Runway pivots to world models to challenge AI incumbents

Runway is shifting its core strategy from video generation to developing world models—which are AI systems that simulate environments to predict behavior. The company’s founders believe the next form of AI intelligence will be built from video and world models rather than text. Runway launched its first world model in December. Currently valued at $5.3 billion, the startup added $40 million in annual recurring revenue (ARR)—defined as annual recurring revenue—in the second quarter of 2026, according to one of its founders. Founded in 2018, Runway has raised $860 million to date, including a $315 million funding round in February.

The pivot places Runway in direct competition with Google and other startups. While Runway seeks to scale its models, the costs of AI development remain high. For instance, according to some estimates, OpenAI shuttered its video platform Sora in March after burning roughly $1 million per day in compute costs with barely $2.1 million in revenue.

The funding landscape highlights the scale of Runway’s competition:

• Runway: $860 million raised to date, with a valuation of $5.3 billion.
• Luma AI: $900 million raised.
• World Labs: $1.29 billion raised.
• OpenAI: $175 billion raised.
• Alphabet (Google’s parent company): $4.86 trillion market value.

Runway’s founders—who met at NYU in 2016 and include co-CEOs Anastasis Germanidis and Cristóbal Valenzuela, alongside chief innovation officer Alejandro Matamala Ortiz—argue that operating outside the Silicon Valley bubble gives them an advantage. Originally from Greece and Chile, the founders built the company in New York rather than Silicon Valley, though they have since expanded to 155 employees across offices in New York, London, San Francisco, Seattle, Tel Aviv, and Tokyo.

Valenzuela believes this outsider status allows the company to bypass established rules. “Rules are just rules they invented. That’s a driving force of how we do things at Runway. They say Silicon Valley is here and that’s where the startups are. Why? Those are just made-up rules. Scrub them all and start again,” Valenzuela, co-founder and co-CEO of Runway, said. Germanidis noted that language models are trained on existing human knowledge from the internet, but moving beyond that requires leveraging less biased data.

Why it matters

Runway’s transition from video generation to world models is a high-stakes bet to unlock capabilities in robotics and science, essential for competing against deep-pocketed incumbents like Google.