Runway bets on world models to challenge Google in AI

Runway, the startup that gained prominence by equipping filmmakers with advanced video-generation tools, is pivoting toward a more ambitious objective: building “world models.” While the AI industry has largely prioritized large language models like ChatGPT, the New York-based company is betting that the future of artificial intelligence lies in systems that understand the physics and mechanics of reality rather than just processing human text.

Co-founder and co-CEO Anastasis Germanidis argues that training models on observational data is the next frontier. “Language models are trained on the entire internet, on message boards and social media, on textbooks — distilling the existing human knowledge,” Germanidis said. “But to get beyond that, we need to leverage less biased data.”

The Path to World Models

Runway has spent the last several years refining its video-generation models, including the recent Gen-4.5. While these tools have become staples for production workflows at companies like Lionsgate and AMC Networks, the startup is now expanding its scope. In December, it launched its first world model, with plans for a successor later this year.

The goal is to simulate environments that can predict behaviors in real time, which has massive potential for:

• Robotics training: Enabling machines to learn in simulated physics environments.
• Scientific research: Creating digital twins of the universe to accelerate experiments.
• Interactive entertainment: Developing more immersive gaming and media experiences.

A Different Kind of Startup

Unlike many Silicon Valley giants, Runway’s origins are rooted in the arts. Founders Germanidis, Cristóbal Valenzuela, and Alejandro Matamala Ortiz met at NYU’s Tisch School of the Arts, viewing their company through the lens of creative expression rather than standard tech industry conventions. This culture of independence has helped them maintain a distinct trajectory.

However, the transition to world models places them in direct competition with deep-pocketed incumbents like Google, which is developing its own Genie model. Kian Katanforoosh, CEO of Workera, notes that the jump from video intelligence to generalized reasoning remains unproven. He emphasizes that success will depend heavily on access to massive compute clusters.

Despite these hurdles, Runway has secured $860 million in funding to date and maintains partnerships with CoreWeave and Nvidia. With a valuation of $5.3 billion and $40 million in annual recurring revenue added in Q2 2026, the company is positioning itself to compete not just in media, but in the fundamental infrastructure of future scientific discovery.