Runway has spent the previous seven years constructing visual-generating instruments for the inventive trade. Now, it sees a brand new alternative for its expertise: robotics.
New York-based Runway is thought for its video and photograph technology AI world fashions, or massive language fashions that create a simulated model of the true world. Most not too long ago, the corporate launched Gen-4, its video-generating mannequin, in March and Runway Aleph, its video editing model, in July.
As Runway’s world fashions began to enhance — and get extra life like — the corporate started to obtain inbound curiosity from robotics and self-driving automotive firms wanting to make use of the tech, Anastasis Germanidis, Runway co-founder and CTO, informed TechCrunch in an interview.
“We predict that this means to simulate the world is broadly helpful past leisure, though leisure is an ever rising and large space for us,” Germanidis mentioned. “It makes it far more scalable and price efficient to coach [robotic] insurance policies that work together with the true world whether or not that’s in robotics or in self driving.”
Germanidis mentioned working with robotics and self-driving automotive firms was not one thing Runway initially envisioned when it launched again in 2018. It wasn’t till robotics and different firms in different industries reached out, that the corporate realized their fashions had a lot broader use circumstances than they initially thought, he mentioned.
Robotics firms are utilizing Runway’s tech for coaching simulations, Germanidis mentioned. He added that simply coaching robots and self-driving vehicles in real-world situations is expensive for firms, takes a very long time, and is tough to scale.
Whereas Runway is aware of it isn’t going to exchange real-world coaching by any means, Germanidis mentioned firms can get a whole lot of worth working simulations on Runway’s fashions as a result of they’ve the flexibility to get extremely particular.
Techcrunch occasion
San Francisco
|
October 27-29, 2025
Not like in real-world coaching, utilizing these fashions makes it simpler to check for particular variables and conditions with out altering the rest within the state of affairs, he added.
“You possibly can take a step again after which simulate the impact of various actions,” he mentioned. “If the automotive took this flip over this, or carry out this motion, what would be the final result of that? Creating these rollouts from the identical context, is a extremely tough factor to do within the bodily world, to mainly maintain all the opposite facets of the setting the identical and solely take a look at the impact of the precise motion you need to take.”
Runway isn’t the one firm seeking to sort out this. For example, Nvidia launched the most recent model of its Cosmos world fashions, along with different robotic coaching infrastructure, earlier this month.
The corporate doesn’t anticipate releasing a “utterly separate line of fashions” for its robotics and self-driving automotive clients, Germanidis mentioned. As an alternative, Runway will fine-tune its current fashions to raised serve these industries. The corporate can be constructing a devoted a robotics crew.
Germanidis added that whereas these industries weren’t within the firm’s preliminary pitches to buyers, they’re on board with this enlargement. Runway has raised greater than $500 million from buyers like Nvidia, Google and Basic Atlantic at a $3 billion valuation.
“The best way we consider the corporate, is actually constructed on a precept, somewhat than being in the marketplace,” Germanidis mentioned. “That precept is this concept of simulation, of having the ability to construct a greater and higher illustration of the world. After getting these actually highly effective fashions, then you should use them for all kinds of various markets, a wide range of totally different industries. [The] industries we anticipate are there already, they usually’re going to vary much more on account of the ability of generative fashions.”