Runway, which started out supporting film production, now competes with Google in AI

Runway is an AI video startup founded in New York in 2018 by Chilean and Greek founders who met at NYU Tisch. It grew into video generation models like Gen-4.5 and tools that turn text into editable cinematic content, and it has become deeply embedded in film production workflows and advertising agency work. It signed deals with Lionsgate and AMC Networks, and it was also used in Everything Everywhere All At Once.

The company's valuation has climbed to $5.3 billion, and one of the founders said ARR grew by $40 million in Q2 2026. The company has 155 employees, with offices in New York, London, San Francisco, Seattle, Tel Aviv, and Tokyo. The initial mission was to make everyone a filmmaker, but as video models became more sophisticated, the goal evolved into turning everyone into a better filmmaker.

The next bet is the world model. Runway unveiled its first world model last December and plans to release additional models this year. Whereas language models compress existing human knowledge from the internet, message boards, and textbooks, Runway believes that world models trained directly on observational data offer a better path to learning how reality actually works. The logic is that as sensory data accumulates, the models get closer to becoming a digital twin of reality, compressing the pace of experimentation and accelerating scientific progress. However, the leap from video intelligence to generalized reasoning has not yet been proven.

Applications include the following:

• Interactive entertainment
• Gaming
• Robotics training
• Drug discovery
• Climate modeling

In the long term, the company is also targeting biological world models and aging research. But competition is accelerating too. Luma, World Labs, and Google's Genie are moving in the same direction, and Google's Veo is already going head-to-head with Runway's core video business.

The issue is compute. Runway has deals with CoreWeave and Nvidia, but it hasn't disclosed whether it has secured a dedicated compute cluster. Some point out that with $860 million in cumulative funding, it won't be easy to compete against giants like OpenAI and Alphabet, but the founders believe that the absence of Bay Area-style standardization forced them to be more scrappy, and that pressure helped them generate revenue early on.