World-Model AI Startup Odyssey Closes $310M Series B at $1.45B Valuation
Odyssey, founded by self-driving car veterans Oliver Cameron and Jeff Hawke, raised $310 million at a $1.45 billion valuation with Amazon, AMD, and Nvidia all in the round — while committing its AI world-model stack to AWS's Trainium chips.
A $310 million bet on AI that simulates the world
Odyssey, an AI lab founded in late 2023 by autonomous-vehicle veterans CEO Oliver Cameron and CTO Jeff Hawke, closed a $310 million Series B on June 17 at a $1.45 billion valuation — the largest venture deal of the week in AItechcrunch +1. Natural Capital led the round, with Amazon, AMD Ventures, GV, EQT, and CIA-affiliated fund IQT all joining alongside a roster of angel investors that includes Google chief scientist Jeff Dean, Y Combinator president Garry Tan, and Cruise founder Kyle Vogttechcrunch +1. The company has now raised $337 million in totaltechcrunch.
Beyond LLMs: what world models actually do
Where large language models predict the next word in a sequence, world models predict the next state of an environment — complete with physics, causality, and the behavior of objects and agents over timesiliconangle. Odyssey has already shipped three systems: Odyssey-2 Max, which advanced physics accuracy in general-world simulation; Starchild-1, the first real-time multimodal world model; and Agora-1, which enables multi-agent interactivity within a shared simulated spacesiliconangle. Cameron demonstrated one use case in which Odyssey's model recreated a multiplayer version of GoldenEye from pixels, actions, and sounds alone — no access to the game's underlying physics enginestartupfortune. The broader target market spans robotics training, autonomous-vehicle testing, scientific simulation, and interactive mediasiliconangle +1.
Amazon, AMD — and still Nvidia
The compute story is more nuanced than a clean break from Nvidia. Alongside the new funding, Odyssey struck a deal making AWS its preferred cloud provider and committed to optimizing its models for Amazon's Trainium custom chipstechcrunch +1. Amazon VP Ron Diamant described world-model training as "one of the most demanding workloads in AI," arguing Trainium was "purpose-built for exactly this kind of scale"siliconangle. Yet Nvidia's venture arm also participated in the round, and Odyssey's tools currently run on Nvidia H200 and B200 GPUsstartupfortune. Cameron has said serving costs on those chips run between $2 and $4 per person per hour, not counting training — a cost burden that makes cheaper Trainium access strategically valuablestartupfortune.
A crowded race with deep-pocketed rivals
Odyssey is not building world models in isolation. Rival startup Decart.ai raised $300 million at nearly a $4 billion valuation for its Lucy and Oasis models, Yann LeCun's AMI Labs closed a $1.03 billion round in March, and Runway AI raised $315 million in Februarysiliconangle. Google has entered the space directly with Project Genie, which combines the company's Genie 3 world model with its image-generation toolssiliconangle. Natural Capital partner Jay Zaveri has called world models part of a "second wave of AI" — a framing that explains the urgency among chip makers, cloud providers, and investors to plant flags early in a category that is still establishing its first killer applicationsstartupfortune.