Runway, an artificial intelligence startup gaining significant traction, stands apart from the typical Silicon Valley narrative. Its three co-founders, hailing from Chile and Greece, converged at NYU’s Tisch School of the Arts and established their company in New York, eschewing the common path of Stanford alumni or ex-Google engineers with massive seed funding. Depending on who you ask, Runway could be one of the most consequential AI companies today, not just for what it has built, but for its ambitious next-generation goals.

While much of the AI industry has centered on large language models (LLMs) like OpenAI’s ChatGPT and Anthropic’s Claude, built on the premise that intelligence resides in language, Runway is charting a different course. The company’s leadership believes the next evolution of AI will not stem from text but from video and sophisticated ‘world models’ designed to comprehend how the physical world operates, moving beyond mere human descriptions. Co-founder and co-CEO Anastasis Germanidis emphasizes that training models directly on observational data represents AI’s new frontier, suggesting that those who excel here will ultimately surpass those who perfected language-based systems.

Established in 2018, Runway built its initial reputation on pioneering AI video-generation models, including its advanced Gen-4.5, and a suite of tools that transform text prompts into editable, cinematic content. Its technology is now integral to production workflows for filmmakers and advertising agencies, securing partnerships with major media entities such as Lionsgate and AMC Networks, with its tools even contributing to acclaimed films like ‘Everything Everywhere All At Once.’ Currently valued at $5.3 billion, Runway is not resting on its laurels. The company recently expanded beyond video generation, launching its first world model in December, with another planned for this year. These world models are AI systems capable of simulating environments accurately enough to predict their behavior, opening doors for applications in interactive entertainment, gaming, and robotics training.

Germanidis views world models as critical scientific infrastructure, envisioning a future where a single model, trained on vast sensory data, could become a digital twin of the universe. This would allow for accelerated experimentation, potentially compressing scientific progress itself by reducing waiting times for results. However, this ambitious pursuit places Runway in direct competition with formidable players. While other startups like Luma and World Labs are on a similar trajectory, tech giants like Google, with its Veo video model and Genie world model, pose a significant challenge. OpenAI, despite some setbacks like its Sora video platform, and researchers like former Meta chief scientist Yann LeCun and AI ‘godmother’ Fei-Fei Li, are also heavily invested in this domain. Industry experts, such as Kian Katanforoosh, CEO of Workera and a Stanford lecturer, acknowledge that the leap from video intelligence to generalized reasoning via world models remains unproven, underscoring the immense computational resources required for such foundational models.

Despite the intense competition and the need for substantial computing power—Runway has secured deals with CoreWeave and Nvidia and raised $860 million to date, including a $315 million round from AMD Ventures and Nvidia—the company’s founders believe their non-traditional background offers an advantage. Co-CEO Cristóbal Valenzuela suggests that operating outside the ‘standardization’ of Silicon Valley fostered a scrappier culture and diversity of thought, forcing them to generate revenue earlier than many peers. Chief Operating Officer Michelle Kwon confirms the company is not currently rushing to raise additional funds. Early investor Michael Dempsey of Compound highlights Runway’s track record of being ‘early’ and ‘right more often than not,’ attributing it to their agile culture. Ultimately, whether Runway can translate its success in video generation into dominance in world models, and unlock solutions for complex problems ranging from robotics to anti-aging research, remains the high-stakes question defining its future.