What happens when a company raises $860 million to build AI video tools and then tells you AI video isn't really the point?
That's effectively the position Runway CEO Cristobal Valenzuela is staking out. In a recent interview, Valenzuela argued that AI-generated video is a precursor to something more fundamental: world models. A world model is an AI system that doesn't just produce realistic-looking footage — it builds an internal understanding of how physical reality actually works. The difference matters. A video generator can make a ball appear to roll down a hill. A world model understands gravity, mass, and friction well enough to predict where the ball goes before it renders anything.
The distinction has real consequences for what these systems can eventually do. World models are the underlying technology that companies like Google DeepMind are chasing for robotics and autonomous vehicles — applications where the AI needs to reason about physical consequences, not just generate convincing images.
$5.3 Billion on a Bigger Bet
Runway sits at a $5.3 billion valuation, competing directly with Google and OpenAI on AI video generation. Its models have kept pace with the most well-funded labs in the world, which is a credible achievement for a New York company that started as a creative tool for filmmakers and video editors.
But Valenzuela's framing suggests Runway sees its video tools as training ground — both technically and commercially. Generating video requires the model to learn something about how objects move, how light behaves, how scenes change over time. That accumulated understanding, the argument goes, is what eventually becomes a world model.
For the working video editors and content creators using tools like D-ID or Runway today, this probably feels abstract. The near-term question is whether the current generation of AI video tools is good enough, fast enough, and affordable enough to fit into a real production workflow. The world model thesis is more relevant to robotics labs and autonomous system developers than to someone making marketing content.
Still, Valenzuela's framing is a useful signal about where serious AI video investment is actually headed — and why the big labs are spending billions on video that goes far beyond pretty clips.