Runway began by serving to filmmakers — now it desires to beat Google at AI

AI video-generation startup Runway doesn’t have the standard Silicon Valley pedigree. No Stanford founders, no ex-Google founders, no nine-figure seed spherical that purchased them time to disregard income. Its three founders — two from Chile, one from Greece —  met at NYU’s Tisch Faculty of the Arts and constructed the corporate in New York.

Runway additionally could possibly be, relying on who you ask, some of the consequential AI firms right now. Not due to what it has constructed, however due to what it’s attempting to construct subsequent. 

For the previous a number of years, the AI business has largely operated on the premise that intelligence lives in language. Giant language fashions like OpenAI’s ChatGPT and Anthropic’s Claude mirror that guess. 

Runway, alongside different rivals, is making a unique one. Its founders imagine the following type of AI intelligence gained’t be constructed from textual content, however from video and world fashions that learn the way the world works, not simply how people describe it. That distinction sounds educational. Its implications should not. 

Runway co-founder and co-CEO Anastasis Germanidis stated coaching fashions immediately on observational information from the world is the following frontier of AI. The businesses that get there first, he argues, gained’t be those that perfected language.

“We’re principally certain by our personal understanding of actuality,” Germanidis instructed TechCrunch from Runway’s homey sunlight-filled headquarters close to Union Sq..  

“Language fashions are educated on your entire web, on message boards and social media, on textbooks — distilling the prevailing human information,” Germanidis continued. “However to get past that, we have to leverage much less biased information.”

Based in 2018, Runway constructed its repute on video-generation fashions — together with its newest Gen-4.5 — and AI instruments that allow individuals flip textual content prompts into editable, cinematic content material.

Right now, Runway’s know-how powers manufacturing workflows for filmmakers and advert businesses, and the corporate has signed offers with main media gamers like Lionsgate and AMC Networks. Its instruments have even been utilized in movies comparable to “All the things All over the place All At As soon as.” 

Runway is now valued at $5.3 billion and, in response to certainly one of its founders, added $40 million in annual recurring income within the second quarter of 2026.

If Runway’s guess that video technology is the trail to world fashions pays off, the outcome will likely be felt from Hollywood to drug discovery. If it doesn’t, Runway dangers being outpaced by rivals with far deeper pockets — Google chief amongst them.

Taking the leap

Throughout the final six months, the startup has put its plan into motion and expanded past video technology, launching its first world mannequin in December, with plans to launch one other this yr. (World fashions are AI methods that simulate environments properly sufficient to foretell how they’ll behave.) 

Runway isn’t alone in its pursuit of turning physics-aware video fashions to world fashions, with near-term use circumstances in interactive leisure, gaming, and robotics coaching. Startups Luma and World Labs are on a comparable trajectory, and Google has pointed its Genie world mannequin in the identical course.

Everyone seems to be after some model of the identical factor: AI that solves humanity’s hardest issues. That’s removed from Runway’s unique product, but it surely’s the results of each emergent capabilities within the know-how and founders who have been predisposed to observe the place it led.

For his half, Germanidis sees world fashions as scientific infrastructure. The extra sensory information and observations you practice a single mannequin on, the nearer you get to a working digital twin of the universe — one you’ll be able to run experiments on sooner than any lab may. A lot of the scientific course of is simply ready on outcomes, he factors out. When you may compress that ready, you might compress progress itself. 

“If we will construct a greater scientist than human scientists, we will speed up progress in how we perceive the universe and the way we clear up issues,” Germanidis stated.

The moonshot

Germanidis fell in love with programming as an 11-year-old in Athens and got here to the U.S. at 18 to check neuroscience and movie. He turned again to pc science, working at a number of Silicon Valley tech corporations earlier than deciding he’d had sufficient of the tradition. Co-CEO Cristóbal Valenzuela, born and raised in Santiago, studied economics as an undergraduate earlier than working in movie after which software program. One other Santiago native, chief innovation officer Alejandro Matamala Ortiz studied promoting and ran a design agency. 

The three met in 2016 whereas attending NYU’s ITP (Interactive Communications Program), a graduate program that Valenzuela described as an “artwork college for engineers.” 

The co-founders had all aspired to be filmmakers at sure factors of their lives, in response to Matamala Ortiz. So Runway began with a easy mission: Can we use AI to make everybody a filmmaker? 

