Odyssey banked Nvidia's money in February. Its $310m Series B locked the chipmaker out
Odyssey, the world models lab founded by self-driving veterans Oliver Cameron and Jeff Hawke, has raised $310m in a Series B at a $1.45bn valuation, and the most revealing detail is not the size of the cheque but the names attached to it. Natural Capital led the round, with Amazon, AMD Ventures, Google's GV, EQT, and the intelligence-linked fund In-Q-Tel taking part. As part of the deal, AWS becomes Odyssey's preferred cloud provider and supplies the company with its Trainium chips. Nvidia, whose venture arm NVentures had backed Odyssey's Series A only four months earlier, does not appear anywhere in the financing.
That omission carries more weight than a single startup's cap table would normally bear. Trainium is Amazon's in-house answer to Nvidia's dominance over AI computing, engineered for the high-volume, low-latency workloads that real-time world simulation demands. Pairing Amazon's silicon with capital from AMD Ventures, Nvidia's principal chip rival, turns the round into something close to a market verdict. Cameron said the financing provides the compute, infrastructure, and partners needed to push the frontier of general world models and to reach what he called a GPT-3 moment for the field. The honest qualification is that diversification is not defection. Tech Funding News noted that it remains unclear whether the shift reflects genuine conviction in Amazon's chips or simply better commercial terms in a fierce market, and a preferred cloud arrangement is not an exclusive one.
The wider significance sits in the timing. Buyers across the industry are actively searching for ways around Nvidia, from Chinese carmakers designing their own driving chips to infrastructure startups treating Nvidia GPUs as a commodity input. Odyssey is a small data point in that shift, but a telling one, because the same investor base that helped build Nvidia's commercial empire is now writing cheques that route around it. When a company can take NVentures money in February and assemble an anti-Nvidia coalition by June without apparent friction, the supplier's grip looks less absolute than its market capitalisation implies.
Odyssey itself remains an early and unproven bet. The company, started in late 2023, employs roughly 55 people across Palo Alto, London, and Zurich, drawing on alumni of DeepMind, Tesla, Waymo, and Apple, and had raised only about $27m before this round. Its recent work includes Odyssey-2 Max for physics simulation and Agora-1, which allows multiple agents to act within a single shared environment. The underlying premise grew out of the central problem of autonomous driving, predicting what happens next so a machine can act on it, extended to a general model of physics and cause and effect rather than text. A robotics developer could run thousands of training scenarios inside such a simulation rather than on a factory floor.
The capital reflects how crowded and richly funded the category has become. Runway reached a $5.3bn valuation after a $315m round in February, Fei-Fei Li's World Labs raised $230m, Google DeepMind's Genie model is already in use at Waymo, and Yann LeCun's new venture is reportedly raising at a €3bn valuation before shipping a product. Against that field, Odyssey's wager is steeper than most, a 55-person team betting that a technology yet to prove its commercial value will define a new foundation model class. The presence of In-Q-Tel, alongside Odyssey's own identification of defence as a target market, indicates that the addressable customer base for convincing world simulation extends well past gaming and film. Whoever solves it credibly will find demand in robotics, autonomy, and national security alike, which is precisely why the question of whose chips power that work has become a strategic one rather than a procurement footnote.
