "massive compute throughput with tight latency constraints."

The funding reflects a broader shift in AI investment toward physical world understanding. While LLMs have dominated the past two years of venture capital deployment, researchers increasingly view world models as a necessary next step for applications in robotics, autonomous systems, gaming and scientific simulation. Odyssey's team, drawn from DeepMind, Tesla, Waymo, Meta, Apple and Wayve, includes contributors to Google's Gemini language model, DeepMind's Veo video model and Tesla's Full Self-Driving system.

"At Natural Capital, we invest behind ambitious technical teams building what comes next," said Jay Zaveri, general partner at Natural Capital. "We developed deep conviction in Odyssey's research direction, technical leadership and execution, which made this our largest investment to date."

Cameron said the field is approaching a "GPT-3 moment" — a reference to the 2020 release that demonstrated the power of large-scale language models and triggered a wave of investment. The $310 million round provides Odyssey with the compute infrastructure and research capacity to push toward that milestone. The company plans to deploy the technology across robotics, healthcare, education, gaming and defense. For investors, the participation of Amazon and AMD in the round shows major tech companies are betting on world models as the next frontier beyond LLMs, a market that could reshape the $200 billion-plus AI infrastructure sector.

This article is for informational purposes only and does not constitute investment advice.