Amazon, Nvidia, and AMD Bet $310M on Odyssey ML’s 3D World Models

The race for Artificial General Intelligence (AGI) is shifting from text prediction to physical simulation. In a massive show of confidence, industry giants Amazon, Nvidia, and AMD have joined a $310 million funding round for Odyssey ML, a startup dedicated to building advanced 3D world models.

Beyond Language: The Rise of Physical Intelligence

While Large Language Models (LLMs) have mastered the nuances of human dialogue, they remain fundamentally disconnected from the physical realities of our universe. Odyssey ML aims to bridge this gap by developing world models that simulate the physical world in three dimensions. Unlike text-based models that predict the next token in a sequence, Odyssey’s architecture is designed to understand physics, body language, spatial relationships, and complex dynamics.

The technical leap here is significant. By teaching AI to comprehend how objects move, interact, and react to forces, Odyssey is moving toward "embodied AI"—the intelligence required for robots and autonomous systems to navigate the real world safely and effectively.

A High-Stakes Ecosystem of Investors

The $310 million investment brings Odyssey ML’s valuation to a staggering $1.45 billion. This massive capital injection comes from an elite group of backers, signaling that the industry views 3D world modeling as a critical frontier. Alongside the heavyweights Amazon, Nvidia, and AMD, the round includes GV (formerly Google Ventures), investor Elad Gil, and Google chief scientist Jeff Dean. Interestingly, the funding also features IQT, a fund with ties to the CIA, highlighting the strategic importance of this technology.

To power these massive simulations, Odyssey ML has optimized its infrastructure by using AWS as its preferred cloud provider, specifically running workloads on Amazon’s custom Trainium chips.

Moving Toward General Intelligence

La visione dei fondatori di Odyssey ML, Oliver Cameron e Jeff Hawke — entrambi veterani del settore dei veicoli autonomi — si allinea con le teorie più rilevanti della ricerca moderna sull'IA. Il capo di Meta AI, Yann LeCun, ha sostenuto spesso che i modelli linguistici da soli manchino del "modello del mondo" necessario per raggiungere un'intelligenza di livello umano. Allo stesso modo, il CEO di Google Deepmind Demis Hassabis e la pioniera dell'IA Fei-Fei Li (attraverso la sua startup World Labs) considerano la simulazione spaziale e fisica come il passo successivo essenziale verso l'AGI.

Con un team di 55 persone distribuito tra Londra, Zurigo e Palo Alto, Odyssey ML si posiziona all'intersezione tra computer vision e ragionamento fisico, cercando di costruire il "motore" fondamentale che permetterà infine all'IA di abitare e agire all'interno della nostra realtà fisica.

Punti chiave