Runway Evolves from Filmmaker Tools to Google AI Rival

Runway, an AI video generation company, lacks the usual Silicon Valley background. Its founders, two from Chile and one from Greece, connected at NYU's Tisch School of the Arts. They launched the business in New York.

Some view Runway as a key player in AI today. This stems not from current products but from future ambitions.

For years, AI development centered on language as the core of intelligence. Models such as OpenAI's ChatGPT and Anthropic's Claude follow this approach.

Runway and rivals pursue another path. Founders think future AI will arise from video and world models. These learn real-world dynamics, beyond human descriptions. The idea seems theoretical, but results could prove major.

Co-founder and co-CEO Anastasis Germanidis calls training on direct world observations the next AI step. He claims early leaders here will not come from language experts.

"We're basically bound by our own understanding of reality," Germanidis said at Runway's bright headquarters near Union Square.

"Language models are trained on the entire internet, on message boards and social media, on textbooks, distilling the existing human knowledge," Germanidis added. "But to get beyond that, we need to leverage less biased data."

Started in 2018, Runway gained fame with video models like Gen-4.5. These tools convert text prompts to editable cinematic clips.

Now, Runway aids filmmakers and ad agencies in production. Deals include Lionsgate and AMC Networks. Tools appeared in "Everything Everywhere All At Once."

The firm holds a $5.3 billion valuation. One founder notes $40 million in annual recurring revenue for Q2 2026.

Success in video-to-world models could impact Hollywood and drug discovery. Failure might leave Runway behind rivals with more funds, especially Google.

Move to World Models

In recent six months, Runway advanced its strategy. It moved past video generation. The first world model launched in December. Another follows this year. World models simulate environments to forecast behaviors.

Runway joins others converting video models into world models. Uses span interactive entertainment, gaming, robotics. Luma and World Labs follow suit. Google's Genie model heads there too.

All seek AI for tough human challenges. This differs from Runway's start, driven by tech advances and founders' instincts.

Germanidis views world models as science tools. More training data creates digital universe twins for fast experiments. Science often waits for results. Speeding that accelerates discovery.

"If we can build a better scientist than human scientists, we can accelerate progress in how we understand the universe and how we solve problems," Germanidis said.

Founders and Early Days

Runway merch appeared at its March 2026 AI Summit.

Germanidis discovered programming at age 11 in Athens. At 18, he came to the U.S. for neuroscience and film studies. He switched to computer science, worked at Valley firms, then left the culture.

Co-CEO Cristóbal Valenzuela grew up in Santiago, studied economics, then film and software. Chief Innovation Officer Alejandro Matamala-Ortiz, also from Santiago, handled advertising and design.

They met in 2016 at NYU's ITP program. Valenzuela called it an art school for engineers.

All once dreamed of filmmaking, per Matamala-Ortiz. Runway began with a goal: Use AI to let anyone make films.

First model released February 2023, basic by today's standards. Mission grew: Make anyone a great filmmaker, Matamala-Ortiz said.

Team expanded to 155 across New York, London, San Francisco, Seattle, Tel Aviv, Tokyo. "But throughout this process, we learned that these models can understand how the world works, and if you scale them, they can be useful for many other different things," he said.

Applications include robotics, drug discovery, climate modeling. Last year's robotics unit brought real tests and uses, per Germanidis.

He expects multi-modal training on text, video, voice, sensors. Compounding effects matter.

Long-term, Germanidis eyes biological world models for anti-aging.

Facing Competition

Runway led AI video but world models bring new fights. Google, Yann LeCun, Fei-Fei Li, startups compete.

Kian Katanforoosh of Workera notes no proof yet links video smarts to broad reasoning via world models. Still possible. Runway needs compute resources.

Deals with CoreWeave, Nvidia exist. No word on dedicated clusters for frontier training.

"How are you going to build a foundational model without a cluster?" Katanforoosh asked. "I don't think anybody can do that."

Runway raised $860 million total, $315 million in February from AMD Ventures, Nvidia. Matches Luma AI's $900 million, World Labs' $1.29 billion per PitchBook.

But OpenAI raised around $175 billion per Sam Altman. Alphabet values at $4.86 trillion. Google's Veo rivals video, Genie world models.

OpenAI closed Sora in March, costing $1 million daily in compute against $2.1 million revenue estimates.

Resources do not ensure wins. ElevenLabs beat OpenAI, Google on benchmarks with less.

Runway founders see non-Valley roots as strength. Diversity, scrappiness forced early revenue.

COO Michelle Kwon says no rush for more funds despite compute needs.

"Their background has led them to be early, to be right more often than not, and to build a culture that moves incredibly quickly," investor Michael Dempsey of Compound said.

Valenzuela draws from poet Nicanor Parra, anti-rules. "Rules are just rules they invented," he said. "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."

Related on Neura Market

• AI Video Generation Tools
• World Models Directory
• AI for Media Production