From Google DeepMind's Genie 3 to Nvidia's Cosmos, from Fei-Fei Li's World Labs to Meta's evolving Codec Avatars, nearly every major AI player suddenly pivoted toward the same goal: building AI systems that understand and generate three-dimensional, physics-aware worlds.
This wasn't a coincidence, it was recognition of a fundamental truth.
The next trillion-dollar infrastructure layer won't be cloud computing or mobile operating systems. It will be world models: AI systems that can simulate, predict, and generate reality itself.
But only a handful are positioned to become the foundational infrastructure that powers everything from augmented reality glasses to autonomous vehicles, from robotic factories to virtual worlds.
The Competitors: Mapping the World Model Landscape
