- Odyssey, an AI lab building general-purpose world models, raised $310M at a $1.45B valuation in a round led by Natural Capital, with Amazon, AMD Ventures, GV, EQT, and CIA-affiliated fund In-Q-Tel participating.
- The deal names AWS as Odyssey’s preferred cloud provider and gives it access to Amazon’s Trainium chips, a pointed move four months after Nvidia’s venture arm backed Odyssey’s Series A.
- With 55 employees and $27M raised before this round, Odyssey is now one of the most capital-intensive AI bets per head in the market.
Four months ago, Nvidia’s venture arm invested in Odyssey. Today, Odyssey announced it will use Amazon’s chips for its future work.
Odyssey, an AI lab based in Palo Alto that builds world models, has closed a $310 million Series B round at a $1.45 billion valuation. Natural Capital led the round, with Amazon, AMD Ventures, GV, EQT, and the CIA-affiliated fund In-Q-Tel also joining.
As part of the deal, AWS will be Odyssey’s preferred cloud provider and will supply Trainium chips. These chips are Amazon’s answer to Nvidia’s dominance in AI computing and are designed for the fast, high-volume workloads needed for real-time world simulation.
“This round provides the compute, infrastructure, and partners to push the frontier of general world models, and to achieve a GPT-3 moment for the field,” says Oliver Cameron, co-founder and CEO of Odyssey.
The wider AI market is expected to be worth about $602 billion in 2026 and could grow to $3.6 trillion by 2033. Physical AI, which includes robotics, simulation, and autonomous systems, is one of the fastest-growing areas.
From self-driving to simulating everything
Cameron and Jeff Hawke started Odyssey in late 2023, bringing experience from the challenging field of self-driving cars. Cameron was VP of Product at Cruise and CEO of Voyage, which Cruise later acquired. Hawke was the founding engineer at Wayve, where he helped develop the GAIA model series for autonomous driving.
Together, they took the problem of predicting what happens next in the world so you can act on it, and wondered what this would look like on a larger scale, beyond just cars.
Their solution is a general-purpose world model: a causal, multimodal AI system that can simulate persistent, shared environments where many people and AI agents can act simultaneously, and every change affects everyone in the world.
It trains its models on physics, cause-and-effect reasoning, and long-term prediction rather than text. For example, a robotics company could run thousands of training scenarios in Odyssey’s simulated environment instead of testing on a real factory floor.
Odyssey has 55 employees working in Palo Alto, London, and Zurich. The team includes people who previously worked at DeepMind, Tesla, Waymo, Meta, Apple, and Wayve.
A crowded race with deep-pocketed rivals
The world model sector has attracted major investment in the past 18 months.
Runway, valued at $5.3 billion after a $315 million Series E in February 2026, launched its first world model product in December 2025 and gained $40 million in annual recurring revenue in the second quarter of 2026.
World Labs, Fei-Fei Li’s spatial AI startup, launched its Marble product in November 2025 after raising $230 million. Google DeepMind’s Genie model is already being used at Waymo. Yann LeCun’s AMI Labs began funding talks at a €3 billion valuation before releasing any products.
While most competitors focus on video generation or gaming, Odyssey sees itself as a pure research lab, similar to what OpenAI did with language models. Its recent projects include Odyssey-2 Max for accurate physics simulation, Starchild-1 as the first real-time multimodal world model, and Agora-1, which allows multiple agents to interact in a shared simulation.
“We developed deep conviction in Odyssey’s research direction, technical leadership, and execution, which made this our largest investment to date. We believe they have the potential to help define AI beyond language models,” says Jay Zaveri, general partner at Natural Capital.
The Nvidia question
The story of which company provides Odyssey’s computing power is a key subplot. NVentures, Nvidia’s venture capital arm, backed Odyssey’s Series A in February 2026 as part of Nvidia’s broader strategy to fund world-model startups such as Wayve and Runway. NVIDIA is not part of the Series B group.
Instead, AMD Ventures is a new shareholder, and AWS Trainium is now the chip of choice. It is unclear whether this change is due to a real belief in Amazon’s technology or simply better deal terms in a highly competitive market.
What is clear is that the competition for world model infrastructure is now about more than just the models themselves.
IQT’s involvement adds another dimension. The CIA-affiliated fund has supported more AI companies in recent years, and Odyssey’s mission statement specifically mentions defence as a target use.
As world models become better at simulating complex physical and social environments at scale, questions about how and by whom they are used will soon extend beyond purely technical issues.