In accordance with Matamala Ortiz, after releasing their first video-generation mannequin in February 2023 — which is staggeringly unimpressive in comparison with what Runway is placing out right now — that mission advanced into: Can we make everybody a nice filmmaker?  

It required rising the staff to what it’s right now. The corporate has 155 employees unfold throughout places of work in New York, London, San Francisco, Seattle, Tel Aviv, and most not too long ago, Tokyo. “However all through this course of, we realized that these fashions can perceive how the world works, and for those who scale them, they are often helpful for a lot of different various things,” he added.

Issues like robotics, drug discovery, and local weather modeling — the sorts of issues which have stumped researchers for many years. Final yr, Runway launched a robotics unit that Germanidis says has already resulted in real-world testing and deployments.

Germanidis, like others, sees the sphere heading towards coaching a single mannequin on many various modalities — textual content, video, voice, and different sensors — and thinks the compounding impact is the purpose. 

His personal moonshot purpose for Runway’s know-how, given sufficient time and assets, is organic world fashions and anti-aging analysis.

Whether or not Runway can carry its video dominance into world fashions is way from settled, and the competitors isn’t ready round. Runway was among the many first to develop AI video technology, however world fashions are a unique race with deep-pocketed and well-respected rivals. Google, former Meta chief scientist Yann LeCun, AI’s “godmother” Fei-Fei Li, and a rising subject of startups are all chasing the identical purpose. 

Kian Katanforoosh, CEO of AI abilities benchmarking firm Workera and a lecturer at Stanford, identified that nobody has but confirmed the bounce between video intelligence and generalized reasoning through world fashions, however that doesn’t imply it’s inconceivable. He stated that if Runway desires to show its world mannequin guess into actuality, it might want to proceed gathering assets — compute chief amongst them.

Runway has offers with CoreWeave and Nvidia however wouldn’t affirm whether or not it has devoted cluster entry — the form of assured, large-scale compute that coaching frontier fashions requires. 

“How are you going to construct a foundational mannequin and not using a cluster?” Katanforoosh requested. “I don’t suppose anyone can try this.”

Runway has raised $860 million to this point, together with a $315 million spherical in February from strategic companions like AMD Ventures and Nvidia. That’s roughly in keeping with its most rapid rivals, Luma AI and World Labs, which have raised $900 million and $1.29 billion, respectively, in response to PitchBook. 

However Runway can be going up in opposition to incumbents like OpenAI, which has raised round $175 billion per CEO Sam Altman, and tech behemoth Google, whose guardian firm Alphabet is price $4.86 trillion. Google is Runway’s greatest menace. The corporate’s Veo mannequin competes immediately with Runway’s video-generation enterprise, whereas its Genie world mannequin targets the identical longer-term territory Runway is racing towards. 

Katanforoosh nodded at OpenAI, which shuttered its video platform Sora in March after burning roughly $1 million per day in compute prices with barely $2.1 million in income in response to some estimates. His level: Assets alone don’t assure survival. They don’t assure it for Runway both.

Katanforoosh isn’t writing Runway off. He pointed to AI audio startup ElevenLabs, which has outperformed OpenAI and Google on their very own benchmarks, regardless of missing the assets and pedigree of both. Runway, he argues, may observe the same playbook. 

The comparability isn’t misplaced on Runway’s founders. Valenzuela says the startup’s lack of Bay Space “standardization” offers them an edge. Not solely have they got range of thought, he contends, however with out Silicon Valley ties, they needed to be scrappier, missing the struggle chest lots of their friends have entry to that might have insulated them from the necessity to generate income early.

And in response to Michelle Kwon, Runway’s chief working officer, the corporate isn’t in a rush to lift extra funds, whilst compute calls for enhance with scale.  

“Their background has led them to be early, to be proper as a rule, and to construct a tradition that strikes extremely rapidly,” early investor Michael Dempsey, managing companion at Compound, instructed TechCrunch. 

For Valenzuela, that tradition begins with how he sees the world within the first place. He spends no matter free time he has — not a lot, as a co-CEO and new father — studying books, together with the Chilean poet Nicanor Parra, whom he describes because the antithesis of Pablo Neruda: much less formal, much less educational, holding a view that poetry belongs to the individuals reasonably than to guidelines. 

“Guidelines are simply guidelines they invented,” Valenzuela stated. “That’s a driving pressure of how we do issues at Runway. They are saying Silicon Valley is right here and that’s the place the startups are. Why? These are simply made-up guidelines. Scrub all of them and begin once more.”