Odyssey raises $310M at a $1.45B valuation, with Amazon among its investors

Odyssey, a startup building AI-based world models, has raised $310M in a Series B round at a $1.45B valuation. Amazon, AMD Ventures, GV and other investors are taking part, and the company is quickly becoming one of the most visible players in AI beyond conventional LLMs.

The round and the valuation

On June 17, 2026, the company announced a new round led by Natural Capital. For Odyssey this is more than a large cash infusion: it marks entry into the unicorn club, with the startup's valuation reaching $1.45B. Against a backdrop of cooling in parts of the AI market, the round is a strong signal that investors are willing to bet not only on chatbots and assistants, but also on more complex models that try to describe the physical world.

Alongside the venture funds, a notable group of angel investors joined the company. They include Jeff Dean, Elad Gil, Garry Tan, Guillermo Rauch and Cruise founder Kyle Vogt. After this round, the total capital raised by Odyssey has reached $337M. For a startup founded only in 2023, that is a very fast capital build-up, especially given that this is an expensive field that demands large amounts of compute and proprietary data.

• $310M: size of the new Series B round
• $1.45B: Odyssey's current valuation
• $337M: total funds raised
• Natural Capital: round leader
• Amazon, AMD Ventures and GV: among the participants

Why world models

Odyssey works in the category of world models. Unlike large language models, which rely mainly on text and dialogue, these systems learn from physical-world data and try to reproduce environments in line with the laws of physics. That is why the field draws so much interest: if a model genuinely understands space, motion, objects and cause and effect, it opens up applications that ordinary LLMs cover poorly.

According to the company, its models can be used in a range of scenarios, from building video games to robotics. But the wider public probably knows Odyssey best as the developer of technology that generates rich interactive video from text prompts. This is no longer just a clip playing along a prepared script, but an attempt to assemble an environment you can interact with more fluidly and flexibly.

The startup's approach to data collection is telling, too. Odyssey essentially repeats the logic Google Earth once followed: it sends people with backpack-mounted cameras to capture material from the real world. The method fits the founders' backgrounds. CEO Oliver Cameron was previously co-founder and CEO of the self-driving startup Voyage, which Cruise later acquired, where he went on to become vice president of product. CTO Jeff Hawke worked as an engineer at the British startup Wayve, which also works on self-driving transport.

A bet on AWS

Amazon's participation matters for more than its name on the investor list. Odyssey said AWS will become its preferred cloud provider, and its models will be optimized for Trainium chips. For Amazon, it is a chance to show that its AI infrastructure can appeal not only to enterprise customers but also to ambitious labs building new types of models and looking for an alternative to the Nvidia ecosystem.

For Odyssey itself, the pairing looks strategic as well. A company working with physical-world models needs expensive compute, a stable cloud stack and a clear partner for scaling. If the Trainium optimization goes well, the startup will gain not just funding but also tighter infrastructure support. In today's market that is almost as important as the size of the round: AI companies win not only with their models, but with what they train and run them on.

What it means

The large rounds around Odyssey show the market beginning to see world models as a separate, promising branch of AI rather than a side experiment next to LLMs. If investors' bet pays off, the next wave of AI competition will be fought not only over the best text interface, but over systems that can plausibly model space, motion and the behavior of objects in the real world.